ABSTRACT

THE RELATIONSHIP BETWEEN BIOLOGICAL CONDITIONING AND
ANIMAL-ANIMAL INTERACTION IN THE SPATIAL DISTRIBUTION
OF GALLERIA MELLONELLA (LEPIDOPTERA) LARVAE

 

BY

Peter Edward Hogan

The studies in this dissertation deal with the preference

behavior of individual Galleria mellonella (L.) larvae for homotypically

 

conditioned food, its relationship to spatial distribution, and the
effects of degree of conditioning, development (age), and the presence
of conspecifics on these preferences. A mechanism is postulated for
the interaction of these variables with larval activity rhythms in
determining spatial patterns. Also presented is an analytical tech-
nique for the analysis of multiple samples taken from the same indi-
viduals over time.

In Chapter I is presented a review of the literature on
spatial distribution as it specifically relates to invertebrates with
an emphasis on insects and the role of biological conditioning in
spatial distribution. The effects of animal-animal interactions on
preferences for conditioned food are discussed. The studies in this
review are population studies in which individual behavior, in respect

to homotypic conditioning and animal-animal interaction, has not been

thoroughly investigated.

Peter Edward Hogan

Chapter II presents many life history and behavioral consider-

ations directly pertaining to Galleria mellonella larvae, the experi-

 

mental organism for these studies.

Chapter III presents the statistical procedures and the under—
lying assumptions for multivariate analysis of longitudinal data.
Hotelling's One and Two Sample T2 Statistics for Multivariate Analysis
are outlined, illustrating a procedure for treating time as a multi-
dimensional variate. Emphasized are procedures for testing the
assumptions of homogeneous variances and non-serially correlated
data. The major objective is to arm behavioral biologists with the
necessary analytical techniques enabling logical decisions about tests
of significance on repeated individual measurements over time.

Chapter IV tests the hypothesis that preference of individual
Galleria larvae for homotypically conditioned food is a function of
the degree of conditioning and age of the larvae. Larvae were given
a two-choice situation in which one food lump was variously conditioned
and the other was non-conditioned. Individual preference was recorded
every 12 hours for the total developmental period. Depending on the
degree of conditioning, larvae are attracted to conditioned food with
a threshold effect being indicated at high levels of conditioning.
Older larvae demonstrate stronger, initial, attraction to conditioning
than do young larvae, although there is some confounding with early
experience factors, but initial preference is not a good indicator
for long-term preference. There are indications that age differences
in preference may be related to differential activity. Adult Galleria,

on the other hand, avoid highly conditioned food, avoidance being

Peter Edward Hogan

measured by oviposition preferences. The consequences of this behavior
are discussed in relation to larval density.

Chapter V treats the hypothesis that individual larval pref-
erences for conditioned food are a function of the presence of con-
specifics. A resident larva (always an older larva) in a conditioned
food lump has no effect on individual preference for that food lump.
However, if two larvae encounter each other in the open, that is
when neither larva is in a food lump, both preferences for conditioned
food are lowered, as measured by comparisons with an isolate situation.
The preference of the younger larva is most affected initially but
soon increases, possibly as a function of an interaction with the
older larva. A hypothesis was proposed and tested that the behavior
of the younger larva is a function of the initial encounter between
the two larvae, extra conditioning performed by the older larva on the
already conditioned food lump, and a second encounter later in develop-
ment. The effect of the initial encounter is supported by the data,
but the effects of the other two variables remain inconclusive.

Future tests are proposed.

In Chapter VI are summarized the results of these studies.
Individual behavior and distribution of Galleria larvae is a function
of the degree of homotypic conditioning, age of the individual larva,
presence and age of a conspecific, location of the conspecific, and

possibly activity differences between larvae.

 

THE RELATIONSHIP BETWEEN BIOLOGICAL CONDITIONING AND
ANIMAL-ANIMAL INTERACTION IN THE SPATIAL DISTRIBUTION

OF GALLERIA MELLONELLA (LEPIDOPTERA) LARVAE

 

BY

Peter Edward Hogan

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Zoology

1975

I dedicate this dissertation to my wife, Peggy,
whose incalculable patience, understanding, and

help have made all phases of this study possible.

ii

ACKNOWLEDGMENTS

I wish to express my appreciation to Dr. Martin Balaban, my
advisor, for his help and encouragement throughout my graduate career
and for his critical evaluation of my research and this dissertation.
I am also grateful to the members of my guidance committee including
Dr. John A. King, Dr. James H. Asher, and Dr. Stanley C. Ratner for
their constructive comments during this research.

Special thanks go to Dr. James H. Asher for the time spent
helping me with the mathematical analyses in this dissertation and
to Lincoln Gray for his invaluable help in writing the computer pro-
grams used in the analyses. Without their help this dissertation
would not have been possible.

This research was supported in part by NIH Animal Behavior
Training Grant #5 T01 GMOl751-OS BHS and by teaching assistantships

provided by the Department of Zoology.

iii

TABLE OF CONTENTS

LIST OF TABLES . .

LIST OF FIGURES .

CHAPTER
I. BIOLOGICAL CONDITIONING AND SPATIAL DISTRIBUTION .
Introduction .
Literature Review
Spatial Distribution .

Sense Organs .
Food and Spatial Distribution
Environmental Factors in Spatial Distribution

Sex and Genetic Factors in Spatial
Distribution .

Density Factors in Spatial Distribution

Biological Conditioning, Defined .

Physiological Aspects of Biological Conditioning .

Protection From Toxic Substances .
Control of Sex .

Effects on Morphological Changes .
Effects on Growth

Effects on Population Dynamics .

Biological Conditioning and Spatial Distribution .

iv

Page
ix

XV

11

16

19

CHAPTER

II.

III.

IV.

BEHAVIORAL AND LIFE HISTORY CONSIDERATIONS OF THE
GREATER WAX MOTH, GALLERIA MELLONELLA (L.)
(LEPIDOPTERA: GALLERIIDAE) . . . .

 

Behavior and Life History

Spinning Behavior and Preferences

Food Preferences .

Rearing and Maintenance of Stock Cultures

HOTELLING'S T2 STATISTIC FOR MULTIVARIATE ANALYSIS:

A "NEW" APPROACH TO LONGITUDINAL STUDIES AND
SERIALLY CORRELATED DATA .
Introduction .

Hotelling's One Sample T2 Statistic

Hotelling's Two Sample T2 Statistic

The Assumptions Underlying Hotelling's Procedure .

Multivariate Normal Distribution .
Equality of Variance-Covariance Matrices .
Serially Uncorrelated Data .

Summary

BIOLOGICAL CONDITIONING AND SPATIAL DISTRIBUTION IN
GALLERIA MELLONELLA (L.) LARVAE

 

Introduction .

Materials and Methods
Husbandry
Apparatus
Experimental Procedures
Analysis .
Pilot Experiment .

Data Reduction .

Page

65
65
71
75

78

83
83
86
93
99
100
102
105

107

108
108
110
110
111
111
117
127

130

 

 

CHAPTER Page
Results . . . . . . . . . . . . . . . . . . . . . . . . 130

I. Within-Group Comparisons for Randomness of
Larval Response to Conditioning . . . . . . . . . 131

II. Between-Group Comparisons of Larval Responses
to Increasing Degrees of Conditioning . . . . . . 134

A. Response of Test Larvae to Conditioning
Done By Varying the Number of Days of
Conditioning Within Any One Age of
Conditioning Larva . . . . . . . . . . . . . . 134

B. Response of Test Larvae to Conditioning
Done By Varying Ages of Conditioning Larvae
at Constant Days of Conditioning . . . . . . . 147

III. Between-Group Comparisons of the Effects of
Larval Age on Response to Conditioning . . . . . . 158

IV. A Brief Analysis Relating to Initial Pref-
erence to Conditioned Food and the Amount of
Larval Movement as Conditioning Increases . . . . 170
A. Does Larval Initial Preference to Con-
ditioned Food Change as Degree of Con-
ditioning Increases . . . . . . . . . . . . . 170

B. Does Amount of Larval Movement Vary in
Different Amounts of Conditioning? . . . . . . 179

Control for Food Limitation . . . . . . . . . . . . 181

V. Can Response to Conditioning Be Overridden

by Varying the Food Source? . . . . . . . . . 182

VI. Do Adult Galleria Exhibit Preferences to
Conditioning Done by Larvae? . . . . . . . . . 185
Discussion . . . . . . . . . . . . . . . . . . . . . . . 18S

Larval Responses to Increasing Degrees of

Conditioning . . . . . . . . . . . . . . . . . . . 187
Age Differences in Response to Conditioning and
Possible Mechanisms . . . . . . . . . . . . . . . 188
Size of Responding Larvae . . . . . . . . . . . 188
Activity Differences . . . . . . . . . . . . . . 189

vi

 

CHAPTER
Sensory Differences Between Ages .
Early Experience Factors .

Response to Biological Conditioning as a
Bioassay of Behavior .

Possible Functions of Biological Conditioning
Protection .

Biological Conditioning as a Preferred
Consistency

Mobility .
A Signal for the Presence of Other Larvae
Biological Conditioning and Feeding Behavior .
Biological Conditioning and Population Density .
V. THE RELATIONSHIP BETWEEN BIOLOGICAL CONDITIONING AND

ANIMAL-ANIMAL INTERACTION IN THE SPATIAL DISTRIBUTION
OF GALLERIA MELLONELLA (L.) LARVAE .

 

Introduction .

Materials and Methods
Husbandry
Apparatus
Experimental Procedures
Analysis .

Results

Experiment I: Do Galleria Larvae Group in the
Absence of Initial Conditioning and Is Their
Distribution in Such a Situation a Function
of the Presence of a Conspecific? . . . . . .

Experiment 11: When Initial Conditioning is
Present, Is Preference for Conditioning
Affected by the Presence of a Conspecific,
the Age of the Conspecific, and Whether
‘the Conspecific is a Resident in the
Conditioned Food?

vii

Page
191

192

193
196

196

197
197
198
198

200

201
201
204
204
204
204
208

209

231

(CHAPTER

Experiment III: Are the Lowered Responses of
Younger, 7 Day Old Larvae, When a 13 Day Old
Larva is Simultaneously Introduced, a Result
of the Changing Biological Conditioning,
Interaction With the 13 Day Old Larva, or
Both? . . . . . . . . . . . . .
Discussion .

Individual Responses to a Conspecific and
Biological Conditioning

Possible Mechanisms for the Interaction of
Biological Conditioning and Larval-Larval
Interaction in the Spatial Distribution
of Galleria Larvae .

VI. SUMMARY
Isolate Behavior .
Interactions Between Conspecifics and Biological

Conditioning .

BIBLIOGRAPHY

viii

 

Page

255

274

276

281
290

290

292

295

 

Table

3.1.

4.2.

4.3.

4.4.

4.5.

LIST OF TABLES

Artificial dietary medium for laboratory rearing of
the Greater Wax Moth, Galleria mellonella (L.)

Components of the Two Sample T2 Statistic

Larval weights corresponding to the larvae in
Figure 4.1. Weights were calculated for 500
randomly selected larvae of a particular age
as well as for the larvae of a particular age
actually used for experimentation

Twelve experimental groups in Figure 4.2 were
randomly selected and their AM and PM mean
vectors were compared using Hotelling's One
(') and Two (*) Sample Analyses . . . . .

Results of Hotelling's One Sample comparison of
each group from Figure 4.2 with an expected
vector whose entries are all 0.5 to determine
which groups are random. Mean vectors (X),
estimated variance vectors (82), F-values,
and D-values for serial correlation are given

Results of Hotelling's One (') and Two (*) Sample
comparisons of the response of 7 day old test
larvae to varying amounts of conditioning (1,3,
5, and 7 days) done by 6, 9, and 12 day old
larvae . . . . . . . . . . . . . .

Results of Hotelling's One (') and Two (*) Sample
comparisons of the response of 11 day old test
larvae to varying amounts of conditioning
(1, 3, 5, and 7 days) done by 6, 9, and 12 day
old larvae . . . . . . . . . . .

Results of Hotelling's One (') and Two (*) Sample
comparisons of the response of 13 day old test
larvae to varying amounts of conditioning (1,3,
5, and 7 days) done by 6, 9, and 12 day old
larvae . . . . . . . . . . . . . .

ix

Page

80

95

112

118

123

136

139

142

 

Table

4.7.

4.9.

4.10.

4.11.

4.12.

4.13.

4.14.

Results of Hotelling's One (') and Two (*) Sample
comparisons of the response of 7 day old test
larvae to conditioning done by 6, 9, and 12 day
old larvae as days of conditioning are held con-
stant at l, 3, 5, and 7 days .

Results of Hotelling's One (') and Two (*) Sample
comparisons of the response of 11 day old test
larvae to conditioning done by 6, 9, and 12 day
old larvae as days of conditioning are held con-
stant at l, 3, 5, and 7 days .

Results of Hotelling's One (') and Two (*) Sample
comparisons of the response of 13 day old test
larvae to conditioning done by 6, 9, and 12 day
old larvae as days of conditioning are held con-
stant at l, 3, 5, or 7 days

Results of Hotelling's One (') and Two (*) Sample
comparisons for age differences in response to
conditioning. The responses of test larvae (7,
11, and 13 days old) to l, 3, S, or 7 days of
conditioning done by a 6 day old larva are
tested . .

Results of Hotelling's One (') and Two (*) Sample
comparisons for age differences in response to
conditioning. The responses of test larvae (7,
11, and 13 days old) to l, 3, S, or 7 days of
conditioning by a 9 day old larva are tested .

Results of Hotelling's One (') and Two (*) Sample
comparisons for age differences in response to
conditioning. The responses of test larvae (7,
11, and 13 days old) to l, 3, 5, and 7 days of
conditioning by a 12 day Old larva are tested

Results of Hotelling's One (') and Two (*) Sample
tests on comparable 4 day segments of the
response vectors shown to exhibit age differ-
ences by a Two-Sample comparison in Tables 4.10,
4.11, and 4.12 .

Hotelling's One (') and Two (*) Sample comparisons
of randomly selected experimental groups from
Figure 4.2 with their replications as a test
of whether behavior is changing over time and
from hatch to hatch

Page

148

151

154

159

162

165

169

172

Table

4.15.

4.16.

4.17.

4.18.

4.19.

4.20.

5.1.

5.2.

5.3.

Chi-square analysis of initial larval preference
to the extremes of conditioning within the
three test ages .

Results of a Chi-square goodness of fit analysis
on the number of times larvae moved between
conditioned and plain food lumps when conditioning
is either low or high . . .

Analysis of variance on developmental times of 7,
11, and 13 day old larvae reared on single food
lumps which were either maximally conditioned
or non-conditioned .

Analysis of variance on 11 day old larvae reared
on normal and deficient diets . . . .

A 2 x 2 contingency table analyzed by chi-square
of larval responses to normal-conditioned and
deficient-conditioned food .

Chi-square goodness of fit test on female adult
preference for egg-laying in low and high
larval conditioned food

Results of Hotelling's One (') and Two (*) Sample
comparisons of experimental groups (P) versus
the replications (R) from Figure 5.1. Com-
parisons are made on the basis of total number
of pairs, regardless of which food lump they
were in

Results of Hotelling's One Sample comparison of
each experimental (P) group in Figure 5.1 with
an expected vector whose entries are all 0.5
to see if grouping (pairing), regardless of
food lump, is different from random in the
absence of homgtypic conditioning. Response
mean vectors (X), estimated variance vectors
(8 ), and D-values for serial correlation
are found in Table 5.1 .

Results of Hotelling's One Sample comparison of
pairs, regardless of food lump, versus an
expected vector whose entries are all 0.5.
These comparisons are shown for the first
four days and the last four days of each
experimental group to see if time in the
experimental situation affects grouping
behavior .

xi

Page

178

180

182

183

184

186

213

218

219

Table Page

5.4. Results of Hotelling's One (') and Two (*) Sample
comparisons among the experimental groups in
Figure 5.1. Comparisons are made for the first
4 days and the last 4 days to see if time in the
experimental situation affects grouping, and all
comparisons are for pairs of larvae regardless
of food lump. For the One Sample (') Test the
groups used for the expected vectors are in the
column headed EV. and those whose covariance
matrix was used are in the column headed CV . . . . . . 221

5.5. Contingency tables with Chi-square Analysis for
each group in Figure 5.1 with the total Chi-
square partitioned to show the contribution of
each time point for each food lump to the total.
Within each box are shown the total number of
pairs of larvae for each food lump at each time
point and their individual Chi-square values. Also
shown are the total pairs at each time point or
food lump and their Chi-square contribution and the
total Chi-square value over all time points and
both food lumps . . . . . . . . . . . . . . . . . . . . 224

5.6. Results of Hotelling's One Sample analysis on the
preference of the younger larva of a pair for the
food lump in which the older larva initially
selected and remained in for the entire experi-
mental period. Comparison is made against an
expected vector whose entries are all 0.5 . . . . . . . 227

5.7. Analysis of Variance on developmental times of two
7, two 11, and two 13 day old larvae reared on
single conditioned or plain food lumps . . . . . . . . . 230

5.8. This table presents the response mean vectors (X),
estimated variance vectors (82), and D-values for
serial correlation for the groups in Figures 5.3
and 5.4 to be used in future comparisons . . . . . . . . 234

5.9. Hotelling's One Sample test on the experimental groups
in Figures 5.4 and 5.5. Total larvae in the con-
ditioned food (paired, singly, or none) is compared
with an expected vector whose entries are all 0.5.
Response mean vectors (X), estimated variance vectors

(82), and D-values for serial correlation are found
in Table 5.8 . . . . . . . . . . . . . . . . . . . . . . 236

xii

Table

5.10. Hotelling's One Sample test on the experimental
groups in Figures 5.4 and 5.5. Pairs of larvae
on conditioned food are compared with an
expected vector whose entries are all 0.25.
Response mean vectors (X), estimated variance
vectors (S ), and D-values for serial corre-
lation are found in Table 5.8

5.11. Hotelling's One (') and Two (*) Sample comparisons
among the groups in Figure 5.4. Assuming no
animal—animal interaction, the response
vectors for two larvae of equal age should be
the same as the response of one larva of the
same age. Also the response vector for either
larvae, when the two test larvae are unequal in
age, should be the same as the response of one
larvae of the appropriate age

5.12. Hotelling's One (') and Two (*) Sample comparisons
among the experimental groups in Figures 5.4 and
5.5 to determine resident versus non-resident effects
on preference for biological conditioning. Assuming
no animal-animal interactions there should not be
any differences in these comparisons. There should
also not be any differences if the effect of a
resident larva on the response larva is the same
as the effect of another larva introduced simul-
taneously with the response larva

5.13. This table presents the response mean vectors (X),
estimated variance vectors ($2), and D-values for
serial correlation for the groups to be analyzed
in Table 5.14

5.14. Hotelling's One (') and Two (*) Sample comparisons
of the 7 day old mean response vectors when paired
with an 11, 13, or 15 day old larva or when an
isolate, all with initial high conditioning on
one food lump

5.15. Hotelling's One Sample (') comparison of the first
3 days of the 7 day old response vectors when with
various aged conspecifics and when isolated, with
an expected vector whose entries are all 0.5 .

xiii

Page

238

240

248

258

261

263

Table

5.16.

5.17.

5.18.

5.19.

5.20.

Table of mean response vectors (X), estimated
variance vectors (52), and D-values for serial
correlation of the 7 day old larvae from the
experimental groups being tested for the
effects of initial encounter, extra conditioning,
and the second encounter with a 13 day old con-
specific .

Results of Hotelling's One (') and Two (*) Sample

comparisons among the 7 day old response vectors
in Table 5.16

Results of Hotelling's One (') and Two (*) Sample
comparisons among the 7 day old response vectors
in Table 5.16 for the last 3 days of each mean
response vector . . . . . .

Results of Hotelling's One Sample comparison of
the first 5 data points of each group from
Table 5.16 with an expected vector whose
entries are all 0.5

Results of Hotelling's One (') and Two (*) Sample

comparisons of various groups replicated through-
out this chapter .

xiv

Page

266

267

271

273

275

LIST OF FIGURES

Hotelling's One Sample T2 Statistic
Hotelling's Two Sample T2 Statistic

From right to left, these are 6, 7, 9, ll, 12, 13,
and 15 day old larvae used for conditioning and
testing preference for conditioning. Note the
size distribution of the larvae

Experimental design, showing the identification
labels and the date when each experimental (B)
group and control (C) group was run

Pilot experiment to see if Galleria larvae response
to biologically conditioned food. Note that
after 16 days of age the response begins declining
during which time the larvae are apparently not
exhibiting a preference for either conditioned
or non-conditioned food

Mean response vectors of 7 day old test larvae to
varying amounts of conditioning (1, 3, 5, and 7
days) done by 6, 9, and 12 day old larvae

Mean response vectors of 11 day old test larvae
to varying amounts of conditioning (l, 3, 5, and
7 days) done by 6, 9, and 12 day old larvae

Mean response vectors of 13 day old test larvae
to varying amounts of conditioning (l, 3, 5, and
7 days) done by 6, 9, and 12 day old larvae

Mean response vectors of 7 day old test larvae to
conditioning done by 6, 9, and 12 day old larvae
as days of conditioning are held constant at 1,
3, 5, or 7 days

Mean response vectors of 11 day Old test larvae to
conditioning done by 6, 9, and 12 day old larvae
as days of conditioning are held constant at l,
3, 5, or 7 days

XV

Page
87

94

112

113

128

138

141

144

150

153

Figure

4.9.

Mean response vectors of 13 day old test larvae to
conditioning done by 6, 9, and 12 day old larvae
as days of conditioning are held constant at l,
3, 5, or 7 days

. Mean response vectors of 7, 11, and 13 day old test

larvae to l, 3, 5, or 7 days of conditioning done
by a 6 day old larva .

. Mean response vectors of 7, 11, and 13 day old test

larvae to l, 3, 5, or 7 days of conditioning done
by a 9 day old larva .

. Mean response vectors of 7, 11, and 13 day old test

larvae to l, 3, 5, or 7 days of conditioning done
by a 12 day old test larva .

. Replications of randomly selected groups from the

experimental design in Figure 4.2

Experimental design for asking whether larvae will
group (pair) in the absence of homotypic
conditioning .

Mean response vectors for percent pairs of larvae
on any food lump when the food lumps were initially
unconditioned

Mean response vectors of the younger larvae in a
pair to a particular "plain" food lump that was
initially selected by the older member of the
pair and in which the older larva remained in for
the entire experimental period. That food lump is
considered the "conditioned" lump for purposes of
these curves .

Experimental design (top) for animal-animal inter-
actions, at two levels of conditioning, when the
conditioning larva is removed and both test larvae
are new to the experimental situation

Experimental design (left) for the effect of a resi-
dent larva on the preference of a test larva for
biological conditioning. The resident larva is
the original conditioning larva

xvi

Page

156

161

164

167

171

211

217

228

Figure

5.6.

5.7.

5.8.

5.9.

5.10.

5.11.

5.12.

Mean response vectors for comparisons (Table 5.11)
of two larvae situations with one larva situ-
ations to find if individual preference for
low conditioned food is affected by the presence
of a conspecific .

Mean response vectors for comparisons (Table 5.11)
of two larvae situations with one larva situations
to find if individual preference for high con-
ditioned food is affected by the presence of
a conspecific . . . . . . . . .

Mean response vectors for comparisons (Table 5.12)
of resident versus non-resident effects on larval
preference for low conditioning

Mean response vectors for comparisons (Table 5.12)
of resident versus non-resident effects on larval
preference for high conditioning .

Mean response vectors of a 7 day old larva and a 13
day old larva with the wandering phase of the 13
day old larva included. These are from experi-
mental group 64PH(l-5)7+13 .

Mean response vectors of 7 day old larvae to high
conditioning when a 15, 13, or 11 day old con-
specific is also present. The mean response
vectors for the 15, 13, and 11 day old larvae
are shown, with their wandering phases, for
comparative purposes .

Response mean vectors for the 7 day old larvae

from all experimental groups in Table 5.16 and
5.17 .

xvii

Page

243

246

251

253

256

260

269

CHAPTER I

BIOLOGICAL CONDITIONING AND SPATIAL DISTRIBUTION

Introduction

 

Biological conditioning has been defined (Allee 1934) as
". . changes produced in the medium by organisms living therein and
the effects of such changes upon other organisms . . . ." Changes in
the medium may be chemical or physical and either of homotypic or
heterotypic origin. Homotypic conditioning is produced by a con-
specific, whereas conditioning emanating from other than a conspecific
is referred to as heterotypic. As organisms inhabit their environment
they modify it, often to such an extent that whole populations are
affected. With invertebrates, notably Lepidoptera, a few studies have
related biological conditioning to spatial patterns, although most of
the literature treats conditioning in relation to its physiological
effects upon other organisms and populations. Since biological con-
ditioning was considered a population phenomenon, these studies looked
at whole populations, making no attempt to ask questions about how
biological conditioning affects individual behavior.

The research for this dissertation deals with an analysis of
spatial distribution (aggregation and/or dispersal) in the wax moth

larvae, Galleria mellonella (Lepidoptera), when maintained in the

 

laboratory. In a simplistic sense, aggregation may be viewed as the

1

result of those factors that "pull" organisms together, and dispersal
as the result of those factors that "push" them apart. Spatial dis-
tribution (dispersion) would be the result of the interactions between
these "pulls" and "pushes." Relationships between biological con-
ditioning, animal interactions, and development (age) are examined.
The basic approach is to analyze spatial distribution by isolating

the critical variables in simple systems of individuals or small
groups. Concomitant with this biological approach is the development
of an analytical model for spatial distribution.

Galleria mellonella was chosen for these studies because of its

 

suitability for behavioral, distributional, and populational experi-
mentation, and because the larvae produce a pronounced physical (and
possibly chemical) conditioning of their environment.

Galleria larvae are vagile, asocial, asexual, short-lived and
herbivorous. Naylor (1959) pointed out that these features allow the
analysis of behavioral and spatial relationships. Vagility is the
ability to actively move from place to place. Asociality is here
defined as the ability of an individual to survive and/or reproduce in
the absence of structured groups. Galleria adults are extremely
fecund, enabling generation of large sample sizes for experimentation.
The implication of being short-lived is that Galleria larvae must
quickly respond to cues. Herbivorous species usually find food and
shelter in a relatively fixed location, and Galleria larvae are not
normally found ranging over vast expanses of territory for their life
needs. The adults lack mouth parts, do not feed, and are not usually

found in close association with the larvae.

A consequence of the preceding traits is that the behavior of
Galleria is a relatively "simple" system for elucidating the critical
variables relating to spatial distribution. Several variables present
in other systems, such as sexual relationships and adult-larval inter-
actions, are not of prime importance.

The following general hypotheses were examined in this dis—
sertation:

l. The spatial distribution of Galleria larvae is a function of
their preferences for homotypic conditioning.

2. Preferences for biological conditioning depend upon the degree
of conditioning and the age of the responding larvae.

3. Animal-animal interactions (presence of conspecifics) alter
preferences for conditioning, and the degree to which prefer-
ences are altered is a function of age and prior residence in

the conditioned situation.

Literature Review

 

§patial Distribution

Dispersion is the spatial structure of a population, as opposed
t0 dispersal which is the dynamic process involved (Surtees 1963a,
1964C). Lack's (1954) definition is more rigorous. He defines dis-
PerSion as a non-random type of distribution, with dispersal being
the movement of young animals from their place of birth. Dispersal is
normally thought to consist of (l) organisms leaving their natal site,
(2) Crossing some sort of barrier, and (3) settling and breeding in a
new place. These criteria have strong genetic implications. However,

for the purposes of my studies, dispersal is being defined as the

process by which an organism or group of organisms move away from some
original site, whether or not breeding occurs in the new locality. In
this context dispersal is the opposite of aggregation, and is merely
one side of a broader concept, dispersion. Wynne-Edwards (1962)
defines dispersion as the placement of individuals and groups of indi-
viduals within the habitats they occupy, and the processes by which
this is brought about. A general definition, such as this, encompasses
other than non-random groupings.

Dispersion, contagious distributions, aggregations, groups,
bunches, clusters, or whatever label one wishes to give such phenom-
enon, are Of widespread occurrence in nature, and are found in varying
degrees of complexity throughout the animal phyla. They range from
chance assemblages to the highly organized social insects and man.
Natural aggregations may result from the response of individuals to
the presence of conspecifics or from the independent response of each
individual to an external stimulus (Bowen 1931), the former often being
called social and the latter asocial. Allee (1927), however, noted
that once established, aggregations may last for extended periods of
time merely due to a lack of disruptive stimuli; whereas dispersal may
be the result of a lack of cohesive stimuli. No social instinct need
be evoked to account for the formation or maintenance of aggregations.
Wynne-Edwards (1962), on the other hand, states that the maintenance
of the spatial organization of animal populations can be attributed to
behavioral interactions among its members. An implication of this

statement is that whatever the causative factors of aggregation are,

individuals would exhibit different distributions if other conspecifics
were absent.

The critical variables being considered in my studies are bio-
logical conditioning and animal-animal interaction. Other causative
factors in spatial patterns are frequently encountered in the litera-
ture, such as the role of the sense organs, food, and various environ-
mental cues. In addition, the mobility of organisms is important and
this may depend on individual variables such as sexual and genetic dif—
ferences. These factors must be either controlled or considered as
variables. It is not the intent of this review to discuss each of
these at length. However, since such variables can affect preferences
for biological conditioning as well as animal-animal interactions, and
since they form the framework for much of the spatial distribution

literature, they will be briefly discussed.

Sense Organs

Sense organs serve as the passageway by which all sensations
from the environment reach the central nervous system and elicit some
kind of response. Auditory, visual, olfactory, tactual, and electrical
signals may all function in dispersionary as well as social behavior.
(See Wynne-Edwards 1962 for a review of this topic.)

Two basic approaches might be used to study the effects of
biological conditioning on spatial distribution. One is to find out
what constitutes conditioning and which senses are employed in respond-
ing to it. The second approach is to demonstrate a response to con-
ditioning, regardless of the senses employed or the characteristics of

conditioning, and begin asking questions about the results, function,

and dynamics of the responses. I have opted for the latter approach
in my studies, although the first approach must eventually be studied.
To study the relationship between biological conditioning and spatial
distribution it is not necessary to know what conditioning consists of
or what senses are used to respond. It is, however, necessary to
demonstrate that conditioning is occurring, that it emanates from the
organism in question, and that the organism responds to it. The
degree of conditioning can be controlled in terms of time spent con-

ditioning and age of the organism doing the conditioning.

Food and Spatial Distribution

Most animals, through their power of movement, play a part in
determining their own population densities. They are able to concen-
trate at places where food and other resources are plentiful and to
Javoid unfavorable situations. According to Wynne-Edwards (1962), food
is always the critical or limiting resource which dictates how high a
population density can be supported in any particular habitat.
Regardless of food supply, however, other variables such as animal-
animal interactions and biological conditioning play a major role in
distribution.

Wynne-Edwards (1962), using some of Jesperson's (1924) data
noted that a high correlation exists between plankton density and the
density of pelagic birds. He concluded that the birds disperse them-
selves on the basis of available food supply. A

Territoriality, a form of dispersion, in birds may be food

limited. Howard (1920) found that territorial size may be reduced in

the presence of abundant food, and may increase in size when food

 

 

> s.

 

becomes limited. The relationship between food, dispersion, and
numbers in a population was discussed by Lack (1954) who concluded that
many bird species are limited in numbers, and in resultant distribution,
by their food supply.

Similar data is available for other Species as well. Calhoun
(1948, 1952) with rats, and Brown (1953) and Strecker (1954) with
house mice, have demonstrated that location of food may be a factor in
influencing spatial distribution under semi-natural conditions. Cal-
houn's (1950) study also shows that population size and competition
are important in determining spatial distribution. However, the
results of such experiments indicate that being near food enables
elimination of other individuals from the area. This is an indication
that animal-animal interaction, not food supply, is the proximal cause
for their distribution.

Food has often been shown to be a variable in the spatial dis-
tribution of invertebrates. Loschiavo (1952) has shown that Tribolium
confusum exhibit strong food preferences which in turn affect its dis-
tribution. In fact, flour is not the preferred medium of the flour

beetle, if a more nutritional medium is available. Galleria mellonella

 

larvae also utilize foods differentially as well as exhibiting food
preferences (Chase 1921; Milum and Geuther 1935; Whitcomb 1946; Beck
1960).

Crowding, often related to food, may affect an organism's
physiology as well as its distribution. In a study on crowding of the

larvae of the meal moth, Ephestia kuhniella, Smith (1969) found that

 

larvae infesting food supplies in an uncrowded state grow faster and

larger than those in crowded substrates. This is rather a common
observation among other species as well, and has been observed in our
laboratory with Galleria larvae. However, Smith did not determine if
stunted growth was due to crowding, food limitation, or excessive con-
ditioning of the medium. Smith also observed that less crowded cul-
tures of Ephestia reproduce faster and probably more plentifully.

Rose (1959) discovered similar results in guppies, as have Sang (1949)

in Drosophila melanogaster, Park (1938) in Tribolium confusum, Brown

 

 

(1946) in trout fry, and Rose (1960) in tadpoles. Smith postulated
that this faster and more plentiful reproduction maximized population
size in the next generation, ensuring maximal use of a food supply.
However, none of these researchers made any attempt to separate the
effects of crowding and food supply from the effects of biological con-
ditioning in crowded cultures.

Smith (1969) also noted a behavioral response of Ephestia
adults in crowded cultures, the adults being more active than isolates.
Such activity may promote the colonization of suitable surrounding
food supplies as they become available and as the old supply becomes
crowded.

A behavioral response of organisms to some aspect of their
food supply may affect their distribution. Butterfly larvae of

Mechantis isthmia avoid the trichomes on their Solanum plant host by

 

spinning a network of silk over the tops of the spines (Rathcke and
Poole 1975). The larvae can then crawl over the tops of the spines
cn1 their "silken scaffolding" and safely feed on the unprotected edges

of the leaves. Most larvae of this genus are solitary feeders, but

larvae of Mechantis isthmia generally feed in small groups of from

 

four to six individuals per leaf, spinning and sharing the webbing
together.

Rathcke and Poole (1975) advanced the hypothesis that such a
silken bridgework is energetically feasible only if several larvae pool
their resources and share the benefits. The ability to spin silk is
widespread among the Lepidopterans, as well as in other orders. Most
larvae spin silk to secure themselves while molting or to attach or
protect their pupa. Therefore, evolution of a feeding web would entail
only an elaboration of an already present ability. However, the
increased silk production could necessitate a change in the social
behavior from solitary to gregarious. Such an hypothesis might also
apply to Galleria larvae if Paddock's hypothesis (1913) that silk
tunnels function to hold the damage combs together after feeding is
tenable.

There are also situations in which food is not limited and the
distribution of a population of organisms is a function of how they
behaviorally exploit a preferred food supply. Aphids feed on the
phloem cells of plants, imbibing soluble nutrients such as sugars and
amino acids. Observations by Way and Cammell (1970) indicated that
aphids normally aggregate in groups of about twenty, and that a rela-
tionship might exist between these aggregations and food utilization.
'They discovered that aggregates of aphids act as "sinks" diverting
nutrients from distant plant parts, and that isolates could not as
«Effectively feed. However, as these aggregates enlarged, population

grownfiuwas halted as a result of decreased natality. Therefore, aphids

10

with spaced out populations, or in small aggregations, do little harm
to a plant and maintain relatively stable populations. Way and Cammell
further noted that different species exhibited behavioral differences
in exploitation of host plants, with a resultant difference in aggre-
gation behavior. Those species normally exploiting a host plant
piecemeal remained closely aggregated, i.e., the cabbage and bean
aphids, whereas those species normally doing little harm to a host
plant remain spaced out in tiny aggregations, i.e., peach and potato
aphids.

There are innumerable examples of the effects of food on
spatial patterns, and food is probably the ultimate factor in any long
term dispersionary study. The main reason for discussing it, other
than elucidating it as an important variable, is that many studies in
spatial distribution failed to account for limited food supplies.
Several studies in this review failed to do so, and, as such, require
cautious interpretation. For example, the results of Surtee's studies
(see "Environmental Factors in Spatial Distribution") are interpretable
on the basis of upward movement away from used up food supplies, rather
than on the basis of Surtee's hypothesis of upward dispersal under
high density as a kinesis. Similar problems arise in Park's work with
Tribolium, and Park admitted that the food limitation hypothesis needs
testing before his physiological and distributional data can be veri-
fied. However, all of my studies described in this dissertation con-

trolled for limited food source as a dispersionary agent.

11

Environmental Factors in Spatial
Distribution

It is well documented that organisms respond to environmental
cues, such as temperature, light, and various structural parameters of
the environment, and that such responses, often associated with physio-
logical tolerances, may affect spatial patterns. Selected examples
from vertebrates and invertebrates illustrate this point.

Studies in habitat selection (Harris 1952; Wecker 1963; Fitch
1974) have demonstrated that mice prefer certain habitats, which may
be related to environmental, genetic, and early experience variables.
Habitat characteristics, such as fallen trees and holes in the ground,
have been shown to affect the spatial distribution of small mammals.

Brand (1955, from Terman 1961), using Peromyscus leucopus, showed a

 

direct seasonal relationship between population distribution and tree
density, density of fallen trees, and degree of slope of fallen trees.
The significant factors in this relationship are probably potential
nest sites and food. The number of rats in city blocks was temporarily
decreased by removing harborage sites (Orgain and Schein 1953), and
Davis (1958) changed the spatial distribution in laboratory populations
of house mice by introducing baffles and additional nest sites.
Moisture, ground depressions, hard textured logs, plant cover,
and high relative humidity all positively influence the movements of
the salamander, Plethodon cinereus (Heatwole 1962). Most of these
variables are directly related to the salamander's physiological
limits. Varying spatial patterns in relation to environmental
Stimuli and physiological tolerances have also been documented in the

ant Formica (Talbot 1934), termites Reticulitermes hesperus and

12

R, tibialis (Williams 1934), grasshopper Melanoplus (Parker 1930),

 

salamander Plethodon (Shelford 1913), and wireworm populations (Salt
and Hollick 1946; Thorpe e£_al. 1946). In some instances such varying
distributions are seasonally correlated.

Often spatial distribution is not so much a result of environ-
mental factors as it is a result of an organism's mobility limits.
For example, Bovbjerg (1952) investigated the aggregation behavior of

the aquatic snail Campeloma decisum and found that these aggregations

 

were a result of an upstream dispersal, probably a rheotaxis, coupled

with an inability to traverse even slight blockages of the streams.
Abiotic variables, such as light, temperature, and humidity,

are the most commonly investigated variables in spatial distribution.

In the spruce budworm (Choristoneura fumiferana), for example,

 

Wellington (1948) demonstrated that young stages aggregated as a result
of being photopositive. Older larvae exhibit a reversal in this ori-
entation. Phototaxis is also the predominant factor in groupings of

the sea urchins Arbacia punctulata and Lythechinus variegatus (Sharp

 

 

and Gray 1962). Arbacia is negatively phototactic. Lythechinus is

 

positively phototactic to artificial light and negatively phototactic

to sunlight. Lythechinus also exhibits an interesting "heaping"

 

'behavior as a response to light which consists of picking up shells
and seaweed and covering its body with them.

There is considerable literature on the spatial patterns in
iSOpods in relation to abiotic factors. Warburg (1964) compared
several isopod species, finding aggregation differences in relation

tx: light, temperature, and humidity. Previously Allee (1926) and

1.

 

13

Cole (1946) had shown that temperature and dryness affect the "bunch-
ing" phenomenon in isopods. Warburg, however, tested several species
and discovered species as well as mechanism differences in response.

In the grain beetle (Sitophilus gganarius), vertical distribution

 

depends upon temperature, humidity, and density of organisms (Surtees
1963a). At low and high temperatures the number of weevils at the
surface increases as a function of density. Increased moisture
decreases surface aggregations. In a study on the saw-toothed grain

beetle (Oryzaephilus surinamensis), Surtees (1963a) discovered the

 

same kind of relationship between density and moisture, but species
differences were observed. Species comparisons of vertical aggregation
behavior in relation to moisture, temperature, and density were there-
fore performed by Surtees (1964e). The general conclusion from these
studies is that accumulation of adults occurs under those conditions
where kinesis becomes minimal. Reduction of individual kinesis levels
has a secondary effect within groups of adults in that it depresses
mutual disturbance which would otherwise elicit dispersal. Accumu-
lation within the physical conditions of temperature and moisture is
therefore affected by changes in orthokinesis and klinokinesis. In
all species examined, specific rates of turning and levels of speed
‘under various conditions of temperature, moisture, and density were
:found. Aggregation would be expected in regions of lowest velocity
and fewest turns. These conditions prevail under the proper con-
ciitions of temperature and humidity. Increasing density has the
Opposing effect of dispersing the weevils away from such aggregations.

However, the effects of biological conditioning and limited food

14

reserves were not controlled in these experiments and both are viable
alternative hypotheses to Surtees' explanation for dispersal.

Surtees (1964f) further discovered that aggregation behavior is
affected by the beetle's early experience with humidity differences.
Beetles reared at 70% R.H. did not accumulate in damp grain, whereas
those reared at 40% R.H. for 14 days did. Humidity-conditioned adults
moved slowly in damp grain, unconditioned adults moved rapidly. These
results add further support to his hypothesis that accumulation in one
area must be seen as the joint result of a reduction in individual
movement and a consequent depression of intra-group disturbance, which
would otherwise elicit dispersal. A similar result was found by

Graham (1958) with Tribolium confusum and I, castaneum. Both species

 

exhibit temperature preferences which affect their dispersal. Early
experience, however, alters these preferences.

Dispersion behavior is often interpreted in light of the
variable being examined, when in fact some other, possibly uncon-
trolled, variable is the actual mechanism. The woodlice literature
is a case in point. Woodlice live in places of high moisture content,
i.e., beneath bark, under rocks and fallen trees, or beneath leaves,
and humidity had been accepted as the controlling variable in their
habitat choice and degree of activity. However, since crevices are
Idark, as well as damp, phototactic and hygrokinetic behaviors were
Ixrobably linked (Edney 1954). It had also been assumed but never
‘tested, by Allee (1926), that woodlice react to contact stimuli, since
(:revices are confining and when crevices are not available they climb

(”1 each other and form aggregations. Friedlander (1964) attempted to

15

sort out the responses of the woodlice Oniscus ascellus, Porcellio

scabes, and Armadillidium vulgare and how they relate to aggregation

 

behavior. His particular interest was with thigmokinetic behavior, or
the response to contact. The results indicate that although the effects
of thigmokinetic reactions are similar to humidity reactions, they are
distinct from them. Such responses also varied with species and

degree of humidity, being most marked at low humidity. The thigmo-
kinetic responses depend upon area and roughness of the contacted
surface, and increases in either results in reduced locomotion.

It would appear that woodlice are drawn to crevices by humidity
responses, and that if the crevice provides the proper thigmokinetic
stimuli, movement is slowed and aggregations form.' However, Fried-
lander did not perform experiments on thigmokinetic behavior as a
result of touching other individuals so attracted to a crevice. It is
reasonable to assume that as the number of individuals increases in a
crevice, thigmokinetic responses to these individuals may override the
tendency to slower movement and lead to dispersal. This is very much
what happens in grain beetle populations, as seen earlier.

It is obvious that as environmental factors vary, so does
spatial distribution. In all of my experiments, however, light,
temperature, and humidity were held constant in order not to be
involved in any spatial distribution results with Galleria populations.
'This was necessary for several reasons, one of which is that there
armaindications that abiotic variables affect biological conditioning.
fhxr example, low temperatures decrease spinning activity and spinning

aurtivity may be the main component of conditioning in Galleria larvae.

16
Sex and Genetic Factors in
Spatial Distribution
The mobility of an organism has important consequences in
relation to the ability to aggregate or disperse. Mobility may depend
upon the individual variables of sex and genetics. McDonald (1968),

using Tribolium confusum, found that males more rapidly disperse than

 

females and that there are strain differences in mobility. Strain
differences may be due to differential response to repellent sub-
stances released by the organisms or to avoiding other adults. This
was not tested. However, he did find that the mobility of males is
lessened in the presence of a female. This supports Naylor's (1959)

finding that male Tribolium confusum are attracted to females and

 

exhibit a different distribution pattern than if females are absent
from the population.

Since genetic differences in mobility and dispersal do exist,
it would be helpful to measure the amount of genetic variation popu-
lations contain for such traits (or at least control them in experi-
mentation) and to explore the relationship between mobility and other
fitness components such as fertility and fecundity. In Tribolium,
certain behavioral traits are easily studied and modifiable by arti-
ficial selection. Dawson (1964) found that individuals of highly

inbred lines of Tribolium confusum spend a greater percent of time on

 

‘the surface of the flour medium than the normal wild type. Sokoloff
(1966) extended this finding to mutant strains and demonstrated that
‘therresponse was maintained independently of population density, in

(nilture containers. The general activity levels of wild type and

certain body color mutants were compared by McDonald and Fitting (1965)

17

who found the latter to be more active. Using Tribolium confusum and

 

I, castaneum, Lerner and Inouze (1968) were able to achieve rapid
response to selection for speed of running hierarchical T-mazes.

Artificial selection has demonstrated that mobility may be
altered, but the exact mechanism is as yet unclear. It is possible
that what is being selected for is response to some cue rather than
increased mobility. Ogden (1969) experimented with some olfactory
cues which mediate dispersal behavior in Tribolium. He attempted
(1970b) to determine dispersal behavior, defined as emigratory movement
by walking in a laboratory apparatus. The apparatus utilized was
that originally designed by Prus (1963), consisting of two shell vials
connected with a U-tube through which a thread was strung. The appa-
ratus was so designed as to allow only one-way movement through the
tube. Ogden tested fourth generation beetles as to their rate of
dispersal from homotypically conditioned medium and fresh flour. In
I, castaneum the rate of dispersal increased with selection, as com-
pared to dispersal from fresh medium, while it was depressed in I,
confusum. Selection, therefore, served only to modulate the normal
response of each species. (I, confusum normally aggregates in homo-
typically or heterotypically conditioned flour, whereas I, castaneum
is normally repelled by both.) Ogden's results clearly indicate a
,genetic component to dispersal behavior. However, it is not clear
Inhether the genetic component relates to increased mobility or increased
preference for preferred medium.

The extent of selection, such as Ogden performed, in natural

INqulations and its consequences to population dynamics is largely

18

unknown. However, some work has been done in this area. Wellington
(1957), using a behavioral test for the ability of the tent cater-

pillar (Malucosoma pluviale) to move towards a light source when iso—

 

lated, found two distinct behavioral populations. Type I larvae were
capable of independent, directed movements as isolates and in groups.
Type II larvae consisted of larvae that were somewhat active if other
larvae were present, larvae that were almost always disoriented, and
larvae which hardly moved at all. The consequences of these behaviors
to spatial distribution will be discussed later. The relevant point
at this time is that these larval types have a genetic component
related to mobility differences. Since such differences were also
found in the wild, Wellington's results are not a laboratory artifact.

Since it was not my intention to investigate the genetic
factors in the spatial distribution of Galleria larvae, sex and genetics
were controlled as closely as possible. For purposes of my experiments,
I attempted to keep genetic variability, relating to spatial distrib-
ution, constant. As discussed in Chapter II, new stocks of larvae were
periodically introduced into the breeding colony and behavioral tests
utilized as indicators of whether response to biological conditioning
fluctuates. The results of these tests indicate a constant response
over time. These results demonstrate that variability in response
differed over time, but the average behavioral response remained con—
stant over the period tested.

Sexual differences were of no concern in my experiments with
(killeria. As near as anyone has been able to determine, Galleria

larvae exhibit no sexual dimorphoism. For purposes of the

19

experiments in this dissertation, therefore, Galleria larvae are

assumed to be sexually undifferentiated. However, there may be a

behavioral dimorphism relating to sexual differences which may account

for a component of the observed changing variance over time. This was

not investigated.

Density Factors in Spatial
Distribution

One of the variables investigated in this dissertation is the
effect conspecifics have on one another in terms of spatial distrib-
Iition and if the results of such animal-animal interactions are altered
as age, number of organisms, and degree of biological conditioning are

unanipulated. The discussion at this point will consider density

«iependent factors commonly associated with population spatial patterns.
latter we will discuss similar considerations in terms of biological

Ccnuiitioning. The density related literature in invertebrates,

Particularly insects, is of main concern to my studies. However,

relevant hypotheses derived with rodents will be briefly touched upon.
Stickle (1946) and Calhoun and Webb (1953), using small mam-

I“5‘15, have shown that stability of spatial distribution is related to

't}\e number of animals in an area. Animals in surrounding areas tend

1:5) move to vacated areas following removal of the residents. Intro-
CIllction of animals into already populated areas causes them to dis-
ID‘ETSG (Calhoun 1948) or to so disrupt the population that a temporary
CI<EC1ine in numbers occurs (Davis and Christian 1956).

Christian (1970) states that there may be a direct relationship

between the degree of aggressiveness and the magnitude of dispersal

20

with increased density and the associated conspecific contacts. As

numbers increase, the level of aggressive interaction may be increased,

and with it the tendency to disperse. Christian was referring to mam-

mal populations. However, these statements are generally applicable

to any population of organisms if the term aggression is dropped,

until we have a better understanding of what it entails, and simply

refer to animal-animal interactions. Therefore, there is a direct

relationship between the degree of animal-animal interaction and the

magnitude of dispersal with increasing density. At one end of the

continuum are low density aggregations with few interactions, and at
the other end is high density dispersal with many interactions. These
statements are particularly relevant in the so-called "asocial" orga-
Itisms which have not yet evolved sophisticated mechanisms or divisions
(If labor to unify large masses of animals and in which there is no
Obvious benefit derived from large aggregations.

Perhaps the best known works dealing with cyclical fluctuation
of Populations are those of Christian (1950), Chitty (1955), and
Kli‘ebs (1966) and Krebs gt a_l_. (1969) dealing with rodent populations.
(‘\1though cursory to the main thrust of this review, these authors
81‘6 briefly mentioned because their theories relate to animal numbers
Eir1d Spatial distribution and are relevant to some other papers to be
discussed, namely Wellington (1957, 1960) and Long (1953).) Myers
£111d Krebs (1974) have written an excellent review of this area.
(Iliristian hypothesized that high densities of organisms produce

Cu
5513955," i.e., disorders of the endocrine system, and thus precipitate

El (kmdine in the population, whereas Chitty felt that the effects of

21

lfigh density were to be found in the offspring, not in the exposed

generation. Chitty further hypothesized that high density stress is

genetically selective and that social interactions within the popu-

lation would change during the population cycle. Krebs tested several

of Chitty's hypotheses, as well as his own ideas about dispersal, and

developed a model of rodent population cycles. The model postulates

that two genotypes are present in rodent populations; one adapted to
survival under crowded conditions but having a low reproductive rate,
and the other being reproductively superior but intolerant of high

densities. As population size increases, the density—intolerant

individuals leave.

Krebs' hypothesis may be applicable to spatial distribution in
idisects. Wellington's (1957, 1960) work with tent caterpillars demon-
:strates the presence of two genetically distinct larval populations
WTuose varying behaviors differentially affect dispersal as density
Cflianges. On the other hand, Long's (1953) paper on two species of
"“3th larvae indicates that behavior does not change as density
i'lcreases, rather, the degree of behavior is altered. Similar results
pertain to other papers discussed throughout this review.

In a series of papers, Surtees (1963a, b, 8 c; 1964a, b, c,
€3” 5 f) studied the dispersal of several species of grain weevils,
EFIFGdn beetles, and flour beetles in relation to environmental cues.
1r}‘€3distribution, however, could not be fully accounted for until
(1‘31“fily effects were considered. A vertical distribution was dis—

c: . . . . . . .
o\Iered 1n the gra1n weev11$ L1t9ph11us gynarius, Oryzaeph1lus

S .
‘~1535323fl22§i§, and Stophilus granarius (Surtees 1963a, 1964d). As the

22

density of weevils increased, the number of weevils aggregating at the
surface of the medium increased. (These experiments are discussed in
some detail under "Environmental Factors" and will be reexamined under
"Biological Conditioning.") Surtees (1963c) also studied aggregations

of Tribolium castaneum and Crystolestes ferrugineus, finding that

 

 

mobility and aggregation behavior were affected by density, temperature,
and moisture. There were, however, species differences. For example,
surface aggregations of Tribolium were larger and more varied with

density, whereas Crystolestes exhibited a more even vertical distrib-

 

ution at all densities. These aggregation responses Surtees interprets
as modifications of individual locomotor activity which affects the
intensity of intra-group stimulation and leads to upward dispersal

and surface aggregations at high density.

In some instances it has been found that age of interacting
larvae, as well as non-conspecific interactions, modify the effects of
density. Salt and Hollick (1946), for example, found that not only
was the spatial distribution of wireworm populations (Agriotes) related
to physiological tolerances based on environmental cues, such as temp-
erature and humidity, but to the presence of other organisms as well,
such as ants. However, the developmental stage also affected con-
specific interactions. Medium sized larvae showed a less marked
aggregation behavior than did small (young) or large (old) larvae.

No hypothesis or test of this phenomenon was offered.

Legay and Chasse (1964) hypothesized that dispersal is dependent

on exploration and pOpulation pressure due to increasing density in

Tribolium. They set up several "enclosures" in which they varied the

23

number of beetles and recorded dispersal behavior. Initially, beetles
rapidly leave the enclosure, but the rate of movement declines as
population pressure decreases. It appeared that tactile stimuli

played an important role in this behavior, and leSsened population
pressure was interpreted in terms of lowered tactual stimuli. Inter-
estingly the enclosure eventually always emptied. Legay and Chasse
interpreted this to mean that exploratory behavior never reaches zero,
even when population pressure is low. Although not thoroughly inves-
tigated, they observed a curious phenomenon. The mean duration of stay
of individual beetles in the enclosure tends to increase with in-
creasing density. According to Legay and Chasse, this is a result of
accommodation to tactile stimuli among the beetles which had not dis-
covered the opening of the enclosure. However, they failed to account
for the fact that as density increases, the number of contacts increases
which results in increased population pressure and movement. Theor-
etically, increased density should increase the probability of an indi-
vidual finding the enclosure opening and dispersing, not increase the
duration of stay in the enclosure.

A possible explanation for increased time spent in the enclo-
sure as density gets high, is that the beetles are responding to bio-
llogical conditioning. Legay and Chasse failed to take into account the
literature preceding them which demonstrates conditioning responses in
Tribolium. I, confusum prefers homotypic and heterotypic conditioning,
whereas I, castaneum is repelled by both. A species comparison would
have been very helpful in their studies. In the case of I, confusum,

an increased mean duration of stay in the enclosure is to be expected

24

with increased density because the conditioning, which they prefer, is
also increasing. The reverse should be true of I, castaneum. Such
experiments would be necessary to separate out density effects from
conditioning effects.

Zromiska-Rudjka (1966, from Ogden 1970) performed such a species
comparison, but again conditioning effects were not accounted for. He
looked at the effects of density on dispersal in I, confusum and I:
castaneum in atrophic (roasted sawdust) and normal medium. Dispersal
in the atrophic medium seemed to be independent of density, but if a
small amount of normal medium is added dispersal becomes density
dependent. He further found that dispersal is only slightly density
dependent in I, confusum. This latter result may have been due to
increased attraction to conditioning as density increases, whereas I,
castaneum is repelled.

Some studies of dispersal indicate that there may be "trans-
developmental-stage" effects of density. That is, the density of one
developmental stage may affect the dispersion behaviOr of future

stages. Long (1953) was interested in the connection between high

larval (Lepidoptera) density and subsequent adult migratory behavior.

 

His long range goal was to see if larval density was associated with
color phases in the adults, as had been demonstrated in locusts. The
results of his experiments show that crowded larval cultures develop
more rapidly than solitary individuals, partly because the larvae go
through fewer instars. If cultures became excessively crowded,
however, development slowed. Long's basic theoretic was that behavior

may be viewed as the outward expression of the ability of an organism

25

to respond to its internal and external environments. The prevailing
pattern of behavior may therefore be considered as the exhibition of
those responses being evoked at any one time. Thus crowding may lead
to distinct behavioral changes. Crowding produces a change in behavior
reflected in higher levels of activity which involves many non-feeding
movements. Feeding activity in solitary larvae, however, is less.

When compared with the solitary condition, crowding repre-
sents a change of environment and accordingly may lead to a change in
the frequency of behavior without necessarily involving a change in

the basic behaviors. For example, the larvae of Plusia gamma, the

 

British migrant silver Y-moth, in crowded cultures frequently show
avoidance reactions to other larvae as well as warding off movements
with the head and anterior region of the body. Isolated larvae
occasionally exhibit non-feeding movements but are mostly inactive.
If, However, a solitary larva is placed in a crowded situation, it
exhibits behaviors typical of crowded larvae. Other differences in
behavior occurred in solitary and crowded cultures, but the most
striking was the general impression that, whereas the solitary larvae
were relatively immobile, larvae in crowds were inclined to "rest-
lessness." A higher activity level in high density situations is not
surprising, as it is possible that the environment Of crowded larvae
initiates extra responses. Long also found that individuals in
crowded cultures spent only one fourth as much time feeding as did
solitaries. Thus, the extra activity is not just feeding but also
entails locomotion, swaying motions, and irritation reactions between

touching larvae.

26

In another study, Long (1955) attempted to demonstrate that
the density of egg clusters is the main determining factor in larval
dispersion behavior. However, his results are inconclusive because
they were confounded by biological conditioning behaviors and a not
very convincing species comparison. This study is discussed under
"Biological Conditioning."

In the field, larval forms of certain Lepidoptera tend to be

 

solitary and such a condition can be reproduced in the laboratory.

In other species living in close aggregations, the population density
considered normal for the species is not known. This leads to dif-
ficulties in interpreting the results of an experimental condition of
isolation versus crowding as a departure from normal. This is also
one of the main reasons I elected to approach dispersion studies with
Galleria by beginning with isolates and gradually increasing group
size. It also becomes necessary to be able to separate density
effects from conditioning effects, which is best accomplished by
first obtaining base line responses to conditioning before asking
questions about animal-animal interactions. This is the major reason,
in such studies as Long's, that it has not been possible to defini-

tively elucidate the critical variables in spatial distribution.

Biological Conditioning, Defined

 

Every organism modifies its environment to some degree,
whether it be the intricate structures built by spiders, birds, or
bees, the making of burrows in aquatic and terrestrial forms, the
secretion of simple structures such as slime tubes of marine annelids,

trail marking in ants, scent marking in mammals, or hive odors in

27

bees. Such environmental modifications are classified as "biological
conditioning."

Biological conditioning is ". . . changes produced in the
medium by organisms living therein and the effects of such changes
upon other organisms . . ." (Allee 1934). Allee's definition contains
two parts. One relates to changes in the medium and the other to the
effects such changes have upon other organisms. My studies are con—
cerned with the effects of biological conditioning on conspecifics,
not with the conditioning per se. Changes in the medium include
chemical as well as physical changes. Allee g£_§l, (1949) also refer
to biological conditioning as ”environmental conditioning" which they
define as the modification of the environment by population-group
activities. Biological conditioning can be homotypic or heterotypic.
If the conditioning an organism responds to is produced by the
organism itself or by a conspecific, it is homotypic. If produced by
other than a conspecific, it is heterotypic. These definitions encom-
pass any modification an organism makes in its environment. However,
careful reading of the literature reveals an implicit attempt at
parcelling or categorization of these phenomena into specific areas.
This categorization seems to be based upon: (1) source of the con-
ditioning, (2) medium in which the conditioning occurs, and (3) dura—
tion of the conditioning.

The source of the conditioning refers to whether it emanates
from the organism or predominantly from the manipulation of environ-
mental materials. The former is usually considered biological con-

ditioning, in a strict sense, and the latter is often classified as

28

"external constructions" or "animal architecture," although neither
category is exclusive of the other. Conditioning emanating from the
organism refers to various organismic secretions and/or excretions.
Falling into this category would be such things as spider webs (Kaston
1964; Witt and Reed 1965), quinone secretions by Tribolium (Alexander
and Barton 1943; Loconti and Roth 1953), spinning behavior in Galleria
mellonella (Paddock 1913; Milum 1952), chemotaxis in Paramecium
caudatum (Dryl 1963), aggregation due to chemical stimuli in planaria
(Reynierse 1967), and nest-entrance marking by honey bees (Butler et_
31' 1969). Examples of conditioning based solely on environmental
materials are nest building in birds (Collias 1964) and various
burrows dug by mammals. This dichotomy is not a discrete one, since
conditioning might easily be a combination of environmental materials
and secretions or excretions. Honey bee hives, termite nests, and
caddisfly cases are combinations of the two.

The second consideration in biological conditioning relates
to the medium in which the conditioning occurs. Most of the litera-
ture, classified as studies in biological conditioning, deals with
Conditioning in some sort of solid medium (i.e., soil, flour, or
grain) or aquatic medium. This is true of most early work in the
field, particularly Allee's studies, and of much of the recent work
with such organisms as Tribolium, grain weevils, Galleria, and
iPIanaria. This also fits in with Naylor's (1959) concept of
llerbivorous organisms that find food and shelter in a fixed location.
liowever, researchers do not include air in their concept of a bio-

logically conditionable medium because the majority of conditioning

29

in air is chemical, which is usually considered chemical communi-
cation (see Johnson §£_al, 1970). At this juncture, it is necessary
to point out that all chemical secretions are considered as biological
conditioning, but not all biological conditioning is of a chemical
nature, much of it is physical conditioning.

Many authors restrict biological conditioning to having long-
lasting behavioral as well as physiological effects, and most air-
borne conditionings are not of this nature, although no one has defined
"long-lasting." Examples of these are sex attractants in phasmid
moths (Taylor 1931), trail phermones in ants and other insects (Carthy
1951; Karlson and Butenandt 1959), and scent marking in mammals (Johnson
1973). This is not true, however, if the chemical conditioning is
constantly reinforced. Besides being of relatively short duration,
many air—borne phermones tend to have short term effects. These
statements do not pertain to other than chemical signals, such as
spider webs. On the other hand, conditioning of all types in a more
solid medium tend to be more long-lasting than in air and have per-
Sistent behavioral and physiological effects. This is probably a
function of the medium, not the stimulus.

A categorization of biological conditioning as to source,
medium and duration is of little help in understanding the phenomenon,
and has created problems for anyone interested in such studies. Part
of'the problem stems from the vastness of the area in which no real
Synthesis has yet been attempted. In fact, the last review of this
area was written by Allee (1934). Another difficulty, as with any

SCientific area, is purely semantic. The major difficulty stems from

30

the fact that studies in biological conditioning originally were off-
shoots of other kinds of studies, and, from a behavioral point of view,
have never come into their own as a field of investigation. Recently,
however, there have been several excellent works dealing with reviewing
some aspect of conditioning, such as a review of communication by
chemical signals (Johnston e£_al. 1970) and a book on animal archi-
tecture (Von Frisch 1974).

It would be more useful to classify biological conditioning,
in natural and experimental populations, on the basis of the pro-
cesses involved. This seems a more practical point of view than
attempting to categorize conditioning on the basis of the medium in
which it occurs or how long-lasting it is. These are aspects of the
problem, but purely artificial as a basis for classification. Allee
§£_al. (1949) originally took such a practical approach and devised
the following classification, with modifications. Biological con—
ditioning consists of: (1) reduction and/or partial distribution of
the available food supply; (2) addition of contaminants to the environ-
ment; (3) liberation of a "growth-promoting," or some other needed
Substance to the medium; (4) fixation by the population of toxic sub-
Stances ("detoxification"); (5) osmotic regulation of an aquatic
environment; (6) physical conditioning of the environment; (7) chemical
Conditioning of the environment; (8) compound conditioning, or com-
binations of categories 1—7.

Before proceeding to specific studies, the following general-

ities should be kept in mind:

31

1. Studies in biological conditioning entail studying the changes
produced in a medium by an organism or group of organisms
as well as the effects of such changes upon other organisms,
particularly conspecifics.

2. These changes may be produced in any medium, i.e., soil, grain,
flour, water, and air.

3. The changes produced can be physical or chemical and homo-
typic or heterotypic.

4. Chemical conditioning in air and physical conditioning in any
medium are usually classified as "Chemical Communication" and
"Animal Architecture or Construction," respectively. However,
both of these phenomena come under the broader classification
of biological conditioning.

5. The effects of changes in the medium upon other organisms may
be physiological, morphological or behavioral and of varying
duration.

6. Biological conditioning does not solely entail adding something
to the medium. It may also be a rearrangement of the medium
or removal of substances from the medium (detoxification).

7. Biological conditioning can be produced by an individual or
by groups of individuals, and therefore is not a purely pOpu-
lation phenomenon. In this context, animal-animal inter-
actions become very important.

The remainder of this review will deal with some of the classic
‘Nork in biological conditioning, exemplifying the various categories

of"conditioning as well as illustrating its physiological and behavioral

32

effects. The physiological and behavioral effects of conditioning
will be discussed in the context of animal-animal interactions. This
is particularly relevant in light of my studies in this dissertation
which are concerned with the function of biological conditioning in
relation to spatial distribution, and how preferences for conditioned
medium are affected by animal-animal interactions coupled with develop-

ment (age) of Galleria mellonella larvae.

 

Physiological Aspects of Biological
Conditioning

 

 

Most of the early literature deals with biological conditioning
in relation to the physiological changes produced in various organisms.
These studies were classified as studies in "Mass Physiology" which
Allee (1934) defined as the result of the aggregations of many indi-
viduals of the same species in a limited space. An excellent review
of this concept is also found in Allee's (1931) book, Animal Aggre-
gations. Basically, studies in mass physiology deal with density and
biological conditioning as they affect reproduction, fertility,
fecundity, natality, mortality, and growth. For example, Edmondson
(1945) found that growth in several species of sessile rotifers was
affected by crowding. Isolates were larger than colonial forms,
Possibly a result of competition for food, but aggregates began
reproducing sooner and lived longer than solitaries. Edmondson's
COnclusion was that crowding acts to enhance survival.

Studies in mass physiology deal with (1) protection from toxic
3Ubstances, (2) control of sex, (3) effects on morphological changes,

(4) effects on growth, and (5) effects on population dynamics.

33

Protection From Toxic Substances
Often a group of organisms will yield each other protection
from a toxic environment to which animals exposed singly would succumb.

The turbellarian flatworm (Procerodes wheatlandi), for example, sur—

 

vives longer in fresh water (it is normally an intertidal dweller) if
present in large numbers and if the water has been previously homo-
typically conditioned. This conditioning is particularly effective if
it consists of some dead and disintegrating flatworms (Allee 1931).

The greater the number of Planaria dorotocephala in a group,

 

the greater the survival of individuals after exposure to colloidal
silver (Allee and Schuett 1927). Similar results have been obtained
with P. maculata and P, lacteum as well as with the brittle starfish
gphioderma brevispina. The greater protection which is afforded by
the mass when exposed to colloidal silver is probably due to the
smaller amount of the toxic substance which each individual removes
from the water in order to lower the strength below the threshold of
immediate toxicity. No evidence has been found for the presence of an
"auto-protective" or "auto-destructive" agent which would protect
individuals from toxic substances. Evidence did indicate, however,
that slime secretion may remove the silver and render the solution
less toxic. Aggregations of mixed species also exhibit the protection
from colloidal silver.

Perhaps the most classic study of mass protection from toxic
Iconditions is Allee and Schuetts's (1927) study on the relation
‘between mass of animals and resistance to colloidal silver. Tests

Were performed on a variety of organisms such as planaria, brittle

34

starfish, isopods, and goldfish. In all cases the results show a
greater protection from toxic substances the greater the mass of
animals present.

It is not necessary to further demonstrate that mass pro-
tection is a common phenomenon. However, the mechanism has also been
investigated. According to Allee (1934), four theories have been
proposed to account for mass protection:

1. Groups of organisms more effectively produce auto-protective
substances which in some manner conditions the medium and
protects the group (Drzewina and Bohn 1921).

2. The mass of animals more rapidly absorbs or exhausts the
toxic agent, so each animal receives a smaller dose (Allee
and Schuett 1927; Breukelman 1932).

3. Grouped animals reduce their metabolism, as measured by
oxygen consumption, compared with isolates. This reduction
favors survival in toxic solutions (Allee and Fowler 1932).

4. The group may be protected by altering the electrical con-
ditions (Drzewina and Bohn 1926), or animals may be mutually
influenced by the presence of other animals without the
diffusion of any substances.

All of the evidence collected by Allee favors the assumption
that the mechanism, at least as it relates to colloidal silver, is
'PTOtection due to the more rapid removal of the toxic element by the
maSS Of individuals or their products. The evidence further indi-
cates that no other factor need be assumed to account for protection

from other substances. However, in most of Allee's studies the

35

organisms also exhibited a general reduction in metabolism. No one
has attempted to decipher if the conditioning merely brings individuals

into a group and the group then, not the conditioning, provides pro-

tection from toxic substances.

Control of Sex

' Attempts have been made to determine the relative role played
by food, crowding, and conditioning in relation to control of sex, but
the most classic attempt was a series of papers in the early 19005
relating to control of sex in Cladocera (daphnids). Cladocera is
normally parthenogenetic, but under certain conditions males are pro-
duced. According to Allee (1934) it had been hypothesized that crowd-

ing of the females in Moina rectirostris produces a higher percentage

 

of males, and experiments testing percent of male Offspring from
female isolates versus female groups seemed to confirm this hypothesis.
Initial conclusions were that the results were due to an increase in
excretory products in the food, but further tests did not verify this.
For some time hypotheses fluctuated between a food explanation and an
explanation based on increased male production due to female excretory
products. Earlier work supported the food hypothesis. Issakowitsch
(1905, from Allee 1934) found he could change the form of reproduction
in Cladocera from parthenogenetic to bisexual by manipulating food
supplies, and similar studies indicated that repeated transfers to
fresh culture water rich in food delayed the appearance of bisexual
fOrms. Metabolic wastes did not appear to affect this transformation
and thus the effect was attributed to decreasing food supplies.

Smith's (1915) work, however, supported the excretory product

36

hypothesis. He grew cultures of Protococcus on excess food in crowded

 

conditions and found excess of male production. He concluded that
crowding exerts its effect through some sort of excretory product.
Stuart and Banta (1931) demonstrated a relationship between the
bacteria present in Cladocera cultures and male production. Male
production increases, within limits, as the number of bacteria
decreases, and there were no indications that the bacteria were
absorbing excretory wastes.

It seems obvious, in retrospect, that sex control is probably
a multiple effect of food, crowding, and conditioning, and that the
mechanism may be different from species to species. Brown and Banta
(1935) postulated such an interaction and their results seem to
support it. They found that the amount of food acts as a broad or
limiting factor in male production, whereas crowding of females and
excretory products act as immediate factors to increase male numbers.
This series of experiments exemplifies reduction and/or partial dis-
tribution of available food supply and addition of contaminants or
some needed substance to the medium as types of biological conditioning.

A second example of how crowding, and possibly conditioning
affects sexual differentiation comes from Loomis and Lenhoff's (1956)
wOrk with hydra. Clonal growth within a mass culture of hydra follows
a segmoid curve and sexual differentiation occurs spontaneously in
growing clones whenever a critical level of population density is
‘reached. This "positive group effect" results from the differentiation-

inducing activity of a volatile chemical or gas secreted by the hydra

37

themselves, and the secretion may be a reaction to animal-animal

interactions at increasing density.

Effects on Morphological Changes

One of the best known examples of the effects of mass physiology
on morphological changes relates to crowding and the phase theory in
locusts. Uvarov (1928) proposed the theory that species formerly sup-
posed to be monomorphic are in fact polymorphic as regards coloration
of the nymphs, form of the pronotum, body size, and relative wing
length of adults. A "gregarious phase," which demonstrate a tendency
to migratory behavior, and a "solitary phase," which lacked such
behavior, were identified in locusts. The gregarious phase is darkly
colored and the solitary phase is green. Uvarov suggested that
locusts pass from one phase to the other, depending on population
density. This theory has since been well worked out in locusts and
other species of Lepidoptera (Long 1953, 1955; Doull 1953; Ellis 1953).
However, it was Faure (1932) who first demonstrated the mechanism.
Crowding causes increased activity and the resulting increase in
metabolism deposits end products in the cuticula giving the gregarious

phase its dark color.

Effects on Growth

It has been repeatedly demonstrated that density affects
growth rate. Allee (1931) reviewed the early literature on the effects
of space limitations on growth rate and Adolph (1931) noted that crowd-

ing in Rana sylvatica tadpoles causes an earlier decline in growth

 

rate than in isolates. If isolated after having been crowded, growth

38.

rate increases. Similar studies in "growth depensation" were run by
Brown (1946, 1951) with trout fry, Magnusen (1962) with Medaka, Peebles
(1929) with echinoderm larvae, and Richards (1958) and Rose (1959)

with tadpoles, all of whom found growth effects produced by grouping
of these organisms. Hypotheses were advanced that growth differences
were due to either spatial, numbers, or conditioning relationships, but
none of these was clearly indicated. These are probably cases of
multiple interactions, but the usual approach has been to single out

a variable of interest and examine its effects on growth.

Homotypic and heterotypic conditioning of the environment and
their effect on growth was studied by Livengood (1937) and Allee e£_al,
(1934, 1940) in goldfish, Shaw (1932) in fish and amphibians and West
(1961) in tadpoles. Most such studies concluded that growth is most
rapid in "mildly" conditioned medium and retarded in overcrowded and/
or over-conditioned medium. Excretions or depleted food supplies were
postulated as the mechanisms, but never demonstrated. Adolph (1931)
decided that growth retardation in crowded conditions was due to
"psychological" effects, not physiological effects. That is, individ-
uals in crowded cultures eat less but live longer than do isolates.

On the other hand, Welty (1934) found just the opposite in fish.
Crowded fish ate more than isolates but grew less. However, Welty
did not replenish the culture water in his experiments and his results
may have been due to growth-retarding contaminants or to group
stimulation to greater activity. In 1921, Robertson gave the name
"Allelocatalysis" to the process of added "contaminants" affecting

growth.

39

Effects on Population Dynamics

Aside from growth effects, mass physiology also affects
fertility, fecundity, mortality, and natality. Perhaps the best
worked out example of mass physiological effects on population
dynamics is that begun by Pearl and Chapman and pursued by Park with
Tribolium. These studies also serve as the basis for most of the work
on biological conditioning and behavior.

Pearl (1927) showed that various populations follow a logistic
growth curve and postulated that increasing density affects growth by
affecting natality and mortality. Chapman (1928) tested this hypoth-

esis with Tribolium confusum and found that an equilibrium condition,

 

measured in terms of numbers of individuals per gram of flour, is
eventually reached, within each population, which is uniform regard-
less of environmental size or initial densities.

Thomas Park (1932) first began his experiments with Tribolium
confusum by asking whether the initial rate of population growth is
greater in beetle populations when more than one pair of beetles is
present. He utilized constant sized environments but variable numbers
of pairs per vial (i.e., 2, 4, 8, and 16 pairs). Measurements of
population growth were taken at 11 and 25 days. At 11 days, the most
eggs had been produced by the initial population of 2 pairs and the
fewest by 16 pairs. At 25 days, this relationship had reversed.
Eventually the populations reach the same equilibrium, regardless of
the initial starting density. Further studies enabled Park (1932) to
elucidate the mechanisms producing these findings. Intermediate sized

Tribolium populations grow more rapidly at 11 days than smaller or

40

larger populations because (1) egg cannibalism favors greatest
increase in minimal sized groups since in crowded groups more eggs

are found and eaten by random movement, and (2) copulations and rec0pu-
lations favor greatest increase in maximal sized groups, recopulation
being stimulating to reproductive productivity. The interaction
between egg cannibalism and copulations thus favors the greatest
initial increase in intermediate sized populations.

As populations of Tribolium inhabit their flour, the flour
becomes progressively more conditioned with time. The extent of this
conditioning is proportional to the population density. Noticing
this phenomenon, Park (1934, 1935, 1936 a and b), Park and Woollcott
(1937), and Park e£_al, (1939) began a series of investigations into
the effects of "auto-conditioned" flour on populations of Tribolium.
'The reader is directed to Park and Woollcott (1937) and Allee et 31.
(1949) for excellent summaries of these studies, only the specifics
will be presented here.

Conditioned flour was found to affect beetle physiology. Con-
ditioning reduced egg cannibalism and fecundity, increased egg via-
bility and duration of larval development, and seemed to have no
effect on egg fertility or the rate of oxygen consumption of adults.
Dilution experiments, in which heavily conditioned flour was graded by
adding variable amounts of fresh flour, showed that beetle fecundity
is proportional to the amount of conditioned medium. Conditioning also
affects metamorphosis. Larval crowding increases larval and pupal
Inortality and conditioning seems important here. However, the duration

of the pupal period is not altered by density, sex, or conditioning.

41

There are two general conclusions possible from Park's studies
with Tribolium: (l) flour becomes progressively conditioned as popu-
lations age, and (2) Tribolium populations decline as they age if
their flour is not renewed. Therefore, conditioning is an expression
of population age and size and the number of beetles is responsible
for the rate and amount of conditioning. It seems fairly clear that
the decline of Tribolium cultures is largely due to reduction in
reproductive rate as a function of conditioning. Even today, however,
it is not clear what constitutes conditioning but it seems reasonable
to assume that it is a combination of food depletion and addition of
toxic waste products. Park and Burrows (1942), for example, have
shown that fecundity is sensitive to changes in the nutritive level of
the medium, and Roth and Howland (1941), Roth (1943), and Loconti and
Roth (1953) have demonstrated that crowded imagoes liberate a gas
from specialized odoriferous glands which causes morphological abnor-
malities. Crombe's (1942) study lends further credence to Park's
data by demonstrating that fecundity is altered by conditioning.

In light of Christian's (1970) hypothesis dealing with changing
behaviors as density increases, the results of Park's work might be
interpreted as changes in the degree of egg cannibalism and repro-
ductive behaviors as population density and conditioning increase.

Biological Conditioning:and
Spatial Distribution

 

 

Other than having physiological and demographic correlates,
biological conditioning is associated with numerous behavioral

responses in individuals and populations. Chemical communication, as

42

reviewed by Johnston gt a}, (1970), is an outstanding example of the
behavioral relationships of conditioning. Chemical conditionings
(specifically phermones) are often classified as releasers, chemical
stimuli initiating immediate behavioral responses, and primers,
chemical stimuli inducing delayed responses (Wilson and Bossert 1963).
Chemical stimuli have been found associated with recruitment in ants
(Karlson and Butenandt 1959; Regnier and Law 1968), sex attractants
(Silverstein and Rodin 1967; Jacobsin gt_al: 1970; Silverstein et_al,
1967; Regnier and Law 1968), and alarm reactions (Maschwitz 1964;
Pfeiffer 1963).

Biological conditioning by means of chemicals is a very common
phenomenon among the animal phyla. It has been well studied in mammals
under the heading of "scent marking." Although scent marks may have
any number of functions, one hypothesis is that they serve to warn
conspecifics of an already occupied territory and function to keep the
population spatially dispersed. Our main concern in this review,
however, is with the invertebrates, particularly insects, and the
reader is directed to Johnson (1973) for a review of chemical con-
ditioning in mammals.

Various physiological and demographic effects of conditioning
have been illustrated, but the relationship between biological con-
ditioning, animal-animal interactions, and dispersion is not as well
worked out. Biological conditioning has historically been thought of
as a population phenomenon since it is an expression of population
density and is a normal and inevitable result of population growth.

As populations grow and maintain themselves they must of necessity

43

modify their environment, and the environment in turn modifies them.
Since biological conditioning is a result of the organisms themselves
and since the degree of conditioning is a function of the number of
organisms present, biological conditioning and animal-animal inter-
actions will be considered as a unit.

One of the classic ecological studies is Edmundson's (1945)
paper on sessile Rotatoria. He found that the distribution of several
rotifer species is partly dependent on responses to heterotypic con-
ditioning by the plants upon which they settle. Some rotifers, i.e.,

Floscularia conifera, seem to respond to homotypic conditioning as

 

well. That is, they are found more abundantly attached to each other
than to the plants. However, this may have been a tactual response
and not a conditioning response at all. Heterotypic conditioning by
algae and texture of the settling surface were found to be important
to the spatial distribution of oyster larvae (Gee 1965). Cole and
Knight-Jones (1949) had earlier discovered that oyster larvae tend to
be gregarious on settlement and prefer to attach themselves when they
have found a surface of suitable texture, a surface that had been
homotypically conditioned by other larvae, or a surface where other
larvae were already attached. Knight-Jones and Stevenson (1951)
found basically the same responses exhibited by the barnacle larvae

Elminius modestus. The conditioning response and response to properly

 

textured surfaces may be linked. That 15, conditioning may consist of
physically modifying the settling surface.
When a species is found to exhibit some sort of predictable

distribution, researchers begin a search for the mechanism(s) involved.

44

The work with planaria is a good example of the procedure. Planaria
form large aggregations (Allee 1931). Fraenkel and Gunn (1961) sug-
gested that aggregations may form mechanically by means of an ortho-
kinesis, where rate of movement depends upon stimulus intensity, or a
klino-kinesis, where rate of random turning depends upon stimulus
intensity. This explanation of aggregation behavior was made in
reference to phototactic responses. However, Reynierse (1967) found
that planaria secrete some substance, possibly species specific, to
which conspecifics are positively attracted. This form of biological
conditioning was interpreted as inconsistent with Fraenkel and Gunn's
requirement that aggregations are accidental and mechanical results of
a kinesis. However, chemical secretions may affect both the rate of

movement and rate of turning. A similar study was performed by Dryl

 

(1963) with Paramecium caudatum in which he found them maintaining
their spatial distribution by an avoidance reaction to chemotactic
stimuli. But none of his data indicate whether the Paramecium were
secreting the chemical.
There are currently three hypotheses in vogue relating to
aggregation formation in invertebrates, specifically in planaria:
1. Aggregation formation is due to the secretion of some chemical
substance to which conspecifics are positively attracted
(Walter 1907; Allee 1931; Reynierse 1967).
2. Aggregations form mechanically as a result of an orthokinesis
or klinokinesis (Frankel and Gunn 1961).
3. Aggregations form as a result of distinctive morphology. That

is, an aggregation forms a distinctive viSual or tactual

45

stimulus to which conspecifics respond (Reynierse 33 31.

1969).

Using various degrees of illumination and natural as well as
artificial planaria aggregations, Reynierse §£.§l: (1969) found support
for all three hypotheses. In other words, aggregation behavior will
only rarely be explainable on the basis of single mechanism hypotheses.
This appears to be a truism, but it is an extremely important con-
sideration for studies in spatial distribution and accounts for much
of the conflicting and inconclusive results seen in the literature.

The most well worked out studies dealing with biological con-
ditioning and animal-animal interactions are those stemming from Park's
work on Tribolium. We have already seen the physiological effects of

conditioning in Tribolium confusum as worked out by Park (1934, 1935,

 

1936a and b), Park and Woollcott (1937), and Park gt_a}, (1939). In
his 1934 paper, Park provided adults with a choice between fresh
(unconditioned) flour and two year old (conditioned) flour, and noted
that, besides the effect on egg production and fecundity, adults
avoided the conditioned medium. Therefore, the physiological conse-
quences of conditioned medium are probably not a result of adult
preferences for over-conditioned medium. Park (1948) further inves-

tigated competition between Tribolium confusum and Tribolium castaneum,

 

 

finding that both species cannot live together for extended periods of

time, and if forced to do so, Tribolium castaneum is the successful

 

competitor. The results of such competition may be due to differ-
ential conditioning of the medium, causing the reduction of available

food, the elaboration of environmental poisons (i.e., excrement and

46

carbon dioxide), or comminution of the flour itself. These possi-
bilities were not tested.
Park assumed the movement of flour beetles to be random. In

1932, Stanley analyzed the movements of Tribolium confusum on the basis

 

of the kinetic theory of gases, considering the beetles as particles
having random movement. He found that the predictions of randomness
were not consistent with experimentally observed populations, and
hypothesized that this assumption is not valid because the beetles are
leaving persistent tunnels in the flour and the probability of re-
traversal of a tunnel is greater than that of making a new tunnel.
Using a theoretical organism, Stanley (1949) mathematically demon-
strated that the probability of re-traversal of existing tunnels
increases with time whereas the probability of constructing new tun-
nels decreases. In theory, as time approaches infinity, the whole
flour mass becomes converted into old tunnels, all traversed an
infinite number of times. In actual practice, tunnel making sta-
bilizes at some point before this occurs because the beetles prefer
using existing tunnels. Observations on Tribolium seem to confirm
Stanley's prediction that spatial distribution is a function of bio-
logically conditioning of the medium (tunnels). However, his theory
does not account for animal-animal interactions which may be a part
of the tunneling behavior observed.

Some of Stanley's observations also shed some light on the
behaviors involved in the egg cannibalism observed by Park. It was

(tiscovered that eggs are laid in the tunnels and as population

47

density increases, egg cannibalism increases as a result of the adults
retraversing existing tunnels.

Rich (1956) confirmed these observations and states that popu—
lation ecology is, in the final analysis, concerned with the inter—
action of natality, mortality, and dispersion. These three factors
produce shifts of numbers over time in a reproducing population. In
Tribolium, however, these shifts in numbers are not constant over
time and egg cannibalism is probably a function of number of adults
present, egg distribution, and the pattern of beetle movement.

One of Park's students, Naylor (1959), made the next major
contribution to understanding the relationship between biological
conditioning and animal-animal interaction in the spatial distribution
of Tribolium. He was interested in the behavioral mechanisms of
dispersal, as distinct from natality and mortality, in Tribolium
confusum populations, and was one of the first to devise a choice
situation. His apparatus consisted of a circular device with vials of
flour around the periphery. The number of vials could be varied as
well as the flour content of each vial. Beetles were placed in the
center of the apparatus and allowed to choose the preferred environ-
ment. Statistical analysis enabled him to distinguish between uniform,
variable, and random distribution after choices had been made.

Several experiments were performed using varying densities of popu-
1ations and the results show that at low densities the beetles either
aggregate or respond to minor differences in the environment. As
denisity increases, "population pressure" overcomes the influences pro-

:duc:ing aggregation and forces uniform patterns of distribution.

48

Naylor's (1959) next series of experiments were designed to
determine what behavioral factors were involved in establishing these
distributions. He first examined male-female interactions, and found
that females spend more time in the flour than do males. At all
densities and in mixed or uniform sex populations, females exhibit a
uniform distribution whereas males aggregate. Both sexes also pre—
ferred conditioned medium. The males' preference for conditioned
medium is consistent with their aggregation tendency, but the same
preference on the part of females is inconsistent with their uniform
distribution. The conditioned vials, however, were initially free
of other individuals and it is possible that the females would not
have aggregated in conditioned medium had other beetles been present.
To test this hypothesis, it is first necessary to demonstrate that the
beetles are capable of behavioral responses to other individuals.
Naylor ran several preference tests to determine this. If given a
choice between vials with trapped beetles of either sex and vials of
only fresh flour, males in single sex groups enter the chambers with
trapped beetles. If the choice is between male or female chambers,
males prefer females. Females, on the other hand, selected non-
occupied chambers, but if all vials contained trapped beetles they
showed a greater tolerance for males. These experiments demonstrate
that beetles behaviorally respond to other individuals but do not
GXplain why females aggregate in conditioned medium, unless condi—
tioning overrides their tendency to be repulsed by other females.

This was not tested.

‘I

49

In 1965, Naylor ran a similar set of experiments except larvae
were used as trapped animals. Females avoided these vials and the
avoidance was density dependent. The same experiments were also run

with Tribolium castaneum (Naylor 1961). The results were similar

 

except that spatial distribution was not as markedly affected by
density and male I, castaneum is repulsed by conditioning.

In general, if sex is ignored, the spatial distribution of
Tribolium is dependent on density. But if sex and conditioning are
introduced as variables spatial patterns are altered. Naylor (1959)
attempted to fit these findings into some kind of population model.
Females enter the vials to feed and oviposit. By distributing them-
selves uniformly, few suitable sites are overlooked and few would be
crowded by hatching larvae. This would increase the probability of
larval survival. It would also be advantageous for females to be able
to respond to levels of occupancy of the habitat and make appropriate
adjustments in dispersal. If females were capable of such behavioral
responses, they might form low density aggregates but disperse if
density became too great. One factor Naylor's model does not account
for is that inbreeding may be promoted at low densities which may nOt
be advantageous to the population.

Naylor's (1961) paper was the springboard for a series of
articles dealing with species comparisons of animal-animal interactions
and biological conditioning. In 1963, Ghent contrasted the behavioral

responses of Tribolium confusum and I, castaneum to fresh and con-

 

<iitioned flour and attempted to assess the influence of these responses

qu)n distribution, population growth, and competition between the two

50

species. A vertical distribution exists in these species dependent
upon sex, density, and conditioning. I, castaneum prefers fresh flour
and I, confusum prefers conditioned flour, regardless of sex or type
of conditioning. I, confusum is attracted to homotypic and hetero-
typic conditioning, whereas I, castaneum is repelled by both, and
these preferences are influenced by population density. As density
increases, the proportion of the total population that the preferred
niches will tolerate decreases. Therefore, at high densities the
distinction between preferred and non-preferred niche becomes hazy.
With increasing density, females of both species show a greater ten-
dency than males to enter a non-preferred niche. The increased den-
sity seems to "drive" beetles to the surface of the flour where the
males normally remain. Females, on the other hand, chose to enter a
non-preferred flour medium. Female I, castaneum will enter conditioned
medium rather than stay on the surface, and female I, confusum enters
fresh flour. It is also generally the females that come to the sur-
face at high densities, not the males. This indicates that the males
are forcing them upwards, as Naylor (1959) suggested, or that the
females are somehow sensing their environment and moving to less dense
areas. Ghent postulated that these behaviors represent an adaptation
that allows aggregations of I, confusum to follow in the occupancy

of a niche as a successor to some species like I, castaneum that
becomes progressively more repelled by the accumulating conditioning.
Since Park (1948) demonstrated that I, castaneum out-competes I,

confusum, this may be a mechanism to avoid competition.

51

In 1966, Ghent pursued this hypothesis by bringing the two
species together and asking whether the resultant distributions would
be as predicted from the preceding facts. He placed a population of
each species in a fresh flour medium and found the following: Both
male and female I, castaneum entered the flour first and began a down—
ward migration. Male and female I, confusum entered more slowly,
possibly as a result of fresh flour being their non-preferred medium.
However, female I, confusum enter more readily than do the males, as
found in Ghent (1963). A vertical stratification is thus established,
with male I, confusum near the surface, female I, confusum in the
middle, and male and female I, castaneum on the bottom. As the male
and female I, castaneum conditioned the bottom flour layer, they became
repelled by it and started a migration back up to the surface. At
the same time the male and female I, confusum are attracted by the
heterotypic conditioning and migrate downward, the conditioning having
overcome male I, confusum's inhibitions about entering the flour. The
two species populations cross somewhere in the middle of the flour.
Concomitant population dynamic data showed that initially egg pro-
duction and survival were higher in I, castaneum until repulsion to
conditioning set in.

On the basis of this experiment, Ghent's (1963) model, that
the preferences of I, confusum represent a mechanism for escaping com-
petition, permitting succession in occupancy of previously unexploited
environments, seems tenable. One may envision a species like I,
castaneum being increasingly repelled by a niche as it becomes con-

ditioned and abandoning that feeding niche before its nutritive value

52

is fully exploited. The availability of such an incompletely
exploited niche may then have furnished the element of survival value
that promoted the selection of conditioned flour preference among
ancestral I, confusum. Ghent beautifully supports this theory with
an argument drawn from the plant kingdom. The larch (tamarach) grows
best on relatively dry, rich uplands. Yet, as a continuous popu-
lation, larch is found commonly only in swamps where its growth rate
is substantially less. The upland sites are normally taken over by
other species with which larch cannot successfully compete and which,
in turn, grow poorly or not at all in swamps. 'Given a choice between
extinction through competition or survival, larch adapted its require-
ments to live in swamps.

Surtees (1963a, b, and c; 1964a, b, c, d, e, and f) investi-
gated the spatial distribution of several species of grain weevils and
Tribolium (see the "Environmental Factors" and "Density" sections of
this review). The basic finding in all these studies was that when
groups of insects are confined at high densities a certain species-
specific percentage of organisms was found at the surface of the
medium. Coupling these density effects with what he had found in
relation to temperature and humidity, Surtees (1964c) developed the
following theory of spatial distribution. Mutual disturbance due to
individual movement elicits negative geotaxis and this plays an
important role in determining group dispersion. Although such dis-
turbances cause an upward movement in restricted experimental popu-
lations, this would result in the spread of a population in a larger

grain bulk firstly by upward movement and then by more random movements

53

resulting in colonization of hitherto uninfested grain. Aggregation,
on the other hand, is a result of environmental stimuli modifying the
generalized pattern of movement, including the irregularity of pathway
or intensity of turning (klinokinesis) and the speed of movement
(orthokinesis) which in turn reduces the disturbance factor and there-
fore dispersal pressure. Thus, aggregation may be expected in regions
of lowest velocity. This latter conclusion was also made by Gunn and
Pielow (1940).

However, Surtees seems to have ignored the literature before
him, particularly Naylor's (1959) and Ghent's (1963) papers on
Tribolium which Show that many of the responses Surtees describes are
due to conditioning effects. Of course, Ghent's (1966) paper tends to
negate many of Surtees' conclusions, which must be very tentatively
interpreted.

While research was progressing on the relationship between
biological conditioning and dispersion, other researchers were attempt-
ing to elucidate the components of conditioning. Park (1934, 1935,
1936) had described the physiological and ecological effects of con-
ditioning but had no idea what the effects were due to. Loconti and
Roth (1953) and Ghent (1963) showed that a major stimulus to dispersal
and aggregation in flour beetles is olfactory, related in some way to
the conditioning of the medium. Some possible components of this
conditioning are quinones, feces, casts, nutritional depletion of the
Inedium, presence of fungi or bacteria and presence of larvae and eggs,
zilthough the principal component was suspected to be quinones (Ghent

1963) .

s4

Ogden (1969) set out to determine the important components of
conditioned medium and to assess their relative importance in the
dispersion responses of Tribolium. He gave adults choices between
fresh flour and flour having some suspected component of conditioning.
He utilized 1, confusum and the conditioning was done by adults and
larvae (larvae do not produce quinones). The results are the following:

1. I, confusum adults are attracted to medium conditioned with a
series of adult densities as well as to medium conditioned by
m§g_mutants (m§g_mutants produce very little quinone).

2. Adults are attracted to larval conditioning and to medium
containing frass (larval excreta).

3. Adults are not attracted to medium containing eggs.

4. Adults are repelled by beetle quinones and give a variable
response to commercially prepared methyl paraquinone, a com-
ponent of beetle secretions.

It would appear that quinones are not the important component
of conditioning, which is in direct contradiction to earlier studies.
However, this debate is as yet unsettled. It is possible that quinones
in low concentrations attract adults and quinones in high concen-
trations repell them, causing dispersal when population density becomes
great. Ogden's (1970) paper however, does not seem to support this
hypothesis.

Ogden (1970) compared I, confusum and I, castaneum with respect
to dispersal, using Prus' (1963) apparatus described earlier. His
:results are consistent with early findings. I, castaneum disperses

:rapidly from homotypically and heterotypically conditioned medium,

55

whereas I, confusum aggregates in both. However, Ogden also found
quinones not to be important to these responses. It was found that
medium conditioned by a "little—quinone-producing—mutant" elicited the
same responses from I: gastafleum_and I, confusum.

My own studies with Galleria conditioning make no attempt to
assess the components of conditioning eliciting dispersion behavior,
although a tactual response to Silk is probably involved. This does
not, however, negate the possibility of chemotactic responses.

The effects of biological conditioning and animal-animal
interactions on spatial distribution might more clearly be seen in
larval populations free from adult-young and sexual interactions.
Lepidoptera populations, i.e., Galleria, provide us with such models
but, for the most part, have not yet been well worked out. One of
the earliest observations pertaining to these variables in a Lepidop-
tera population was made by Brindley (1910) on the behavior of the
processionary caterpillar Cnethocampagpinwora. The larvae seemed to
"march" in single file when on the move and formed a "circulating
mass" just before burrowing into the soil to pupate. This circulating
mass frequently formed and then dispersed. After massing, a new
procession developed with the old leader usually still leading.
Brindley concluded that in any procession one larva has a "greater
tendency to stray" than the others and this one is more likely to act
as a leader. (Wellington, 1957, extends this idea to tent caterpillars
as will be seen shortly.) These larvae are constantly laying down a
silk thread as they march in single file. Since Brindley could not

elucidate a function for this behavior and since these threads

56

contained nitrogenous products, he attributed the behavior to
excretion. Edwards (1910), however, noted that the single file pro—
cessions consisted of larvae arranged in head-to—tail contact. The
whole mass moves along a silken thread which is commenced by the
leader and added to by all the larvae in succession. The number in
the procession is variable, but can be up to 100 larvae in length.
Any larva could be a leader, but once a particular larva begins
leading it continues to do so. At night the procession returns to
the nest. Upon approaching the nest, the leader searches until
encountering one of the many silk strands emanating from the nest and
follows it with the procession behind him. The leader exhibits a
choice in this selection of a silk nest-strand in that it would not
select artificial strands. Edwards further found that procession
integration was helped by the silk strand but the predominant inte-
grating force was head to tail contact. The silk strands were more
important in relocating the nest after a feeding excursion, the silk
strands being predominantly used in finding the nest.

The spinning of silk strands or tunnels is a common phenomenon
in Lepidopterans, as well as other invertebrate orders such as Diptera,

i.e., black-fly larvae, and Trichoptera, caddisflies. Silk trails may

 

have much the same significance as scent-trails in mammals. Accumu-
lation of webs and silk trails over the foliage and ground or to and
from refuge nests may indicate how heavily an area has been travelled,
which in turn could be weighed against the available resources.
Perhaps the result of such behavior conditions the habitat in the same

manner as chemicals do. Wynne-Edwards (1962) says that behaviors are

57

a form of epideictic display. For example, he postulates that the
"circulating mass" behavior of the processionary caterpillars is a
mechanism supplying information on population density. This has yet
to be tested.

There is some evidence that the larvae of Galleria mellonella

 

(the greater wax moth) condition their medium by means of silk trails
and frass (Milum 1952). Newly hatched larvae at first feed on isolated
portions of honeycomb in a bee hive, or beneath the comb's surface
where they form silk tunnels covered with frass and bits of comb.
Gradually these larvae assemble in a mass of webbing, with tunnels
extending throughout the remainder of the combs in search of food.
Oertel (1962) conducted tests to find if Galleria adults are attracted
to old (larval conditioned) hives. He discovered that wax moths find
the conditioned combs in preference to non—conditioned ones and con-
cluded that odor from the old combs was attractive to them.

The social behavior of certain insects, such as ants, bees,
and termites, has attracted considerable attention. However, such a
"high-order" of behavior is restricted to very few species, the others
existing in states from pure solitariness to sub-social aggregations.
Lepidopterous larvae afford many examples covering the extremes of
this range. Those leading a solitary existence generally do so as
the result of the habit of the adult in laying eggs singly or in pairs
rather than as a distinct anti—social facet in larval behavior. In
other cases where eggs are loosely scattered or tightly clustered the
opportunity for aggregation will exist (Long 1955). In a series of

studies, Long (1955) set out to compare the "sub-social" behavior

58

concerned with aggregations in two species of caterpillars, Plusia

gamma and Pieris brassicae.

 

Pieris brassicae (the large white butterfly) lays its eggs in

 

tightly packed groups. The first larvae to emerge help the others to
hatch by eating the tops off of the eggs. Once a large hatch has
occurred, the larvae begin spinning silk which is of prime importance
in their aggregation formation. Eventually a particularly active
larva wanders Off to an adjacent leaf, spinning a silk trail as it
goes. The other larvae follow the trail. As they eat and spin the
whole area is covered with a mat of webbing, and once this mat is
formed future spinning is restricted to the mat or to fresh leaf areas.
Accompanying spinning behavior is a distinct activity rhythm. The
larvae feed and rest on common "mat sites." As they feed they chew a
circular hole in the leaf, extending the mat as the hole gets bigger.
After feeding they return to the mat and then push out again to feed.
This mass behavior seems to depend on a few initiating larvae and
their ability to transmit a sense of activity to other larvae. By
about the fourth instar, the large aggregation begins breaking up into
smaller ones, as a function of the food source, and individual feeding
rhythms are altered.

Larvae raised in isolation were able to join established aggre—
gations. In the first two instars solitary larvae could join an
aggregation by following a silk strand to the group, and were allowed
to join the aggregation only if their size was equivalent to the
individuals in the aggregation. By the third instar, however, soli-

tary larvae exhibited investigatory movements, whereas aggregations

59

remain passive. These investigatory movements interfered with the
normal pattern of group behavior and joining a group became difficult.
It became impossible for fourth instar isolates to join established
aggregations because their individual activity rhythm was asynchronous
with the group's.

Plusia gamma (the silver Y—moth) on the other hand, lays its

 

eggs singly or dotted on the undersides of leaves. Regardless of
this behavior, the eggs laid by a single female hatch in the span of a
day and the opportunity for aggregation formation is present. Unlike

Pieris brassicae, however, the larvae of Plusia gamma exhibit random

 

 

movements, no noticeable silk spinning behavior, and a definite non-
gregarious tendency. If given the proper circumstances, however,

Plusia gamma larvae behave similarly to Pieris brassicae larvae. If

 

 

given crowded experience they aggregate and eat more than isolates.
A larva that was reared in a crowd and then allowed to roam free in a
normal situation has a tendency to stop moving and begin feeding upon
encountering another larva. Isolate reared larvae do not show this
tendency, move more, and eat less.

Long concluded, therefore, that the large aggregations in

Pieris brassicae depend on silk spinning for its formation and partly

 

for its maintenance, whereas Plusia gamma has no such mechanism and

 

forms aggregations only upon chance meetings. In both species,
however, the ultimate factor in aggregation behavior was larval con-
tact. Larval contact could stop a wandering larva or start larvae

moving, although in Pieris brassicae the movements are both silk and

 

larval contact directed. There is also an age variable in relation to

60

aggregation formation, particularly in Pieris brassicae. Larvae not

 

of the same size (age) as those in the aggregation had difficulty
joining the aggregation predominantly because of asynchronous activity
rhythms.

Long's hypothesis about aggregations being a function of the
size of egg clusters is unfounded. There is a correlation between
eggs laid in clusters and those laid singly in the two species
examined, but a sample size of two is not statistically sound. The

larvae of Plusia gamma may not have formed aggregations because they

 

hatched from eggs laid spaced out, but rather because they exhibit no
spinning behavior which was the main cause of aggregation formation

in Pieris brassicae. To further test the relationship between egg

 

cluster size and aggregation behavior we would need a species that
hatches from eggs laid singly and whose larvae spin silk trails.

The relationship between biological conditioning and mobility
(activity) of organisms, as seen in Long's (1955) paper, is a fre-
quently occurring one in the literature. In Lepidoptera, this rela-
tionship has been most systematically elucidated by Wellington (1957,
1960) with populations of tent caterpillars. Wellington was able to

isolate four categories of tent caterpillar larvae (Malacosoma pluviale)

 

on the basis of behavioral orientation to a light source. Type I
larvae are active and orient well when isolated. Type Ila larvae are
active and orient only in groups. Type IIb larvae are less active and
disoriented in groups and as isolates. Type IIc larvae are sluggish
and seldom develop. Various colonies of these larvae were constructed

using varying percentages of each type and their behavior studied.

61

Wellington found that a colony consisting of more than 13% Type I
larvae, more than 29% Type I and 11a larvae, and fewer than 21% Type
IIc larvae were very active and constructed normal, elongated silk
tents.

Comparisons of these laboratory colonies were made with field
data on egg laying and dispersal into new areas and Wellington found
that each larval type served a needed function in the colony. Type I
larvae are capable of independent, directed movements even when iso-
lated. On twigs and foliage they are capable of invading new terri-
tory unmarked by the silk trails of other larvae, although they will
use such conditioning if it is present. Since they are the most active,
Type I larvae are the first to respond to climatic changes, or any
disturbing stimulus. Their movements stimulate some of the other larval
types to follow them to new locations or on food foraging excursions.
Wellington (1957) found that fed and vigorous Type I adults are largely
responsible for relieving population pressure in the original foci by
dispersing some of the next generation to more distant areas. Type I
larvae exert a similar effect by dispersing the colony before parasite
attacks become frequent. Type IIa larvae have a lower activity level,
are incapable of independent orientation, and will travel a straight
course in the presence of a Type I larva or a silk trail. They are
of little value in extending the range of colony activities. However,
they are a valuable component of a colony in that they help transmit
"activity waves" throughout a resting colony, thus insuring that feed-
ing intervals are not too long. Type IIb larvae are relatively

inactive and depend for their survival on the more directed types of

62

larvae or on silk trails for finding food. It is this type of larvae
that comprise the bulk of tent caterpillar colonies. Their presence
insures the survival of at least some of the Type I and Ila larvae
because (1) in tent construction they produce a denser, tougher mat
of silk, since their inactivity causes silk strands to become con-
centrated (a denser tent prevents colony desiccation and predation),
and (2) their presence helps the colony absorb losses due to pre-
dation or parasites, if these attacks are sporadic. Type IIc larvae
are the weakest members of the colony and probably never contribute
to the next generation. Their function in the colony is not well
worked out, but it is detrimental if they are present in large numbers.
Possibly they serve as the vanguard for picking up infections and
diseases. All Type II larvae help subdue the activities of Type I
larvae and keep the colony from becoming too active.

Comparisons of natural colonies or of artificial colonies
made up of various densities of larvae and various percentages of
larval types yielded similar results in that the make up of the colony
largely determines its mortality rate and dispersal ability. More
active colonies suffer many of their losses during the wandering of
small groups led by Type I larvae. The most sluggish colonies are
eliminated simply because they do not wander enough, but stay on their
tents to starve or fall prey to disease.

Studies of natural colonies (Wellington 1960) revealed a
cyclicity of density and types of colonies from 1956-1959. In each
case Wellington found that the initial decrease in active colonies

was a result of the deteriorating quality of the resident stock. That

63

is, the viable proportion Of individuals changed in favor of more
Type II larvae. Increases of active colonies was a result of the
elimination of the less viable and moribund sluggish colonies from
the pOpulation by extrinsic and intrinsic mortality factors. This
increase occurred only after a period of decreasing density.
Wellington's results indicate that Chitty's (1960) hypothesis
may be applicable to insect populations. Chitty (1960) hypothesized
that populations are capable of limiting their own densities through
deteriorations in quality that occur whether or not extrinsic
mortality factors are operating, and that such reductions in vitality
may lead to increasingly severe effects of some mortality factors,
such as weather, upon the weakened population. This hypothesis
assumes an "intrinsic weakness" within the population. The intrinsic
weakness in the case of tent caterpillars is the Type II larvae,
particularly Types IIb and c. The more active Type I larvae produce
Type I adults which fly off to oviposit. The less active Type II
larvae produce less active Type 11 adults which oviposit near their
birthplace. This gradual buildup of large numbers of Type II larvae
leads to sluggish, slowing, deteriorating colonies whose numbers far
outweigh the few active colonies established by active adults. After
an initial increase these colonies will die. Intermediate colonies
will also die out when their more sluggish members succumb (i.e., to
sudden weather changes) because they are not numerous enough or
sluggish enough to produce the necessary tent silk and mass of insulat~

ing bodies.

64

There are strong indications from these studies with Lepidop-
tera that, exclusive of environmental factors, spatial distribution
is dependent upon biological conditioning involving a preference on
the part of the larvae, some sort of social behavior be it sex or
animal-animal interactions, and the probability that such responses
are age dependent. My studies strongly support this view. However,
unlike the studies presented in this review, my studies have been
approached from the point of view that we must first understand indi—
vidual behavior before the myriad interactions within a population are

explicable.

CHAPTER II

BEHAVIORAL AND LIFE HISTORY CONSIDERATIONS OF THE
GREATER WAX MOTH, GALLERIA MELLONELLA (L.)
(LEPIDOPTERA: GALLERIIDAE)

 

Behavior and Life History

 

There is a rapidly accumulating literature on Galleria

mellonella, although it is a relatively new species to behavioral

 

studies. Galleria mellonella (L.) is of biological interest because

 

of its developmental patterns, food habits, ecological and behavioral
adaptations, and general adaptability as an experimental organism.
The larvae and adults have been used in studies of biochemical,
nutritional, and digestive processes (Haydak 1936, 1940; Niemierko
and Cepelewica 1950; Niemierko and Wlodawer 1950, 1952; Niemierko

and Wojtezak 1950; Zielinska 1952; Przelecka 1956; Dadd 1964, 1966;
Balazs 1958; Young 1961, 1964), physiology (Bell 1940; Wojtczak
1952a, b, and c; Pipa 1963; Edwards 1966; Harshbarger and Moore

1966; Alexander 1970), development (Paddock 1913; Andrews 1921; Chase
1921; Haydak 1940; Whitcomb 1942; Beck 1960), comparative arthropod
anatomy (Borchert 1933; El-Sawaf 1950; Milum 19523 and b), and com-
parative ecology and behavior (Paddock 1913; Plath 1934; Milum and
Geuther 1935; Haydak 1936; Milum 1952a and b; Dutky et 31. 1962;

Burkett 1962; Oertel 1962, 1963; Woolever and Pipa 1970).

65

66

The origin of Galleria mellonella is not well documented, but‘

 

it is probably of wholly Oriental origin (Paddock 1913). This orga—
nism was even known to writers of antiquity. Aristotle tells of the
injury they caused in bee hives. Today, however, Galleria has a world-
wide distribution (Paddock 1930; Imms 1948; Skaife 1954) and was
introduced into the United States sometime around 1805, shortly after

the introduction of the honey bee (Apis mellifera).

 

Galleria mellonella (L.) is a parasite in honey bee (Apis

 

mellifera) hives, the larvae feeding on the pollen, honey-laden brood
combs, and possibly bee larvae. It is of economic importance since
it is estimated that large losses to the honey bee industry are caused
by this and related species (Whitcomb 1942), particularly to combs in
storage. It is not yet clear how a bee hive first becomes infested
with Galleria, but once infestation occurs the hive is eventually
destroyed. There appears to be an indirect relationship between
"strength" of a hive and Galleria's ability to invade it. Paddock
(1913) states that if the bee colony has a vigorous queen and is kept
strong, Galleria cannot enter the hive to deposit eggs. Oertel (1962)
records a similar finding, and Whitcomb (1942) notes that weak,
diseased, starved, or otherwise abnormal colonies are prey to the wax
moth, and in such colonies the combs are entirely destroyed. However,
wax moth injury is secondary since strong colonies will defend them-
selves well against attack.

It appears that Galleria (larvae) is chiefly dependent upon
honey bee hives for food, but it has been reported to infest other

hives as well. Oertel (1962) found Galleria to be a pest of the

67

stingless bees of Brazil, and further states that Walrecht (1958)
found them in wasp nests (Vespa_and Vespula). However, Plath (1934)
and Sladen (1912) could find no evidence of such infestations in
bumble bee nests. Oertel (1963), on the other hand, found that under
certain conditions, bumble bee cells can serve as a host of Galleria.
He further hypothesizes that perhaps bumble bee nests help to main-
tain wax moth populations in the absence of honey bee combs.

The life cycle of Galleria mellonella is typical of holometab-

 

Olus insects, consisting of egg, larval, pupal, and adult stages.
There are several overlapping generations per year in the wild, and
over-wintering can occur as larva, pupa, or adult (Paddock 1913).
Each female moth lays hundreds of eggs in clusters. Paddock
(1913) reports the eggs to be elliptical in shape, measuring 0.48 mm
in length and 0.43 mm in width, and pearly white in color. Andrews
(1921) says the eggs are spherical or pear-shaped, depending on where
laid. Length of egg stage is variable, depending on temperature.
Paddock (1913) reports the egg stage to last 10-12 days whereas
Whitcomb's (1942) data indicates length of egg stage to be 5—8 days
at 24-27°C. In our laboratory we found the egg stage to last 5
days at 32:1°C. Lowering the temperature will delay hatching.
When first hatched the larvae are white and about 3 mm long
(Paddock 1913). Chase (1921) reports newly hatched larval length
to be 2 mm or less, which we have found to be more accurate. The
young larvae are extremely active, with prominent thoracic legs and
the ability to move forward or backward with equal rapidity (Milum

1952). Older larvae exhibit a range of colors from whitish to dark

68

gray, are about 18 mm long, have a small head and large body, and
have lost the rapid movements characteristic of young larvae. Older
larvae are also round in shape, whereas young larvae are dorsal-
ventrally flattened.

Paddock (1913) reports the length of larval life to be 35-40
days, and Chase (1921) and Beck (1960) state that larval development
consists of several instars. Development of the larvae is heavily
dependent on temperature, light, food source, and density. Accounts
in the literature differ because these variables differed or were not
specified, observations were taken in the field and the laboratory,
and what constitutes larval development was not operationally defined.
For example, the average duration of development (from hatching to
emergence as adults) has been reported to be 29.3:_l.5 days (Haydak
1940), 42-49 days (Metalnikov 1908), 90-120 days (Paddock 1913),
42-53 days (Andrews 1921), 33-54 days (Chase 1921), 62-64 days
(Borchert 1933), and from 28 days to 4 or 5 months (Whitcomb 1942).
In many of these cases the rearing conditions were not specified.
Whitcomb (1942), however, reared eggs at 85-95°F, which is about
the temperature normally found in an active bee hive. Milum (1952)
agrees with this data, saying that the average larval period is 28.85
days at 35°C. In our laboratory, however, we found the average length
of larval life (from hatching to cocoon spinning) to be 18-20 days at
32:}°C.

During the larval stage, larvae spin extensive "silken"
tunnels from modified salivary glands and mouth parts. These tunnels

are spun throughout their food source and become covered with frass

69

and food. It would appear that at no time during larval development
do they leave the food or the tunnels.

Galleria larvae exhibit a rapid growth rate under optimal
conditions and the size distribution between ages is extreme (Figure
4.1). Beck (1960) has demonstrated that, at 35°C and under continuous
darkness, larval weight doubles daily during the first 10 days of life.

One or two days before beginning to spin a cocoon, the larvae
enter the "wandering phase." During this phase, they leave the food,
probably stOp eating, and search for a pupation site. They continue to
spin silk during the wandering phase but no longer in discrete tunnels.
It has been hypothesized that the mechanism initiating this phase is
a cessation of eating, but this has yet to be tested. The wandering
phase lasts about two days, at the end of which time the larvae settle
down and begin spinning a cocoon.

The larvae usually pupate in cracks and crevices in the hive
walls although they may pupate in the refuse and webbing under the
combs (Paddock 1913). Whitcomb (1942) observed cocoons in the refuse
under the comb and in the mass of tunnels and refuse of wax on the
hive frames. Once a suitable pupation site has been selected, the
larvae spin a dense, tough silken cocoon, presumably utilizing similar
behaviors as those described by Van der Kloot and Williams (1953a) for
Cecropia, and pupate within this protective structure. Before changing
into a pupa, however, three flap-like slits are cut in an exposed end
of the cocoon to facilitate emergence as an adult (Milum 1952). The
duration of the pupal stage ranges from 8 to 62 days, depending on

temperature (Whitcomb 1942), after which time eclosion occurs and the

70

adult stage begins. Males eclose a few days before females, and
are generally longer-lived.

The adults are small, 15 mm long moths with a wing span of
30-32 mm (Paddock 1913; Whitcomb 1942; Milum 1952) and are dark gray
in color. The size and color of the adults is affected by the type
of food consumed by the larvae and by the length of the developmental
period (Whitcomb 1942).

There is probably no direct interaction between the adults and
larval or pupal stages. This hypothesis is supported by several
observations. Galleria adults die shortly after mating and have never
been observed caring for either eggs or larvae. The adults lack mouth
parts and do not eat (Oertel 1962). Their only function in life is
reproduction and invasion of new bee hives.

Galleria adults exhibit a sexual dimorphism (Whitcomb 1946;
Paddock 1913). Males are smaller than females, and the forewing of
the male is deeply scalloped while that of the female is straight
edged. The female also has two short, prominent pointed palps on the
front of the head, whereas the male lacks these palps (Milum 1952).

Mating occurs at night (Paddock 1913), but several reports
in the literature indicate mating occurs during the day as well. If
maintained in constant darkness, mating occurred at all hours of the
day in our laboratory. Once mated, a female lays anywhere from one
to several hundred eggs in the cracks and crevices of the bee hive.
There is, however, considerable conflict in the literature on just
how many eggs a female lays. Paddock (1913) says 1018, Whitcomb

(1942) says less than 300, and Milum and Geuther (1935) say an average

71

of 754 eggs per individual female. The stimulus for egg laying seems
to be copulation with a male. Non-mated females, however, will
eventually lay their eggs, which do not hatch.

How adults leave a hive to infest a new hive has not been
investigated. Once they have left a hive, however, a new hive is
probably invaded by the female laying her eggs in cracks and crevices
on the outer walls of the hive. The newly hatched larvae could then

work their way into the hive and reach the combs.

Gpinning Behavior and Preferences

 

Galleria larvae exhibit three distinct spinning behaviors:
(l) discrete larval tunnels, (2) diffuse webbing during the wandering
phase, and (3) discrete cocoons in which pupation occurs. There are
no studies dealing with these behaviors or with how the result of
such behavior affects the behavior or preferences of the other larvae
in the population. Such studies would be of value when one considers
that the larval phase of most holometabolus insects is the longest
phase and larval behavior could have significant effects on preceding
and succeeding generations. For example, the size and color of the
adults is a function of the degree of larval crowding and food source.
Developmental sequence and mating might be affected by larval behavior
and development. My studies with Galleria have shown that biological
conditioning, of which spinning behavior is a part, functions in the
spatial distribution of the larvae, and that larval preferences for

conditioning are a function of age and number of larvae present.

72

By definition, Galleria larvae biologically condition their
environment. Spinning behavior is an obvious physical conditioning,
but there may also be chemical conditioning from secretions and
excretions, such as frass. These biological conditionings, although
not called such, have been alluded to in the literature. Milum (1952)
found that newly hatched larvae feed on isolated portions of the comb,
forming silken tunnels with added frass and bits of comb. They
gradually assemble in a mass of webbing with tunnels extending through
the remainder of the combs in search of food.

Several functions for larval spinning behavior have been
postulated, but never tested. Paddock (1913) says that the only time
the larvae are at the mercy of the bees is just after they hatch and
are trying to chew into the comb. Once they chew into the comb they
make tunnels directed toward the bottom of the cells. These tunnels
afford protection and food for the larvae and also lead to their
desired feeding place, the center of the comb. When the center of the
comb is reached, the larvae leave their tunnels and wander over the
bottom of the cells or tunnel along the midrib from cell to cell. If
disturbed, they seek their tunnels for protection. Eventually a
massive silken gallery has been spun which gives almost complete pro-
tection from the bees. From this central gallery feeding is extended
out along the bottoms of the cells or the middle of the comb. The
silk is spun wherever the larvae go so that soon the bottoms of the
cells are replaced by a layer of silk threads covered with excrement
and particles of chewed wax. As outward feeding continues the threads

of silk are extended to cover the new feeding grounds and not only

73

serve to protect the larvae, but also act as a scaffold to support the
damaged cells. Therefore, the feeding area gradually extends from
the point of infestation to the entire comb, and, if the comb has
insufficient food, they feed in the refuse beneath the comb. These
feeding larvae consist of overlapping generations and all stages are
found feeding and spinning at any one time.

These observations suggest that larval spinning functions as
(1) protection from the bees, (2) scaffolding to support damaged cells,
and (3) a mechanism to direct feeding behavior. Implicit in these
observations is the idea that the silk tunnels may be involved in the
spatial distribution of the larvae. The fact that all larval ages
are present at any one time also raises the question as to what inter-
actions might be occurring and whether there is preferential use of
the silk galleries according to age.

The hypothesis that larval tunnels serve as protection from
the bees is particularly interesting in light of Free's (1961)
article on the stimuli releasing the stinging response of honey bees.
Free discovered that bees preferentially sting dark colors, rough
surfaces and rapidly moving objects. Galleria larvae move in rapid,
jerky motions, but in their light, smooth tunnels their movements may
not be detected.

There is no literature on spinning behavior or preferences
during the wandering phase. The assumption appears to have been made
that a wandering larva is oblivious to all else during this phase

except finding a suitable pupation site. The mechanism(s) for the

74

transition in spinning behavior from tunnels to webs and from webs to
cocoons has not been investigated.

There are several studies dealing with cocoon spinning behavior
in Lepidoptera. Van der Kloot and Williams (1953a and b) discuss the
role of the internal and external environments in relation to cocoon
spinning in Cecropia. Certain preferences were found in relation to
location and position of the cocoon. However, the initiation of
spinning does not invariably depend upon the presence of an external
environment which would allow the construction of a normal cocoon.
Caterpillars will spin in environments where it is impossible for the
organism to construct more than a single sheet of silk. The longer a
silkworm is kept in an inadequate environment, the greater the likeli-
hood of its beginning to spin. It appears that there is an increasing
internal factor, acting as a stimulus for spinning, which lowers the
threshold for external stimuli to a level where the larva will spin
in a normally inadequate environment. Van der Kloot and Williams'
experimental results also emphasize the importance of tactile and
proprioceptive stimuli in cocoon spinning.

Some data on Galleria demonstrate preferences for cocoon
sites. Whitcomb (1942) records that when larvae spin cocoons they
firmly attach themselves to the sides of the hive, the frame, or some
such solid object. Some larvae, however, prefer tunnels and hive
debris as cocoon sites. Woolever and Pipa (1970) experimentally
investigated the spatial and feeding requirements for pupation of
last-instar larvae. They discovered that emergence of pupa from LSU's

(the first stage of cocoon spinning) can be significantly delayed by

7S

depriving the larvae of "sufficient" free space. Last-instar larvae
also require a minimum of two days' exposure to standard diet before
metamorphosis will occur.

Anything affecting the natality, mortality, or dispersion of
the larvae will have consequences to the species as a whole. Since
Galleria is a young "society," that is there are no adult-young inter-
actions from which the larvae might behaviorally learn, it would be
advantageous for the larvae to have the ability to pick and choose
those biotic and abiotic cues best suited for its development and
success as a species. There are indications of this ability in the
literature, particularly with reference to variables affecting
developmental time, natality, and mortality. My studies indicate that
the relationship between Galleria's preferences for biological con-

ditioning and dispersion is a close one.

Food Preferences

 

Certain areas in a bee hive have better food sources than
others. Older hive combs are often deserted by the bees and not
maintained, whereas areas in which the bees are actively rearing their
own larvae will be rich in honey and pollen. In an already infested
hive, some areas will have been conditioned by previous Galleria
generations, while others will not. Such conditioned areas may
(1) indicate an area already well worked over or heavily populated,
or (2) such areas may signal an immediately usable food source accom-
panied by immediate protection. There are indications in the liter-
ature that Galleria larvae exhibit preferences in selecting a food

source on which to develop.

76

Beck (1960) reported on an artificial dietary medium for

laboratory rearing of Galleria mellonella larvae which yielded optimum

 

growth rate, as measured by the number of instars. It had earlier
been reported that larvae are capable of utilizing beeswax as a source
of dietary lipids (Dickman 1933; Waterhouse 1959). Yet Haydak (1936,
1940) found beeswax not to be a required nutrient, since growth and
maturation occurs in its absence. This appears to be true, but Beck
(1960) notes that if beeswax is included in the diet, larval growth is
significantly improved. Beeswax also has an effect on the physical
consistency of the medium, rendering the diet somewhat less compres—
sible. I found that larvae avoided diets lacking wax. However, it was
not determined if this avoidance was due to some missing nutrient or
to the physical consistency. Paraffin has a similar physical effect,
but less of a stimulating effect on growth. Since paraffin is indi-
gestible, its stimulating effect was attributed entirely to the altered
physical properties of the diet, and the better growth on beeswax
diets was attributed to some added nutrient (Beck 1960). Young (1961)
found that beeswax and some fatty acids enhance larval growth.
Niemerko and Wlodawer (1950) reported that the excreta of the larvae
contains 60-65% fatty acids. This percentage is similar to that found
in beeswax, an indication that the larvae eat the wax. However,
undigested wax is a major part of the fecal matter.

Whitcomb (1946) felt that larvae get nourishment from
impurities in the wax. Chase (1921) found that many of these impur-
ities are from bee and Galleria frass, dead moths, and various larval

casts. Much earlier, Braun (1867) maintained that wax moth larvae

 

 

77

devour combs in order to obtain pollen and other nitrogenous sub—
stances, such as cocoons and excreta. Most of this research on larval
nutrition lack tests of actual larval preferences for various diets.
Chase (1921), however, experimented with several food sources. The
one she found to be most successful was a mixture of chopped-up comb
wax, bee bread, soft dead moths, and honey kneaded into a small loaf
from which slices were cut. Newly hatched larvae first selected for
food the dead moths, and as development proceeded their preference
switched to the comb. Andrews (1921) found that larvae could not be
raised solely on dead moths, which may account for the switch in
preference noted by Chase.

Haydak (1936) hypothesized that Galleria larvae are seeking
pollen or other materials and that burrowing in wax is really a
secondary, acquired characteristic which serves as self-protection.
However, his experiments did nothing to shed any light on this
hypothesis. He raised larvae on several diets, but only in those diets
containing dried yeast did the larvae produce large quantities of
silk. Growth was also greatest on yeast diets. Larvae reared on
non-yeast diets produced very little silk and covered their sparse
tunnels and cocoons with frass. (It is interesting to note that the

lesser wax moth, Achroia grisella, normally spins only sparse tunnels

 

and exhibits the behavior of covering these tunnels with frass and

other debris.) The low silk production on deficient diets may be a

function of insufficient amino acids essential for silk production.
The quantity and quality of larval food affects certain adult

morphological characteristics. Slowly developing larvae, as a result

78

of poor nourishment, low temperatures, or high humidity, transform
into small adults. Larvae fed on dark brood combs transform into
adults that are dark gray or almost black. Larvae reared on other
deficient sources have been recorded to produce small adults ranging
from silvery-white to gray in color, depending on the food source
(Paddock 1946). We noted in our laboratory that overcrowded larval
colonies are slow to develop and produce tiny adults. This is
probably a function of the food supply becoming deficient, or might
also be related to substances being added to the medium by the crowded
larvae.

The larvae of Galleria mellonella (the greater wax moth) and

 

Achroia grisella (the lesser wax moth) both infest honey bee hives,

 

although they occupy ecologically dissimilar niches. The larvae of
these two species are virtually indistinguishable. Milum and Geuther
(1935) stumbled upon a possible method for distinguishing between
these species, based on food preferences. Young Galleria larvae will
develop on brood comb, but not on comb honey. Older larvae will
develop on both. Achroia larvae appear to develop on almost anything.
Although not so stated, it would appear that Galleria will preferen-

tially select brood comb when in its early instars.

Rearing and Maintenance of Stock Cultures

 

Stock cultures of the greater wax moth larvae were maintained
in 150x75 mm crystallizing dishes. The dishes were covered with a
circular piece of plexiglass with a 15 mm hole in the center. The
plexiglass cover was held in place with two pieces of masking tape,

one on each side, and the center hole was covered with tape. This

79

rearing container was originally patterned after a similar one used
by Beck (1960). Beck, however, plugged the central hole with cotton
and held a glass top on with a layer of vaseline. These procedures
proved rather cumbersome. The vaseline layer was intended to prevent
larvae from escaping. However, behaviorally, larvae exhibit a down-
ward movement, and since the vaseline melts in an incubator, I found
masking tape to be more efficient. The cotton plug used by Beck was
for ventilation. However, I found that when pupation occurred the
larvae either pupated in this plug or easily escaped through the hole.
Since sufficient ventilation occurred between the plexiglass top and
the crystallizing dish, the central hole was covered with masking
tape. This arrangement also prevented water condensation from forming
inside the dish.

About 200 grams of dietary medium was added in a loose layer
to each dish. Larvae were then added, the top secured, and each dish
placed in an incubator. The larvae were raised in total darkness at
32:1°C and 75% R.H. (Egg, larval, pupal, and adult stages were all
reared in incubators under identical conditions).

The dietary medium (Table 2.1) was a 3%, by weight, wax diet
based on the diet used by Beck (1960). This diet was prepared as
follows: The beeswax was melted on a small hotplate. While the wax
was melting, the pablum and yeast were weighed and mixed together.
Honey, water, and glycerin were weighed as a unit, added to the pablum
and yeast, and thoroughly mixed. The melted beeswax was then mixed in.
This mixture was usually made in 1000 gram quantities and then equally

divided among five crystallizing dishes for colony use.

80

Table 2.1.

Artificial Dietary Medium for Laboratory Rearing of the Greater Wax
Moth, Galleria mellonella (L.)

 

 

——-.

 

 

. Amount
Constituent (G.)
Honey 24.25
Glycerin (U.S.P.) 21.34
Water 8.73
Pablum, mixed cereal 32.98
Active Dry Yeast 9.7
Beeswax (with impurities) 3.0
100.00

 

The colony was maintained as follows: Adults were put in a 5
gallon fish tank with a plexiglass top. Several pieces of wax-coated
weighing paper were folded, accordion-fashion, and placed in the tank.
Females laid their eggs on these papers which could then be retrieved
at will. As new adults eclosed in the colony they were added to the
tank so as to maintain a constantly breeding colony. To remove the
eggs, the breeding tank was placed in a refrigerator for five minutes.
This slowed adult activity so the top could be removed in order to
retrieve the eggs without the adults escaping. The addition of
adults was also done at this time. These procedures were performed at
8 AM every day, and the breeding tank was then returned to the incu-
bator with fresh egg papers.

The eggs were cut off of the paper and put in a 100x15 mm
.Plastic petri dish. These dishes consisted of two petri dish bottoms
tllped one on top of the other. This was done to allow extra space for

létrge egg lays and to prevent newly hatched larvae from escaping.

'Theese egg dishes were devoid of food so that newly hatched larvae

81

could be seen and time of hatching easily determined. Newly hatched
larvae are too tiny to be found in food, which would make hatching
time impossible to calculate. Hatching time was needed to determine
future larval ages and to synchronize the colony. Once the eggs had
been cut off the paper and put in these dishes, the dishes were placed
in the incubator.

As soon as eggs began hatching, the contents of the petri
dish were put in a crystallizing dish containing food, and left-over
eggs were removed 24 hours later. Due to this procedure, the larvae
in any particular crystallizing dish consisted of larvae hatched in a
24 hour period of time. For example, one—day-old larvae are actually
1 day old 1.12 hours, and so on for all ages. The number of larvae in
any one dish was variable, depending on the size of the hatch.

When larvae in the crystallizing dishes entered the wandering
phase at about 16 days of age, as indicated by their tendency to move
upward in the dish, they were culled. About 100 larvae were removed
and placed in a new crystallizing dish with fresh food, where pupation
occurred. Pupation appeared to last 7-9 days, although this was not
precisely measured. This procedure was followed because it was found
that adult eclosion was delayed if the culture dishes in which the
larvae pupated were crowded.

When the adults in the new crystallizing dishes emerged they
were put in the breeding tank and the cycle continued.

The initial larval stocks of Galleria were obtained from a

10¢:al fish bait wholesaler, Victor A. Wilicht of Bath, Michigan. In

an.:attempt to keep inbreeding at a minimum, fresh stocks of 500 larvae

82

were purchased from this dealer once a month and mixed with our estab-
lished colony. Of course, all animals were from the same original
colony so some inbreeding occurred. However, since hundreds of adults
were being bred at any one time, inbreeding was probably at a minimum.
As a check on whether these procedures were resulting in behavioral
changes, periodic behavioral tests (controls) were run. These tests
consisted of asking whether larval responses to biological conditioning
changed with time. The results showed that behavioral responses to
conditioning remained constant. These tests are discussed in Chapter

IV.

CHAPTER III

HOTELLING'S T2 STATISTIC FOR MULTIVARIATE ANALYSIS:

A "NEW" APPROACH TO LONGITUDINAL STUDIES AND
SERIALLY CORRELATED DATA

Introduction

 

Scientists have long been perplexed by the appropriate means
to analyze phenomenon in behavioral development, the major problem
being how to analyze developmental data and satisfy the assumption of
independence. Classically two kinds of experiments have been used in
developmental studies. One is the cross-sectional experiment which
measures different individuals at each age or time period. Although
this approach meets the assumption of independence of measurements, it
is not an appropriate method for analyzing individual behavioral
changes over time. It also often necessitates a small sample size at
each time point, particularly if the time series is a long one. The
Second approach is the longitudinal study in which an individual is
repeatedly measured throughout time, with the result that each suc-
ceeding measurement incorporates the preceding ones. 'This is an

aFHIropriate procedure for looking at changes in individual behavior
oVer time but may violate the assumption of non-serially correlated
‘aartau It is the longitudinal study which is of concern in this paper.
Biologists seem well versed in statistics when it comes to

‘3cnisidering the assumption of normality and homogeneity of variances

83

84

in their statistical procedures, but serial correlation is another
matter. To avoid violating the assumption of independence of the error
terms underly-all parametric statistics, biologists have typically done
one of the following with longitudinal data:

(1) They ignore the statistical assumptions and analyze their data
as if measurements had been independently collected. As Gill
and Hafs (1971) point out, such a procedure leads to misinter-
pretation of results because of correlated error terms.

(2) Often a "grand mean" is calculated from the raw data and
analyses are carried out on these means. This procedure
eliminates the problem of independence but it also throws
away much valuable data. It also may lead to incorrect con-
clusions since the analysis may conclude that two groups are
not different on the basis of their means, when in fact the
curves for each group are very different.

(3) Biologists often circumvent the assumptions in parametric
statistics, except for independence, by resorting to non-
parametric analyses. However, most non—parametric statistics
are not very powerful or sensitive.

The objective of this paper is to present a statistical pro-
CEdure for the analysis of developmental data by treating time as a
mlliltidimensional random variable and discussing the statistical con-
‘Sjlierations necessary before drawing inferences from such data. This
‘VilJl be done by first discussing the test statistic and then the

aSsumptions underlying it and how these assumptions may be tested.

85

The method to be presented is also valid for conventional multivariate
analyses drawn from cross-sectional data.

In 1931, Hotelling presented a generalization of the familiar
univariate Student's t-test on means of normal populations to tests
involving the mean vector of responses drawn from the multivariate
normal population. Hotelling's procedure was originally intended for
situations in which one wishes to ascertain whether two samples of
organisms with p measurable characteristics arose from the same popu-
lation. For example, one may wish to ascertain if body weight, ear
length, and tail length are the same in two populations of mice given
different treatments. Such data may be analyzed by comparing each
group to an expected mean vector or by comparing the groups to each
other. These procedures have come to be known as the "One" and "Two"
Sample Hotelling's T2 Statistics for Multivariate Analysis, respec-
tively, and both utilize a matrix algebra approach to their solutions.

For developmental data, however, the time scale is of interest.
By modifying Hotelling's procedures such an analysis is possible. The
modification consists of replacing the "p measurable characteristics"
With "t measurable time units" and treating time as a multidimensional
variate. By coupling this modification with the Durbin-Watson test
for serial correlation and Box's Test for heterogeneous variances,

all1 of the assumptions of a Multivariate Analysis of Variance may be
teSted.

Morrison (1967) presents a detailed discussion of Hotelling's

2
IF procedures from which the formulas in this paper were taken, but

the basics will be presented here, particularly as they relate to the

86

time modification. However, Morrison does not deal with the assump-
tions underlying this test and these will be thoroughly discussed.
The procedure for the time modification was first brought to my
attention by Dr. James Asher who utilized such an approach in a study
of parthenogenesis and genetic variability (Asher 1970).

The nomenclature in this paper will consist of the following:
Any letter with a wavy line under it refers to a matrix, i.e., A or a.
Entries within a matrix will be designated with lower case, subscripted
letters, such as xij for the ith row and jth column of x. The number
of rows in a matrix will be designated as N, which during the current
discussion will also be the number of replications. The number of

columns in a matrix will be designated as t, which is also the number

of time units.

Hotellingis One Sample T2 Statistic

 

The procedure for the one sample test is presented in flow
chart form in Figure 3.1.

Let the response variate 5 consist of t components distributed
according to the multinormal low with mean vector u and nonsingular

~

Covariance matrix 2. The null hypothesis to be tested is:

 

 

 

 

 

 

 

 

- . _ j _ _ _ _
“1 “ol “1 “oi
H . . _ . and its . f .
o' . . alternative is: H1 . .
Luts Diet- LUL Luot

III Irractice the population mean vector u and the entries in the popu-

1ation variance-covariance matrix are unknown.

87

Input (Raw data) Matrix A
~ “N\“‘-\l§ Durbin-Watson Test for
(L Serial Correlation
Mean Vector (X) é}"’”””flflflflflflg’r’

Find B X

Find transpose of (B‘)

l

Sums of Squares Matrix C = B'B

~

Covariance Matrix 8 = (l/N-l)C

Find Correlation Matrix

(B)
Inverse of covariance matrix (8-1) \L

Test Ho: r.. = 0
1]

(t-tests on entries in R)

Calculate T2 = N(X-XO)'S-1(X-Xo)
J/ N—t 2
F ' t(N—l) T
Hypothesis Testing
Ho: u = no
H0 is not rejected if H0 is rejected if
2 t(N-l) 2 t(N-l)
T < _.___._ - - ___.___ - _
__ N_t ‘17. t, N t T > N_t Fa. t, N t
CCHTfidence Regions on Mean Vector Simultaneous Confidence
(Ellipsoid intervals) Intervals for Linear

Compounds of Means

Fig. 3.1 2
Hotelling's One Sample T Statistic.

88

The calculations necessary to test the null hypothesis are
derived from a matrix of N independent observation vectors and t time

units. Let A be such a raw data matrix, such that

l . . . t
r" 7
1 a11 . . . a1t
A = .
N aN1 . . . aNt

 

 

and each entry in the matrix is some measurable characteristic of
interest over time. It is necessary that N > t since the degrees of
freedom for the test statistic are t and N-t, and if N f-t this test
cannot be made.

1

From this matrix are calculated the X, B, B', C, S, and S7

matrices, where

2 X!
II

the mean VBCtOI‘ Whose entries are the means calculated

from each column in A. For example,

X = [a1, a2, . . . at] / N

~

B = the matrix resulting when each entry in X has been sub-

tracted from each entry in the corresponding columns of A.

1 . t
1 311 a1 alt at
B = .
N aNl-al aNt-at
__ .__J

 

 

89

B' = the transpose of B

C = the sums of squares and cross-products matrix calculated
as 13'?

S = the variance-covariance matrix calculated as C/N-l

8'1 = the inverse of S

For the one sample test it is necessary to be able to determine,
a_priori, an expected mean vector (30) against which the calculated
mean vector (i) is to be compared.

The test statistic is

2 -l - -

= -—- ' _
T N(§ 50) (§ §o)°

2CD

When the null hypothesis is true, the quantity

_ N—t 2
F ' t(N-1) T

has the F distribution with degrees of freedom t and N-t.

2 t(N—l) - -
< ———.——— - =
If T __ N-t Fa, t, N-t’ Ho. 5 50°
2 t(N-l) . - -
If T > N-t Fa; t, N-t’ Ho. 5 # Eo'

If the null hypothesis is not rejected, it is possible to calculate a
confidence region for the mean vector as follows:

(N-l)t
N-t a; t, N-t

N(E-go)'§-l(g-Eo)‘:
This confidence region is represented by a single number which T2
“Knlld have to be bigger than before the null hypothesis could be
rej ected. This value will change as N and t fluctuate, or as the
0“level is varied.

If the null hypothesis is rejected, simultaneous confidence

11l‘tervals for linear compounds of means are calculated as follows:

9O

- 1 (N-l)t
: _ | ._____ v 1
s25 \Igeée [it Fa, 1., NJ 5.211153,
1 , (N-l)t
JET? §§ [N—t F01; t, N—J

where, a = a vector the experimenter specifies, and

~

 

+-

Z><I

 

(N-l)t _ 2
N-t Fa; t, N-t - T a; t, N-t'

For example, if we wanted to look at the confidence interval
around entry 51 of the mean vector, then a = [l 0 0 . . . 0]. Such a
procedure enables us to determine which of the responses have probably
led to the rejection of the hypothesis on the mean vector. The
probability that all such intervals generated by different choices of
a are simultaneously true is l-a (Morrison p. 121).

If one wishes to determine whether the means are correlated,
a correlation matrix may be calculated from the variance-covariance
matrix. This is done by pre- and post-multiplying the variance—
covariance by a diagonal matrix whose entries are l/JOE- along the
diagonal and zeros on the off-diagonal positions. The resultant
matrix is the correlation matrix with ones on the diagonal and corre-
lation coefficients on the off-diagonals. Each off—diagonal element
can then be tested for significant correlation by calculating

Sr ='J(1-rij2) / (N-2) (see Sokal and Rohlf 1969, p. 516) and com-
13'
Intring the result with an expected calculated value from Table Y of

ROhlf and Sokal (1969).

Example 3.1.—-An example will help clarify the single-sample
techniques. Ten isolate larvae were given a two choice situation for
thl‘ee days, the choices being between a biologically conditioned and
anunconditioned food lump. If the conditioned lump was preferred
‘1 31 was recorded, otherwise a zero was recorded. The results are to

‘3 tested against an expected vector whose entries are all 0.5,
aSSuming a random response.

91

The raw data matrix (A) is

t1 t2 t3

1‘_1 1 T

1 1 1

1 1 1

1 0 1

A: 1 1 1
” 1 1 0
1 1 1

0 1 1

1 1 1

10L1 1 1

 

 

The sample variance-covariance matrix of responses is
7.100 —0.011 -0.01T

S = -0.011 0.100 -0.011

 

 

-0.011 -0.011 0.100

We wish to test, at the a = 0.05 level, the null hypothesis that the
observations came from a population with mean vector

30 = [0.5 0.5 0.5].
The quantity (:1 - 130) = [0.4 0.4 0.4].
The test statistic is
T0.2795 1.2705 1.270;“

2
T = [0.4 0.4 0.4] 1.2705 10.2795 1.2705 0.4

 

 

 

 

1.2705 1.2705 10.2795 0.4J

61.714286, and

(IO-3)

15.9994, and d.f. = 3, 7.

Since this is a two-tailed test, the tabulated critical value
a/Z; 3 7 = 5.89. Since the observed F is far in excess of the
tabular, ’
'th
111

(Df F

ed P, we conclude that the null hypothesis is untenable and
at the responses arose from a population of larvae with considerably
gher responses to conditioning.

92

Since the null hypothesis has been rejected we would want to
look at the simultaneous confidence intervals for linear compounds of
the means. Let us calculate one such interval for the second mean of

the calculated mean vector i = [0.9 0.9 0.9].
Let a = [0 l 0].

Then,

1 . (N-11t -
N a Sa [:N-t Fa/Z; t, N{£] - 0.4766

and the interval for the second mean is

 

0.4234 5 0.9_: 1.3766.

This result says that if we rejected H because of a deviation
in the second mean then any mean would have toobe outside of this
interval to be considered different. By regulating the entries in
vector a we can set the variance attributable to any mean and ask
similar’sets of questions.

If the null hypothesis had not been rejected we would have
calculated a confidence interval for the mean vector.

Then,

- - -1 - - .(10-1)3
N(§'§o)'§ (5-50) 1 10-3 F01/2; 3, 7 < 22'7183'

. Therefore, if TZ-i 22.7183 the null hypothesis is not rejected.
Sane our calculated T2 = 61.714286, we rejected the null hypothesis.

From the variance—covariance matrix we can also calculate the
Correlation matrix which turns out to be

F_" _-1
1 —0.111 -O.111
R = 1 -O.111

l

 

 

“”1 From the correlation matrix a matrix can be calculated to test
c.ether the observed correlations are significant. This is done by
alculating S = 2 for each of the off-diagonal
rij (l-rij) / (ll-2)

 

erltTies and comparing that value with Table Y. For example, let us
00k at entry r12 of R.

\[[l-(-0.lll)2]/(10-2) or

0.3514.

 

(I)
ll

12

93

From Rohlf and Sokal's (1969) Table Y we find that with one
independent variable and d.f. = 8 a critical value of 0.632 at the
a = 0.05 level. Since our calculated value is less than the critical
value we cannot reject the null hypothesis. In other words, the
means are not correlated.
Hotelling's Two Sample T2 Statistic

Hotelling's Two Sample T2 Statistic (Figure 3.2) is an elabo-
ration of the one sample test. Two response mean vectors are compared
to each other, rather than comparing a mean vector with an expected
vector. Two raw data matrices are required, both of which have mea-
surable time units. The test statistic, however, utilizes a pooled
variance-covariance matrix derived from the two sums of squares and
cross products matrices.

Suppose that two independent random samples of observations on
a multidimensional variate (i.e., time) have been obtained from two
eXperimental treatments. It is to be assumed that the variates have
a multivariate normal distribution with the same, although unknown,
CoVariance matrix § of rank t. However, the distributions may not
necessarily have the same location parameters and we desire a test of
the null hypothesis that the population mean vectors are equal,

“°‘ E1 = B2
as QRposed to the alternative hypothesis

”1‘ E1 * E2

The mean vectors and sums of squares and cross products
matriCes are computed from their respective raw data matrices and a
IPooled Variance—covariance matrix is calculated. The following table,

taken from Morrison (1967, p. 125), summarizes these procedures.

94

Input (Raw Data) Matrices A and A
7 ~ Durbin-Watson Test
(on each matrix) for

{ serial correlation

 

Mean Vectors XI and X

Find B = Al-Xl and B = A -X

1

' I '
Find Bl and 82

Find Sums of Squares and Cross Products Matrices

= v . = 1
91 51 51’ 92 52 92

Calculate a Pooled Covariance Matrix
§ = (91 + 92)/N1 + N2 ' 2_________> Box's Homogeneity of
Variance Test
M=N LogeISI-
Inverse of Covariance Matrix (S-l) éf”””’ Z(Vk LogelSkl)
~ k

1

Calculate:

2 = N1N2

N + N

\L 1 2

Calculate:

N1 + N2 - t 2

= T
(N1 + N2 - 2)t

i

Hypothesis Testing
°‘ 51 = 52

_ _ -1 _ -
_ 1 -
T (5 52W (E15)

F

 

 
  

H0 is not rejected if H0 is rejected if

 

T2 < (N1+N2-2)t 2 (N1+N2-2)t
~—_________ >

_ N1+N2-t'1 Fa; t, N1+N2‘t-1 T N1+N2-t-1 Fa; t, N1+N2-t-l

ElliPSOid Confidence Intervals Simultaneous Confidence Intervals
or the Mean Differences for Linear Compounds of Mean
Differences
Fig. 3.2 2 Test for Parallelism of two Curves.
Hotelling's Two Sample T Ho: Cu = Cu

~~ ~~

Statistic.

95

Table 3.1

Components of the Two Sample T2 Statistic

 

 

 

 

 

 

 

 

 

 

 

Sample 1 Sample 2

Sample Size N1 N2
Mean vector X1=[x11, . . . Xlt] X2=[x21, x2t]
Matrix of sums of C C
squares 8 products ~l ~2
Pooled covariance 1
matrix § — N1 + N2 2 (91 + 92)

The test statistic is
N N
2 _ l 2 — — , -l - -
T ’ N1 + N2 (51'52) § (51'52)

and the quantity

N1 + N2 - t - l 2

= (N1 + N2 - 2)t T

 

F

has the F distribution with degrees of freedom t and N1 + N2 - t - 1.

It is important to note that the two raw data matrices being

cOmPaI‘ed must be of equal dimension, that is N1 = N2 and t1 = t2. If

the dimensions are unequal it becomes impossible to determine the

degrees of freedom for the test statistic, or to obtain the pooled
(N + N - 2)t
nce matrix. If T E'Nl + N2 _ t _ 1 Fa; t, N1 + N2 _ t _ 1,

tfm:nu11 hypothesis is not rejected, and the confidence region for the

 

n1 . .
ea“ dlff'erence is

96

 

_ _ 6) _1 _ _ 0) N1 + N2 ’(N1 + N2 - 2)t
(X—x-'S(x-X- _<_———-—-—— F.
-1 ~2 ~ ~ ~ ~2 ~ NlN2 N1 + N2 - t - 1 a, t, N1 + N2 -t—l,

where 6 = the population mean of (Kl—X2). If
> (N1 + N2 - 2)t F

N1 + N2 - t - l a; t, N1 + N2 - t - l,
rejected and the simultaneous confidence intervals about the vector

2

T the null hypothesis is

 

of mean differences is

 

N +N (N1+N2—2)t

 

 

 

 

 

 

- - 2
a'(X -X ) - a'Sa F . < a0 <
~ ~1 ~ ~ ~~ NlNz N1+N2"t-l (I, t, N1+N2"t"1 — ~~ —
N +N (N +N -2)t
- - l 2 1 2

a'(X -X ) + a'Sa F .

~ ~ ~2 \ ~ ~~ NlN2 N1+N2-t-l a, t, N1+Nz-t-l
where

(N1+N2-2)t F = T2
N1+N2-t-l a; t, N1+N2-t-1 a; t, N1+N2-t-l.

A correlation matrix may be calculated as in the one sample
Case, except now it is calculated from the pooled variance-covariance
matrix. By pooling the two sums of squares and cross products
matrices we have assumed that we have homogeneity of variances and
thus need not calculate a correlation matrix for each covariance
matrix. This assumption will be dealt with shortly.

If the null hypothesis is rejected, a profile analysis may be
performed on the two response mean vectors, by asking if the popu-
lation mean profiles are similar, in the sense that adjacent curve
segments are parallel. In behavioral development, for example,
parallel curves would indicate similar behaviors, even if at different

-leVels. The null hypothesis is Ho: Cul = Cuz.

97

The test statistic is

 

N N
2 _ l 2 - - -l - -
T - m— Q‘.1‘§2)'9'(E§§') 9951252)
1 2
where
C = a (K-l) x k patterned matrix, having the effect of sub-
7 tracting each row in the matrix from the preceding row.
If T2 5' N1+N2-t Fa: t-l N +N -t the null hypothesis is not
(N1+N2-2)(t-1) ’ ’ l 2 rejected.

An analysis specifically for repeated measures has been worked
out for Hotelling's procedure (Morrison 1969, p. 133) in which serial
correlation is handled internally by the test statistic. However,
this procedure has not been worked out for the case of two samples,
has no profile analysis and is limited in scope as to the questions
it will answer.

The maximum number of experimental groups that can be analyzed
bY‘Pkrtelling's two—sample test is two. This means that multiple
comparisons must be performed for more than two samples, and presents
theProblem of determining an appropriate a-level for testing the null
hYPOthesis. This is up to the experimenter, but must be decided 3
331231, If in 100 comparisons, for example, 20 tests turned out to be
Significant at the a = .05 level, then 5 of those tests would be
eXPeCted to be significant on the basis of chance alone. The a—level
ShOHLd therefore be set at most at the a = 0.01 when such comparisons
are attempted.

There are several multivariate tests in the literature which
allow Many treatment groups to be simultaneously compared. There are,

'however, two problems with such tests. Firstly, they are much less

98

powerful than Hotelling's T2 test even when the a-level is adjusted.
Secondly, if the results of such a test indicate significant dif-
ferences, the experimenter has no way of knowing which treatment
groups caused the difference. In such a case the only recourse would
be pair-wise comparisons to pull apart the differences.

Example 3.2.--In the experimental design of Example 3.1 sup-
pose that we now have two experimental groups each of which is pre-
sented with a different degree of biological conditioning and we are
interested in asking if larvae respond identically to the different

degrees of conditioning.

The raw data matrices are

 

 

 

 

 

 

t1 t2 t3 t1 t2 t3
1 II. 1 1‘ 1'0 0 0'
1 1 1 0 0 0
. 1 1 1 . 0 0 0
1 0 1 1 1 1
51 = 1 1 1 A2 = . 0 0 0
1 1 0 ~ 0 0 0
. 1 1 1 . 1 1 1
0 1 1 1 1 1
1 1 1 1 1 0
10 Li 1 31 10_1_ 0 .13
31 = [0.9 0.9 0.9]. $2 = [0.5 0.4 0.3].
The pooled coVariance matrix is
f— a
.189 .106 .078
.106 .183 .094
L_._078 .094 ~1§Z.
The quantity (XI-12) = [0.4 0.5 0.6].
The test statistic is
2 1 8.1347 -3.8829 -l.6138 0.4
'r =-£Igl§l%%- [0.4 0.5 0.6] -3.8829 9.5404 —3.5564 0.5
* -l.6l38 -3.5564 8.7436 0.6

11.8652, and

_ 10 - _
F - T33§13180 _32)31 (11.5156) = 3.5156, d.f. = 3. 16.

 

99

The tabulated critical value of F0.05; 3, 16 = 3.24. Since

our calculated F is larger than the critical value we reject the null
hypothesis and conclude that larvae respond differently to different
degrees of conditioning.

The simultaneous confidence intervals are calculated as they
were in Example 3.1, except the two sample formula is used. For

example if we wanted to look at the first time period, then a -
[l 0 0] and the confidence interval becomes 0.7212. ~

Knowing that the response means are different we now want to
test for parallelism between the two curves. This is done by first
constricting the (k-l)(k) patterned matrix, C, as follows

1 —1 0

20
ll

0 1 -1

The test statistic is

'l'! _. ( )(1 ) 0 1 0 S O I -| O ' - l

1 0 '1 1 —1 0 0.4
-1 1 0 1 -1 0.5
0 —1 0.6

1.0, and

p = 10 + 10 - 3

10 + 10 - 2)(3 — 1) (1'0)

(
0.4722, d.f. = 2, 17.

The tabulated critical value for F0.OS; 2, 17 = 3.59. Since

the calculated F is less than the tabulated value the null hypothesis
$anPOt-13e rejected and we conclude that the curves are parallel. This
cipdlcates that although the response mean vectors are significantly
1fferent, these organisms are behaving the same way but at different
1:;:::- That is, there is no interaction between response and age of

%¥H§f§fiflflmptions Underlying
W5 Procedure

The assumptions underlying Hotelling's T2 Statistic are:

(1) multivariate nOrmal distribution, (2) equality of

100

variance-covariance matrices, and (3) serially uncorrelated data.

These assumptions must be satisfied.

Multivariate Normal Distribution

Just as in the univariate case, data for multivariate analysis
must be normally distributed. Procedures for testing this assumption
are handled by Sokal and Rohlf (1969) and appropriate transformations
are discussed, which are the best ways to correct for lack of
normality. If non-normality is present, then the probability value
for F is an underestimate which causes errors in the direction of
announcing too many significant results (Cochran 1947). In addition
to its effects on the validity of tests of significance, non-normality
is accompanied by a loss in efficiency in the estimation of treatment
effects and corresponding loss of power in the F- or t-tests. Sokal
and Rohlf (1969), however, show that the consequences of non-normality
are not too serious. Only very skewed distributions would have a
marked effect on the significance level of the F-test or on the
efficiency of the design.

Given certain prerequisites, data that is not normally dis-
tributed may be used with Hotelling's T2 Statistic. For example, the
data discussed in Examples 3.1 and 3.2 is binomial data, which is a
discrete distribution not a continuous one. It would seem, therefore,
that this data should be transformed to fit a normal distribution.
fhawever, the crucial part of Hotelling's test is the difference
Vector (3'80) in the one sample or (81-32) in the two sample.

Although the raw data is discrete, i.e., ones and zeros, the data

being tested comes from (X-Xo) or (XI-1(2) which may yield any number

101

between 1 and 0. This then becomes a continuous distribution, as N
approaches m.
The Central Limit Theorem states that as sample size increases,

the means of samples drawn from a population of any distribution will

 

approach the normal distribution. There are not very many statis-
ticians, however, willing to categorically state how big the sample

size must be before this theorem is true, but a general rule of thumb

is > 30. However, for illustrative purposes smaller samples are
described in Examples 3.1 and 3.2. For actual test data calculate,
a_priori, a sample size necessary to satisfy the assumption of normality
based upon the Central Limit Theorem. The procedure (personal communi-
cation with Dr. John Gill of Michigan State University) is

N = 3/W(1-fi)

where N sample size

a

any value the experimenter would like to be able to

differentiate. This might be done by picking the

smallest difference one wishes to be able to detect.
Let us suppose we have run a pilot experiment in the two choice

situations described earlier to determine Galleria's preference to

the highest degree of conditioning we intend to test, and find that

the largest response mean in the mean vector is 0.96. This is W. It

is now possible to calculate the N that would be necessary, given

that W = 0.96, to satisfy the assumption of normality. The result is

N 3/0.96(1-0.96)

78.125.

Therefore, an N greater than 78 would be required.

102
Equality of Variance—
Covariance Matrices

A test for heterogeneous variance is not necessary in the one
sample test. Hotelling (1931) originally devised the statistic with-
out the underlying assumption that the variances within a group are
homogeneous. This is reasonable since the test is designed to examine
multivariate factors among which homogeneous variances would not be
expected.

When comparing two experimental groups, however, the assumption
of equal variance-covariance matrices must be satisfied. Just as in
the univariant case where the assumption is made that the observations
are normally distributed about their population mean values with con-
stant variance, so an analogous assumption that the variates are
multinormally distributed about their mean values with constant
variance-covariance matrices is made in the multivariate analysis of
variance. Cochran (1947) has shown that if ordinary analysis of
variance methods are used when the true error variance differs from
one observation to another there will be a loss of efficiency in the
estimates of treatment effects as well as a loss of sensitivity in
tests of significance. The validity of the F-test for all treatments
is probably least affected, but t—tests from a pooled error may seri-
ously distort the significance levels.

Box (1950) has devised a method for testing the assumption of
homogeneous variances in multivariate analysis.

Example 3.3.--Let us return to Example 3.2 and calculate Box's
test for homogeneous variances.

M = N 16ge |§l — E (Vk loge |§k|).

103

where

N = Vk’ the total of the degrees of freedom,

Sk = the unbiased estimate of each covariance matrix being
~ compared,

Vk = degrees of freedom,

S = the pooled variance or covariance matrix, and

II = the symbol for determinants.

In Example 3.2,

loge|§1l = —6.948348, v1 = 9
logeISZI = -6.1862086, v2 = 9
ioge|§ | = —5.9251272, N = 18
M = 18(-6.l862086)—9(-6.948348)-9(-6.1862086)
M = 6.8592.
Calculate

2
_ 2t — 3t — 1 (z 1/v1 — 1/N),

l - 6(t+l)(k-l) k

where k = the number of matrices being compared.

A

 

Fl = 1/2 (k-l)t(t+l).
Therefore,
A _ 2(3)2 + 3(3)-l

 

1 ‘ 6(3+1)(l-1)

0.1806, and

'11
ll

1/2 (1-1)3(3+1)
= 6,

and (l-A1)M is distributed as X2 with f degrees of freedom, or

1
(1-0.1806)6.8592 = 5.6204

In this case A1 = 0.1806, f1 = 6, (1-A1)M = 5.6204 is referred

2 . . .
tc> tables of X with 6 degrees of freedom. The probability for the
Occurrence of a value as great or greater than this, when the
variances and covariances are in fact homogeneous from one group to

104

the next, is thus between 0.5 and 0.1, and there is therefore no
reason to doubt the homogeneity of the data.

The chi-squared approximation is good if k and/or t do not
exceed 4 or 5 or if some of the degrees of freedom are small. If,
however, k and/or t are greater than 5 or the degrees of freedom are
large, Box proposes an F—distribution approximation where

fit-muz) 21 v 2 2
A2 ”L6(k-1) k / k] ‘ UN

 

and M/b is referred to an F-table with f and f degrees of freedom

 

where 1 2
f + 2 f
f2 = A1 A 2 and b = 1 - A -If /f
2'1 1 12

It will be noticed that Box's approximation allows the simul—
taneous analysis of any number of matrices, but does not provide a
method for distinguishing which matrix is causing problems when the
test is heterogeneous. Therefore, it is necessary to perform pair-
wise comparisons in this event.

If the test proves to be heterogeneous there is no approximate
method available for Hotelling's two-sample test and the test should
be run as if the variances were homogeneous. If this is done a
decision must be made pertaining to the validity of the significance
test. If two covariance matrices are heterogeneous the pooled covari-
ance matrix will be smaller than one and larger than the other. There
will thus be a loss in sensitivity because the variances have been
inflated, making it more difficult to reject the null hypothesis.
Therefore, we can be confident of a significance test in which the
ruill hypothesis has been rejected, at the usual a-level of 0.05.

If a_priori an experimenter suspects finding heterogeneity of

vrrriance he should increase the sample size. Ito and Schull (1964)

Turve demonstrated that, for Hotelling's two-sample statistic, as N

105

approaches infinity the test statistic and significance tests are
unaffected by heterogeneous variances. The general rule of thumb is

that N must be greater than or equal to 30 for this to hold true.

Serially Uncorrelated Data

The greatest difficulty with longitudinal data is determining
if the error terms are correlated and what to do if they are. However,
it does not necessarily follow that longitudinal data is correlated.

If the assumption of independence is satisfied, the analytical pro-
cedure is valid whether or not the observations themselves are serially
correlated. Cochran (1947) has shown that if correlation exists the
following are the consequences: (1) if the correlation is positive,

the treatment means are less accurate than the mean of an independent
series, but are estimated to be more accurate, and (2) if correlation
is negative, these conditions are reversed.

Durbin and Watson (1950, 1951) and Watson (1955) have worked
out several regression models for testing the assumption of serially
uncorrelated data. There are several of these tests, depending on
the kind of data or experimental design, and the reader is directed
to the original papers. However, for the purposes of my design and
resultant discrete data the Durbin-Watson "One-and-Two—Way Classifi-
cation was used (see pp. 166-168 of Durbin—Watson 1951). The cal—
CUIations are much too laborious to be presented at this time, but

entail a regression model consisting of

u + oi + Bj + 2..

Y..
ij 1]

Where

. . . . h
the observation in the ith column and Jt row,

.<
ll

11'

106

+ B

.. co 5 ants and
1110113 n t ’

zij the error term.

Least squares estimates of u, ai and 8i are calculated and designated

as m, Ai-m and bj-m

where
m = the sample mean of all observations,
. . .th
Ai = the mean observation in the 1 column,
. . .th
bj = the mean observation in the 3 row.

Thus the residuals are given by

Z.. = Y.. - A. - b. + m, and
1] 1J 1 J
the test is made by calculating
£(AZ..)2
c1=——————1J
2 ,

XZ ..
13

where AZ.. = the first differences of the residuals when arranged as a
single time series.

The calculated d-value is compared to a lower (dL) and upper
(dU) range of critical values. If d < dL there is positive correlation°

,

if d > no correlation exists. If d < d < d the test is incon—
L-— ._

U
elusive. Durbin and Watson (1951) have calculated several tables of
upper and lower d-values. However, most of my research deals with
large matrices and I had to generate other tables in situations where
k' (the number of time units) was greater than 5. This was done on the
Computer using Durbin and Watson's (1950) formulas on p. 427.

Regardless of the outcome of the Durbin-Watson test for serial

(xxrrelation, Hotelling's T2 statistic is run normally. If correlation

exists, then by examining the variance-covariance matrices we can make

a (decision about the significance tests' validity. Positive serial

107

correlation means that the variances are over—estimates making it
difficult to reject the null hypothesis, whereas negative correlation
means the variances have been under-estimated making it easy to
reject the null hypothesis. Therefore, with positive correlation the
significance tests should be run at the a = 0.05 level and at the

a = 0.01 level when negative correlation is the problem.

However, tests for negative correlation are rarely run unless
the experimenter has an a_priori reason to suspect it. For example,
it is a common practice in econometric work (Durbin and Watson 1951)
to analyze the first differences of the observations rather than the
observations themselves, on the ground that the serial correlation of
the transformed errors is likely to be less than that of the original
errors. It is possible that the transformation has over-corrected,
thus introducing negative serial correlation into the transformed

errors .

Summary

Statistical procedures, and the assumptions underlying them,
are presented for the multivariate analysis of longitudinal data.
Hotelling's One and Two Sample T2 Statistics for Multivariate Analysis
are outlined, illustrating a procedure for treating time as a multi-
dimensional variate. Emphasized are procedures for testing the
assumptions of homogeneous variances and non-serially correlated
data. The major objective is to arm behavioral biologists with the
Imecessary analytical techniques enabling logical decisions about

tests of significance on repeated individual measurements over time.

CHAPTER IV

BIOLOGICAL CONDITIONING AND SPATIAL DISTRIBUTION
IN GALLERIA MELLONELLA (L.) LARVAE

 

Introduction

 

Stimuli affecting spatial distribution are such environmental
parameters as light, temperature and humidity (Allee 1926; Heatwole
1962; Surtees 1963a, 1964e 6 f; Friedlander 1964; Warburg 1964),
food (Park 1938; Brown 1946; Loschiavo 1952; Wynne-Edwards 1962;

Beck 1960), sex and genetic factors (Wellington 1957; Naylor 1959;
McDonald and Fitting 1965; Ogden 1969), and density (Long 1953;
Wellington 1957; Naylor 1959; Surtees 1963a, b 8 c, 1964a, b, c,

e 6 f). This paper, however, is concerned with the relationship
between biological conditioning and spatial distribution.

Biological conditioning refers to changes produced in a
medium by organisms living therein. The results of such homotypic
conditioning are physiological or behavioral. The physiological
effects of biological conditioning have been extensively studied
(Uvarov 1938; Allee 1931, 1934; Adolph 1931; Park 1932, 1934, 1935,
1936a 8 b; Brown 1946, 1951; Long 1953). It has been demonstrated

‘thaI.biological conditioning affords protection from toxic substances,
it; involved in control of sex, affects morphological changes and

growth, and affects population dynamics, such as natality and

108

109

mortality. Historically, biological conditioning has been considered
to be a population phenomenon. It is an expression of population den-
sity and is a normal and inevitable result of population growth. As
populations grow and maintain themselves, they of necessity modify
their environment and the environment in turn modifies them. However,
individuals also condition their environment and it has not yet been
demonstrated if the effects of population conditioning are different
from the effects of individual conditioning. Since biological con-
ditioning is a result of the organisms themselves and since the degree
of conditioning is a function of the number of organisms present or
of the length of time an organism is present, biological conditioning
and animal-animal interactions have been considered as an inescapable
unit. However, they can be separated by experimental procedures.
The effects of either, or the interaction between the two, may alter
the behavior of individuals with resulting effects on populations.
Biological conditioning has been demonstrated to affect spatial
distribution of populations of organisms, particularly invertebrates
(see Chapter I). It has been shown in Tribolium (Park and Woolcott
1937; Park 1948) that homotypic conditioning alters the movements of
these beetles and that adult populations of Tribolium seem to prefer
weakly-conditioned medium. Naylor (1959 and 1965) extended these
findings to include animal-animal encounters, such as interactions
Iaetween the sexes, and homotypic conditioning, and he found that the

distribution of Tribolium confusum and I, castaneum was only explain-

 

able when all three factors were accounted for. The interaction

betrween these three variables seems a common phenomenon among

110

invertebrates, particularly insects. For example, Long (1953, 1955)
with butterfly larvae and Wellington (1957) with larvae of the tent
caterpillar both demonstrated similar interactions.

Although biological conditioning affects spatial distribution
and distribution may also be altered by other variables, the inter-
actions between animal-animal interactions and homotypic conditioning
remain confounded. Studies dealing with biological conditioning deal
with whole populations, and little, if any, data have been collected
on the effects on individual behavior.

It is the purpose of this paper to look at the role of homo-
typic conditioning in the dispersionary behavior of isolated Galleria
mellonella (L.) larvae, the assumption being that before we can under-
stand population behavior we must first understand individual behavior.
The general hypothesis being tested is that response of isolate larvae
to biological conditioning will change as a function of the degree of
conditioning and age of the test larva. In testing this hypothesis a

bioassay of behavioral preferences to conditioning was developed.

Materials and Methods

 

Husbandry

Rearing Galleria mellonella is a relatively simple task and

 

was described in Chapter II. All experimental larvae were drawn from
tflie colony dishes which were maintained in darkness in incubators at
31:}°C and 75% R.H. Due to the manner in which egg hatches were
‘ulllected (see Chapter II), the age of larvae within any colony dish

had a maximum possible variability of 24 hours. It was therefore

111

necessary to have a criteria for consistently determining age from
hatch to hatch. Weight of larvae was used as the criteria. Table 4.1
shows the mean weights of 500 larvae, at each age, randomly selected
from appropriate colony dishes as well as the corresponding weights of
larvae actually used for experimental purposes. Figure 4.1 shows the

size distribution of larvae corresponding to Table 4.1.

Apparatus

The experimental apparatus consisted of 100 x 25 mm plastic
petri dishes with two 4:0.05 gram food lumps in each dish. Food was
prepared as described in Chapter 11. Within any one dish the food
lumps were placed on opposite sides of the dish so that test larvae

could be introduced between them and allowed to make a choice.

Experimental Procedures

 

The experimental design is shown in Figure 4.2. Four con-
ditioning periods (1, 3, 5, and 7 days) and three ages of conditioning
larvae (6, 9, and 12 days of age) were used. Preference tests were
run with 7, 11, and 13 day old test larvae for each conditioning age
and/or period. Each box in the design has an identification code
(1E, 2B, 3B, and 37C) where E stands for experimental group and C for
control group. There is one control for every three experimental
groups based on days of conditioning. For example, 37C is the control
.for groups 1E, 2E, and SE at one day of conditioning. Each group has
a sample size of 80.

The date in each box of Figure 4.2 indicates the day on

which the test was begun for each group. Due to the size of the

112

Table 4.1
Larval weights corresponding to the larvae in Figure 4.1. Weights were
calculated for 500 randomly selected larvae of a particular age as well
as for the larvae of a particular age actually used for experimentation.

 

 

A e Randomly Selected Experimental Larvae
(D: 5) [Mean weights(mg) [Mean weights(mg)
y .1 standard deviation] :_standard deviation]
6 0.53 :_0.012 0.71 :_0.009
7 1.27 :_0.029 1.49 i 0.010
9 5.29 :_0.090 6.33 :_0.069
11 12.45 :_0.221 12.72 I 0.014
12 22.14 :_0.710 22.56 :_0.320
13 34.40 1 0.760 33.82 :_0.221
15 57.60 i 0.700 55.40 :_0.250

 

 

Figure 4.1
Frc>m right to left, these are 6, 7, 9, ll, 12, 13, G 15 day old larvae
“Seed for conditioning and testing preference for conditioning. Note
the: size distribution of the larvae.

113

.::h mez much» now Houucou new macaw Amy Hmucosfiuomxo
some con: came onu ecu agenda :owumowmfiucoew on» «cwzocm .cuwmov Huucoafluomxm

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

~.e engage
ee\N\e ee\m~\m ee\e\e ee\e\e
one one uee ume
- 2 v
ee\~\e , ee\~\o ee\m\e ee\m~\m ee\m~\m ee\m~\m ee\e\e ee\e\e ee\e\e ee\m\e ee\e\e ee\e\e Ana
men mmm men men mmn mHm men mew mew mew mew mmm qua
- . o
,
ee\H~\m ee\-\e ee\n\m ee\e~\o
I.
uee one owe uHe
- - - = H
eh\HN\m ee\H~\m en\H~\m ee\mm\e ee\-\e ee\N~\e ee\n\m ee\n\m ee\n\m eh\e~\e ee\e~\e ee\e~\e In.
mem mnm mNN mam mom me” me" meH meH mmH meH meg an.
V
eh\efi\m ee\e\e ee\na\e ee\e~\e “nu
8e eon 68 oh A
k V
ee\ea\m ee\wa\m ee\ea\m ee\w\e ee\e\e ee\e\e ee\n~\e ee\mfi\e ee\nH\e ee\efi\e ee\efi\e ee\e~\e an.
mNH mHH meH me we we we mm me mm mu me
a. e e a. o e S e e a. e e A. emu
k m m — defer

 

 

 

 

 

02_ZO_.—_OZOU *0 mummvmo

114

experimental design and the total number of larvae required (total N =
3840) it was necessary to test groups at different times and with
different populations (hatches) of larvae. Therefore, the test date
for every block of three experimental groups and their control was
randomly determined, a_priori. Every four groups form a natural
entity of days of conditioning, age of conditioning larvae, and age of
test larvae and were therefore run at the same time. For example,

1E, 2E, 3E, and 37C or 19E, 20E, 21E, and 43C were each run as a unit.

Experimental group 8E has been chosen to elucidate the experi-
mental procedures. Food was prepared as discussed in Chapter II,
Table 2.1, and 160, 4:0.05 gram food lumps were weighed on a Mettler
P120 Scale. These food lumps were flattened on wax paper until they
were approximately 6 mm thick and 157 mm in circumference. One food
lump was then placed in each of 160, 150 x 25 mm plastic petri dishes,
the lump touching the periphery of the dish. In 80 of these dishes
a conditioning larva was added, in this case a 9-day old larva. The
other dishes have no conditioning larva. All dishes were then put in
a Jamesway Single Stage Incubator-Hatcher, Model 2528, for the length
of time conditioning was to occur; in this case 5 days.

At the end of the 5 day period the following manipulations
were performed. The conditioning larva, which is now 14 days old, is
:removed from each of the 80 dishes by gently tapping the silk tunnels
ill has spun with a pair of forceps. This causes the larva to poke
iits head out and it can be carefully removed with forceps. Once all
1arVéle are removed, the 80 non-conditioned food lumps are placed one

in eaiczh of the conditioned dishes opposite the conditioned lump. This

115

is done wearing surgical gloves since pilot experiments indicated
larvae will selectively respond to food handled with bare hands. The
distance between food lumps is approximately 35 mm across the center
of the dish. A small black dot is placed on the side of the petri
dish opposite one of the food lumps for identification purposes. The
food lumps receiving the identification dot were randomly selected so
as to alleviate the possibility of response to the dots. An identifi-
cation number was also put on the top of each dish.

Once the two food lumps are in the same dish, a single, in
this case 7 day old, larva of proper size is removed from the appro-
priate colony dish and placed in the petri dish between the lumps and
the top is placed on each dish. The total procedure of removing the
conditioning larvae, adding the plain food lumps, and seeding all 80
dishes with a test larva takes 5 minutes. At the end of the 5 minute
period the location of each larva is recorded and the dishes placed
in a Napco (Model 330) incubator at 32:1°C and 75% R.H. in darkness.
The incubator holds 8 shelves of 80 dishes per shelf with the dishes
stacked 5 deep.

Every 12 hours the dishes were removed from the incubator and
data collected as to which food lump the larvae were residing in. The
dependent variable, therefore, is position of the larvae over time.
Except during the wandering phase, larvae were never found between

.fbcni lumps. Every 24 hours the position of the dishes on each shelf
as Well as shelf position in the incubator was randomized to prevent
anyllarval responses to discontinuities of temperature, humidity, or

light: within the incubator.

116

In order to record position of the larvae it was necessary to
bring the dishes into the light for 5 minutes, every 12 hours. There
is some indication that Galleria larvae are photonegative and I may
have been recording the preferred food lump in which to escape from
light. However, differences in response to degree of conditioning
are still evident. Data were collected every 12 hours until pupation
occurred, at which time the experiment was terminated.

The control group for SE, namely 39C, was set up exactly as
the experimental group except that none of the food lumps were con-
ditioned. The controls, therefore, control for food differences and
the handling of one of the food lumps at the end of the conditioning
period. The response of a larva in a control group will be random if
these variables have no effect on distribution or preferences.

Every group in Figure 4.2 was set up exactly as just described,
the only differences being the number of days of conditioning, age of
conditioning larvae, and age of test larvae.

For purposes of ease in referring to this experimental design
and manipulations, I will refer either to the group number, 1E or 10E,
or I will use a CAT-value, where C = the number of conditioning days,
A.= age of conditioning larva, and T = age of test larva. Therefore,
CAT = l, 6, 7 refers to an experiment in which 1 day of conditioning
was done by a 6 day old larva and the response of a 7 day old larva
was; tested. For group 85, CAT = 5, 9, 7; for group 27E, CAT = 1, 12,
13 and so on. If multiple comparisons are desired, the CAT-value

can [)3 used to indicate these as well. For example, if we wanted to

100k .at l, 3, 5, and 7 days of conditioning done by a 6 day old larva

117

at the start of conditioning and tested with a 7 day old larva then
CAT = (1,3, 5, 7), 6, 7.

In all Tables and Figures in this paper test larvae are
referred to as 7, 11, or 13 days old. It must be remembered, however,
that these are the ages when a particular experiment is started, and
that, for example, four days into an experiment a 7 day old larva is
11 days old, an 11 day old larva is 15 days old, and a 13 day old
larva is 17 days old. The abscissars in all graphs in this paper

are calibrated in days, following initiation of the test not age of

test larva.

Analysis
The dependent variable chosen for analysis within each exper-
imental group is the mean response to conditioned food over time. Each
experimental group has 80 dishes with 1 larva per dish. If a larva is
in the conditioned lump it received a score of 1, if in the non—
conditioned (plain) lump it received a 0. In the case of 8B, for
example, this would generate an 80 x 20 matrix whose entries are
either one's or zero's, 80 being the number of dishes or larvae and
20 being the number of 12 hour intervals. A mean for each 12 hour
interval can be calculated which generates a l x 20 vector whose
entries are the mean number of larvae in conditioned food at each 12
hour time interval. For actual analysis, however, 24 hour intervals
‘werre utilized, since there was no difference in response from AM
(morning) to PM (afternoon) readings (Table 4.2). The wandering
phase was also not included in the analysis because wandering larvae

exhitxited atypical behavior. This was done because some data

118

. So. w H: meoceewe> msoecomouepoxz

.wouoo> moanewe>oo pmewwda on» see: much» 0:» Beam ease umou m.vefimamm oco one you nouoo> wouooexo one

 

 

 

 

 

Q
.HeeweeHeeeeeH u He .me. u e we ew.wu :e eee Hn.H u He Hewe. u e we eH.w «HeeH eHewe.
ewe. a e ee ew.w a hen eHeHe.
enH. ene. eeH. eee. mee. ewH. eHH. eHH. eeH. wew. eeHeeme
eeH. ene. HHH. Hee. Hee. HHH. eeH. HeH. weH. wew. ”me
: eeH. ene. HeH. Hee. Hee. ewH. Hee. eeH. HHH. wew. we
.me.H nen. ewe. ene. eee. wHe. ene. nee. nee. eee. ewn. 2am meH
tee.H eH H en H wHe. ewe. nee. wHe. eee. eme. eee. nee. ene. ewn. 24x,
eeH. nnH. nnH. meH. eeH. wnH. wnH. enH. eeH. eew. eeHeewe
.ne.H an. enn. enn. wen. nen. wHe. een. nen. een. ewe. 2mm en
tee.H ee e em». mnn. mnn. wen. hen. emn..wemw we». em»..hme. 2mm.
0 O
Hew. eee. Hee. eee. HHH» eee. emH. enH. wew. eww. e H ewe
ree.H een. ene. eee. wHe. ene. nee. wHe. nen. ewn. wee. 2e“ we
«ee.H ew e een. wHe. ewe. eee. ene. eee. wHe. wen. ewn. wee. z<m
Hew. new. wew. new. wew. eww. eew. eew. eew. eew. eeHeeme
tee.H nee. nee. ewe. ewe. eee. wee. eee. wHe. nee. nee. 2e“ ew
.ee.H we e ewe. wHe. wee. ewe. nee. wee. ewe. ewe. ene. nee. z<
peep e we Hm eH e e n e e e e w H mzoee
m.xom exam cw 05mm.

 

.wemxaece onawm nauoze use n.vo:o m.wceHHeuo:-weem: vehemeoo
ewe: mnouoe> sees 2m emu 2< nwogu ecu eeuooHem xfiaoecmw one: N.e enamem cw museum kucoawnomxo o>~039

NJ 0.3.3.

119

. 30. W 3 ouceeaenH esoecowonepozz

 

 

 

 

 

.He>HeeHeeoeeHuHe ee. u e we eH.w u ee eee ee.H u He Hewe. u e ee He.w u HeeH eewe
*
eee. ene. Hee. ene. eee. eeH. eeHeeme
Hee.H eee. ewe. wHe. ewe. wHe. wen. sew mew
Hne.w wH e wHe. NHe. wHe. ewe. eee. een. 24%
eee. Hee. HeH. eeH. eeH. eew. eeHeeee
w
Hee.H eee. nee. nee. eee. wHe. nee. 2em mew
HHe.H eH e wHe. wHe. nee. ewe. eee. nee. 2<m
new. eHw. nHw. eHw. eHw. wew. eeHeoee
Hee.H nee. nee. ene. nee. an. ewn. 2mm eeH
HHe.w ee e nen. een. een. een. een. ewn. 24m
wew. wew. eew. eew. eew. eew. eeHeeme
Hne.H wHe. wHe. wHe. nee. wHe. nee. sew meH
Hee.w ee e eee. eHe. wHe. ene. ewe. nee. z<m
eeee e.xee e we e e e e w H eeeee

memo Ge maee

 

H6353 we e32.

120

.mmo.ww: eeocewwec» msoocemowmuenz

.HOpoo> ooeeewm>oo ueomwefl
on» see: macaw one Scum came peep n.vemeaew eno exp new Hepuo> empooQXe one

 

 

 

 

 

 

 

 

.ee. u e um eH.w u ee eee ee.H u He ”ewe. u e ee ee.w «HeeH eewm.
.o>wm: ocooew mm3 o: «>1 .
. H . H eH .ewe. u e ee ee.w u Hen eeHe.
eeH. eeH. eeH. eew. eeHeeee
w
ee.w nen. een. wen. eee. sew mee
ew.w eH e nen. nen. enn. eee. 2<x
eee. eee. eee. ewH. eeHeeee
ene. eee. eee. ewH. 2mm
: eee. eee. ene. ewH. 2mm mee
ew.e ewe. nee. eee. eee. 2em
Hee.w ne H ee H eee. nee.[mme. eee. z<u
eHH. HHH. eeH. eeH. eoHeeee
ne.w nee. ene. ewe. een. ZMM mew
ew.w wH H .hee. ene. eee. ewn. z<x
eww. eww. eww. eww. eeHeeme
Hw.w eee. eee. wee. wee. sew mew
Hee.w me e wee. wee. wee. wee. z<m
umoe 0 mm Hm v m N H macaw
m.xom

memo cw oaeb

 

H.e.eeoee w.e oHeee

121

reduction was necessary, and will be discussed shortly. Therefore,
the data matrices are 80 x 10 for 7 day old test larvae, 80 x 6 for 11
day old test larvae, and 80 x 4 for 13 day old test larvae. The
respective mean response vectors are 1 x 10, l x 6, and 1 x 4.

Statistical analysis is by means of Hotelling's One- and Two-
Sample T2 Statistics for Multivariate Analysis (see Chapter III),
although some Chi-Square and ANOVA analyses are employed. An a_priori
decision about the a-level for tests of significance had to be made.
This was necessary because multiple comparisons were to be made,
positive serial correlation was suspected and the variances between
groups may be heterogeneous. Serial correlation and heterogeneous
variances make it more difficult to reject the null hypothesis, whereas
multiple comparisons increase the probability of rejecting a null
hypothesis on chance alone. However, the maximum number of comparisons
made with any one group is 10, which means we would expect less than
a 0.5 in 10 chance of a significant difference on the basis of chance
alone at the 0.05 level. It was therefore, determined to use an
a-level of 0.05. However, the null hypothesis is two-tailed and the
a-level is actually 0.025. A two-tailed hypothesis was utilized
because there was no a_priori basis upon which to assume response to
conditioning to be attraction or repulsion.

After making comparisons with Box's test for heterogeneous
variances, I found that, almost without exception, the variances in
each comparison to be heterogeneous. In Hotelling's Two-Sample
Statistic, a covariance matrix is calculated by pooling the two co-

variance matrices of the groups being compared. If the respective

122

covariance matrices are homogeneous, the pooled covariance matrix is
the best estimate of the two. If the two covariance matrices are
heterogeneous, as they are in my data, then the pooled covariance
matrix is inflated and is not the best estimate. Ito and Schull
(1964), however, indicate that heterogeneity of variances does not
affect the power of the test or the level of significance if N is
very large, however, large N is not well defined. If the null hypoth-
esis is rejected when the variances are heterogeneous we may be sure
of this decision, but if it is not rejected then we may be accepting
false null hypotheses.
Because of these considerations, I also ran Hotelling's One-
Sample statistic on those comparisons with heterogeneous variances and
in which the Two-Sample statistic had not rejected the null hypothesis.
In this procedure the mean vector of the group with the greater vari-
ance is used as an expected vector (BV.) and is compared to the other
test mean vector using the covariance (CV.) of the group with the
smaller variance. This procedure has the effect of letting the smaller
covariance matrix be the "best estimate" of the two groups being com-
pared and asking whether those groups are still not different. If
the groups are still not different then we know that it is not because
of heterogeneous variances. If the two groups are found to be dif-
ferent with the One-Sample procedure, then we know that non-rejection
<3f the null hypothesis with the Two-Sample test was a result of
heterogeneous variances. This piece of information coupled with
exfilllljmation of the variance-covariance vectors in Table 4.3 gives us

additional information about the behavior of the experimental animals

123

Table 4.3
Results of Hotelling's One Sample comparison of each group from Figure 4.2
with an expected vector whose entries are all 0.5 to determine which
groups are random. Mean vectors(X), estimated variance vectors(52), F-
values, and D-values for serial correlation are given.

 

Time in Days
Grog? 1 2 3 4 5 6 7 8 9 19. F D
IE .500 .550 .437 .450 .462 .475 .462 .475 .487 .475 0.86 1.55*
.253 .251 .249 .251 .252 .253 .252 .253 .253 .253

.478 .587 .612 .650 .662 .650 .625 .625 .637 .587 1.31 1.51*
.253 .245 .240 .230 .226 .230 .237 .237 .234 .245

.512 .625 .650 .675 .700 .687 .737 .712 .737 .675 3.12** 1.54*
.253 .237 .230 .222 .213 .218 .196 .207 .196 .154

.512 .475 .487 .562 .537 .550 .500 .512 .500 .475 0.94 1.34*
.253 .253 .253 .249 .252 .251 .253 .253 .253 .253
.550 .662 .650 .637 .625 .625 .625 .612 .550 .462 1.94 1.42*
.251 .226 .230 .234 .237 .237 .237 .240 .251 .252

.562 .700 .762 .750 .800 .837 .800 .812 .712 .612‘ 8.59** 1.39*
.249 .213 .183 .190 .162 .138 .162 .154 .207 .240

6B 2 .662 .725 .787 .812 .887 .875 .912 .900 .875 .700‘ 25.97** 1.48*
52 .226 .202 .169 .154 .101 ,111 .081 .091 .111 .215

.525 .487 .487 .462 .500 .475 .500 .575 .475 .387 1.84 1.53*
.253 .253 .253 .252 .253 .253 .253 .247 .253 .240

.625 .750 .787 .750 .812 .787‘.762 .775 .775'1712 6.82** 1.57%
.257 .190 .169 .190 .154 .169 .185 .177 .177 .112

32
X'
32
22 .625 .737 .837 .837 .837 .862 .856 .862 .837 .712 11.40** 1.46*
S
i
32

 

2B

3B

4B

5
i
52
X
52
37C :2
X»
52
2
SE 82

38C 2

 

 

7E

85 .257 .196 .158 .158 .158 .120 .129 .120 .158 .207
.637 .862 .857 .850 .925 .950 .912 .962 .925 .857 74.72** 1.161
.254 .120 .138 .129 .070 .048 .081 .057 .070 .138

x .462 .512 .462 .557 .462 .487 .475 .487 .457 .550 1.26 1.38*
5 .252 .255 .252 .252 .252 .255 .255 .255 .249 .250

'XT‘ .725 .800 .887:.857‘.875 .912 .900 .875 .925 .757 23.48** 15651
52 .202 .162 .101 .138 .111 .081 .091 .111 .070 .196

X .625 .762 .875 .875 .950 .925 .937 .862 .862 .812 48.62** 1.48*
$2 .237 .183 .111 .111 .048 .070 .059 .120 .120 .154

125 x .700 .687 .712 .725 .800 .850 .875 .862 .812 .725 12.76** 1.55*
s2 .215 .218 .207 .202 .162 .129 .111 .120 .154 .202

40C X .562 .525 .537 .500 .525 .500 .500 .525 .437 .337 1.58 1.41”
82 .249 .253 .252 .253 .253 .253 .253 .253 .249 .226

= 2.24 at P = .025; D = 1.71 and D = 2.28 at P = .05.

F
(10,70) L u
*‘k

F g =
(10,70) 2.59 at P .01.

D-VEIIue was inconclusive.

124

Table 4.3 (cont'd.)

 

Time in Days

 

 

 

 

 

Group 1 2 3 4 5 6
13E 1 .457 .512 .487 .512 .512 .512 0.49 .971
s2 .249 .255 .255 .255 .255 .255
145 X .657 .687 .650 .725 .687 .700 4.55** .981
52 .254 .218 .250 .202 .218 .215
15E 1 .575 .687 .687 .725 .775 .762 7.07** .721
52 .247 .218 .218 .202 .177 .185
410 X .457 .500 .525 .475 .525 .557 0.89 .971
52 .255 .255 .255 .255 .255 .252
165 i .725 .712 .687 .675 .687 .687 5.28** .891
52 .202 .207 .218 .222 .218 .218
175 ' x .757 .850 .800 .850 .812 .800 17.70** .851
52 .196 .129 .162 .129 .154 .162
18E x .712 .875 .912 .950 .900 .950 66.80** .76*
52 .207 .111 .081 .048 .091 .048
42C 2 .475 .475 .425 .412 .450 .457 0.85 .891
2 .255 .255 .247 .245 .251 .249
195 ' .657 .812 .825 .862 .850 .812 l4.82** .951
2 .254 .154 .146 .120 .129 .154
20E 1 .657 .812 .850 .887 .887 .900 27.52** .901
2 .254 .154 .129 .101 .101 .091
215 ’ .675 .957 .925 .912 .875 .850 78.01** .061
52 .222 .059 .070 .081 .111 .129
45c 1 .487 .487 .462 .457 .475 .462 0.42 .791
52 .255 .255 .252 .249 .255 .252
225 X .637 .850 .887 .862 .875 .862 19.90** .051
52 .254 .129 .101 .120 .111 .120
23E 2 .762 .912 .925 .912 .925 .900 51.52** .901
s2 .185 .081 .070 .081 .070 .091
24B 22 .857 .950 .900 .912 .912 .862 82.21** .78*
s .158 .048 .091 .081 .081 .120
44c ‘ .462 .487 .487 .512 .500 .512 0.24 .16
52 .252 .255 .255 .255 .255 .255
*
5(6 74) = 2.59 at P = .025; DL=1.83 and DU=2.16 at P = .05.
9
**F(6 74)= 5.08 at p = .01.
3

I . .
D-value was 1nconc1u51ve.

125

Table 4.3 (cont'd.)

 

Time in Days

 

 

 

 

 

Group 1 2 3 4 F D

255 X .550 .550 .550 .562 0.46 2.24
52 .251 .251 .251 .249

26E X .662 .662 .650 .650 5.04* 2.21

2 .226 .226 .250 .250

275 i .662 .757 .775 .775 10.49** 2.51
52 .226 .196 .177 .177

45c ‘ .462 .525 .487 .525 1.10 2.52
52 .252 .255 .255 .255

28E 1 .562 .662 .675 .675 5.74** 2.081

2 .249 .226 .222 .222

295 i .750 .825 .875 .857 25.55** 2.57
52 .190 .146 .111 .138

505 i .787 .857 .875 .862 24.76** 2.14
s2 .169 .138 .111 .120

46C 1 .500 .462 .475 .475 0.57 2.14
82 .255 .252 .255 .255

313 X .812 .850 .800 .812 27.05** 1.961
52 .154 .129 .162 .154

325 x .775 .887 .912 .875 45.59** 2.12
s2 .177 .101 .081 .111

335 12 .850 .950 .957 .925 81.84** 2.20
s .129 .048 .059 .070

47c 22 .450 .500 .487 .462 0.66 2.51
s .251 .255 .255 .252

345 i .925 .975 .987 .975 424.55** 2.55
52 .070 .025 .012 .025

555 x .875 .975 .987 .962 674.79** 2.47
52 .111 .025 .012 .057

365 22 .650 .762 .750 .737 7.54** 2.24
s .250 .183 .190 .196

48c 22 .450 .457 .462 .475 1.25 1.961
s .251 .249 .252 .255

*

F(4,76) = 2.96 at P = .025; DL=1.89 and 0U=2.10 at p = .05.
*1-

3.59 at P .01.

Foam"

I . .
D-value was inconclu51ve.

126

that we would not otherwise have. This will be discussed later, but
an example will help now. If we look at the estimated variances in
Table 4.3, we find that the variances decrease as degree of condition-
ing increases, that is, groups 2B, 5B, 8B, and 115. In Table 4.4 the
Two-Sample test comparing 2E to SE gave an F2-value of 1.84 and the
null hypothesis was not rejected. However, Box's test says that the
variances are heterogeneous indicating that the pooled covariance
matrix for this test was not the best estimate, and that we may have
accepted a false null hypothesis. The One-Sample comparison of
groups 2E and 5E yielded an Fl-value of 3.44 which is significant at
the .01 level. Therefore, the heterogeneous variance was responsible
for originally not rejecting the hypothesis. We now know that the
variances between the two groups are different and that it is greater
in the group (28) responding to the lowest degree of conditioning.
These facts tell us that larvae in the two groups are not responding

in the same way to different degrees of conditioning even if their mean

 

response is consistent from one experiment to another. It appears
that response to low conditioning is much more variable.

The One-Sample procedure, therefore, is being used only as a
tool to help dissect the data. The Two-Sample procedure is still the
legitimate one and will be used for the final conclusions on the
experiments. It should be pointed out, however, that the One-Sample
procedure does not alter any of the conclusions, it only reinforces
them and adds to our understanding of what is happening biologically.

Since both One- and Two-Sample procedures were used, some

experimental groups were multiply compared as many as 20 times which

127

necessitated a reconsideration of the appropriate a-level for hypoth-
esis testing. All tables in this paper show the critical values for
a-levels of .025, .01, and .001. Any rejection of the null hypothesis
with the Two-Sample test, even at the .025 level, is valid since it

is difficult to reject when the variances are heterogeneous. Since

the One—Sample test has eliminated the source of the greatest vari-
ability, any rejection at the .025 level must be cautiously interpreted
because homogeneous variances make it easier to reject the null

hypothesis.

Pilot Experiment

 

One of the reasons Galleria larvae were selected for these
studies is that they produce discrete silken tunnels in the food as
well as deposit copious amounts of frass. This is an obvious form of
physical homotypic conditioning, although there may be chemical con-
ditioning as well. However, it was first necessary to determine if
such conditioning would elicit a response from conspecifics before
the design in Figure 4.2 was run.

To determine whether Galleria larvae respond to homotypic
conditioning an 80 dish experiment was set up with CAT = 7, 6, 7.
Data was collected every 24 hours, and the mean response vector is
shown in Figure 4.3 as compared to an expected vector whose entries
are all 0.5. The response vector is significantly different from
random with F = 20.80, where F(14’ 66) = 3.18 at P = 0.001. The
curVe has also been divided into "Larval" Phase and "Wandering" Phase.

As a result of this pilot study it is obvious that larvae

respend to conditioning, in this case they are attracted to it. When

o
6

Larvae in Cond. Food

128

"Larval" Phase

80‘
70‘
mi
sol
40‘

501

 

¢——- CAT=7,6,7.

10‘ .---- Expected vector.

 

 

0 f U I I I V 1 I 1 I T I r fi

1 2 3 4 S 6 7 8 9 10 11 12 13 14

Time in Days

Figure 4.3
Pilot experiment to see if Galleria larvae reSpond to bio-
logically conditioned food. Note that after 16 days of age
the response begins declining during which time the larvae
are apparently not exhibiting a preference for either con-
ditioned or non-conditioned food.

129

the larvae were 16 days old (day 10 of the experiment), however, what
appears to be a change in behavior occurs. The curve begins declining
in what looks like repulsion to conditioning, and it would appear that
the larvae are moving to the non-conditioned food lump. Observations
from the experimental situation showed that larvae were all over the
dishes, mostly not in any food lump.‘ This is what the literature
terms the wandering phase during which time larvae are "looking" for

a place to pupate and seemingly exhibiting no preference for con-
ditioned versus non-conditioned food. This is not an artifact of the
way in which the data were calculated for the wandering phase. As
previously described, a "one" was recorded if a larva was in the
conditioned food lump and a "zero" if in the non-conditioned food
lump. If a larva is on neither food lump, as is often the case during
the wandering phase, it is effectively removed from consideration in
the analysis. The fact that the response curve in Figure 4.3 drops
below 50% does not mean that the larvae are now all in the non—
conditioned food. What it does mean, however, is that 20% of the
larva pupated in the conditioned food and the other 80% pupated in
either the non-conditioned food or elsewhere in the experimental
dishes. This same phenomenon occurs in every group in Table 4.2,
regardless of degree of conditioning or age of test larvae. It would
appear, then, that we are dealing with a behaviorally different
organism once it reaches 16 days of age and there is no 3_priori basis
upon which to calculate an expected vector for the wandering phase.
For these reasons analyses in this paper deal only with the test larva

Lu) through 16 days of age. The use of 16 days of larval age as the

130

cut-off point was determined from Figure 4.3 and from the literature
which indicates that wandering begins 2 days before pupation. The
mean pupation time for the larvae in this pilot experiment was 18

days and I subtracted 2 days to arrive at this 16 day figure.

Data Reduction

 

Aside from the fact that some data reduction had resulted from
elimination of the wandering phase from analysis, it was desirable to
further reduce the amount of data being analyzed for convenience
sake as well as for the fact that the central memory of the computer
was being exceeded with such large matrices. I decided to analyze
response mean vectors consisting of only PM data points, that is, 24
hour time intervals rather than 12 hour time intervals.

In order to make this decision it is first necessary to demon-
strate that daily AM responses are not different from daily PM
responses within the same group. Twelve groups from Figure 4.2 were
randomly selected and their AM and PM vectors analyzed with Hotelling's
One- and Two-Sample T2 Tests. The results of these comparisons are
given in Table 4.2 and it will be noted that in every case the null
hypothesis could not be rejected. Therefore, AM data points were

eliminated from future analysis.

Results
The following description of the results has been divided
into six basic parts, the first three of which directly relate to the
hYPOtheses that response to conditioning is a function of the degree

0f cIO‘nditioning and the age of the test larva. First, within-group

131

comparisons will be examined as to whether larval response to con-
ditioning is random, attraction, or repulsion. Second, between-group
comparisons of larval responses to conditioning will be analyzed as
the amount of conditioning increases. Third, between-group comparisons
of the effects of larval age on response to conditioning will be
examined. Fourth, a brief analysis will be presented relating to
initial preference to conditioning and amount of larval movement in
conditioning. This analysis allows us to ask whether a larva's
initial preference for conditioning is a good predictor of its long-
term response and whether different degrees of conditioning alter the
number of times a larva moves between the conditioned and non-
conditioned food lumps. Fifth, data on whether response to con-
ditioning can be over-ridden by varying the food source will be
presented to see if larval response to conditioning is affected by
the quality of food on which the conditioning is done. Finally,

data on whether adult Galleria exhibit preferences to larval con-
ditioning will be presented. This last analysis is for the purpose
of discovering if larvae and adults differentially respond to larval
conditioning, and if they do, to propose some possible mechanisms for

the differences.

I. Within-Group_Comparisons for Random-
pgss Of Larval Response to Conditioning

 

 

Hotelling's One-Sample T2 Statistic was used to compare the
response mean vector of each group in Figure 4.2 with an expected
‘Vector whose entries are all 0.5. This procedure enables determining

IMhich groups are randomly distributed and which are non-random, i.e.,

132

significantly attracted or repelled by conditioning. The results

are shown in Table 4.3. The mean vector (i) and estimated variance
vector (S2) are given for each group and should be referred to in all
future comparisons in this paper. If for any two groups the reader
wishes a pooled variance vector, he need only average the two vectors
of concern or if the standard error is desired, it can be calculated
for any entry in the variance vector as Sz/N-l (1.96). Table 4.3
also lists the F-values for each test as well as the D-values for
serial correlation of the error terms.

The results in Table 4.3 are good indicators for the compari-
sons needed as tests of the other hypotheses in this paper. All con-
trol groups (37C—48C) are randomly distributed as predicted. The
only experimental groups exhibiting randomness are 1E: CAT = l, 6, 7;
2E: CAT = l, 9, 7; 45: CAT = 3, 6, 7; 13E: CAT = l, 6, 11; and 25B:
CAT = l, 6, 13. In groups 1E, 13E, and 255 conditioning was by a 6
day old larva for 1 day and examination of the size of a 6 day old
larva in Figure 4.1 indicates a random response might be expected.
The other random responses were by groups 2E and 4E, both of which
involve 7 day old test larvae. The corresponding groups for 11 day
old test larvae (14E and 16E) and for 13 day old test larvae (26E
and 28E) show non-random responses, indicating there may be age dif-
ferences in response to the same degree of conditioning. Perhaps 7
day old larvae require the presence of more conditioning before they
perceive it as different from non-conditioned. If 7 day old larvae

are not responding to conditioning at the same level as 11 or 13 day

133

old larvae, the F-values should reflect such a trend. I will come
back to this point at a later time.

Examination of the estimated variance vectors ($2) in Table 4.3
also shows a consistent phenomenon. If the variances are compared
within a test age at constant days of conditioning, i.e., 15, 2E,
and 3B, the variances get smaller as age of conditioning larva gets
bigger. If the variances are compared within a test age at a constant
age of conditioning larva, but across days of conditioning, i.e., 1E,
4E, 7E, and 105, the variances get smaller as days of conditioning
increase. If variances are compared for constant days of conditioning
done by the same age of conditioning larva but the age of the test
larva is varied, i.e., 2E, 14E, and 26B, the variances get smaller as
the test larva gets older. This latter comparison can only be made
for the first four days of response because the ages of test larvae
are variable. These trends indicate that as degree of conditioning
and/or age of test larva increases the response to conditioning
becomes more stable, that is there is less variability in response.
This will be further discussed later.

The results of the analysis in Table 4.3 show that the
majority of responses to conditioning are non-random and in the
direction of attraction, not repulsion, as shown by the mean response
vectors. In the next two sections we will look at whether this
attraction is the same for all degrees of conditioning and age of

test larvae.

134

II. Between-Group Comparisons of Larval
Responses to Increasing Degrees of
Conditioning

 

 

 

There are two ways in which degree of conditioning may be
viewed in the experimental design of Figure 4.2. One is to hold age
of conditioning larvae constant and vary the number of days of con-
ditioning within the three test ages. The assumption here is that
the longer a larva spends in a food lump, the more conditioning it
leaves. The second approach is to hold days of conditioning constant
for the three test ages and vary the starting age of the conditioning
larvae. The assumption is that older (bigger) larvae leave more con-
ditioning than younger (smaller) larvae.

A. Response of Test Larvae to Condition-
ing Done by Varying the Number of Days
of Conditioning Within Any One Age of
Conditioning Larva

The F-values in Table 4.3 indicate that response of test
larvae to conditioning increases as days of conditioning increase at a
constant conditioning larval age. For example, when compared to an
expected vector whose entries are all 0.5, the F-values for the
response of 7 day old larvae to 1, 3, 5, and 7 days of conditioning
done by a 6 day old larva are 0.86, 1.94, 6.82, and 23.48 respectively.
Therefore, to test the hypothesis that response to conditioning
indreases as the number of conditioning days increases at a constant
conditioning age, all between-group comparisons were made where such
a situation exists.

Tables 4.4, 4.5, and 4.6 show the results of these comparisons

at ieach test age. Presented are the comparisons made, the two F-values,

135

the results of Box's Test for Heterogeneity of Variances and the T2
Test for Parallelism of two mean vectors. If two vectors are parallel
it indicates that the responses are not a result of some interaction
over time, but rather the larval response to two different degrees of
conditioning involves the same behavior but at different levels. In
other words, parallel responses indicate that the quality of behavior
is the same but the quantity is different. All responses are parallel,
except for IE versus 105 and all variances are heterogeneous, except
for 27E versus 36E. Figures 4.4, 4.5, and 4.6 graphically show the
mean response vectors for the comparisons in the respective tables.
Examination of Figures 4.4, 4.5, and 4.6 show a definite
trend to increasing response as days of conditioning increase at a
constant conditioning age, regardless of age of test larvae. However,
if we view these graphs in terms of age of test larvae there appears
to be a difference. Older larvae respond at a higher level than do
younger larvae to the same degree of conditioning (whether this is
significant will be tested later). This can be seen, for example, by
comparing Figure 4.4 A, B, and C for 7 day old test larvae. There
is also a trend for the response vectors to become more closely
grouped within a test age as age of conditioning larvae increases.
For example, the three groups of curves in Figure 4.5, reading from
left to right exhibit a tendency to become bunched, possibly indicating
an inability on the part of the responding larva to distinguish
between highly conditioned food beyond a certain level, or above a

Certain level they respond maximally. There may also be a threshold

136

Table 4.4

Results of Hotelling's One(') and Two(*) Sample comparisons of the
response of 7 day old test larvae to varying amounts of condition-
ing (1, 3, 5, and 7 days) done by 6, 9, and 12 day old larvae.

 

 

 

Comparisons F1 F2 Box's TTTY'Test for
EV. CV. Test Parallelism
1E vs. 4E 2.31' 1.46 H P
1E vs. 7E 3.26* P
1E vs. 10E 6.62** H
4B vs. 7E 5.04"' 1.69 H
4B vs. 105 3.75** H
7B vs. 10E 4.97"' 1.24 H P
2E vs. SE 3.44" 1.73 H P
2E vs. 85 4.83"' 1.84 H P
2E vs. 11E 3.35** H P
5E vs. SE 1.58 0.64 H P
5E vs. 11E 2.39* H P
8E vs. 11E 5.20"' 1.45 H P
3E vs. 6E S.30"' 2.06 H P
3E vs. 9E 3.29** H P
3E vs. 12E 4.20"' 2.02 H P
6E vs. 95 4.49"' 1.73 H P
6E vs. 12E 1.03 0.57 H P

12E vs. 9E 2.96" 2.01 H P

Fl(10,70)= 2.24 at P = .025. F2(10,149)= 2.15 at P = .025.

' Fl(10,70)= 2.59 at P = .01. ** F2(10,149)= 2.46 at P = .01.

"Fl(10,70)= 3.48 at P = .001. ***F2(10,149)= 3.24 at P = .001.
EV.= expected vector and CV.= covariance matrix for the One Sample

Test(').

HHeterogeneous variances (P 5.05) .

PMean response vectors are parallel -(P$.05) .

137

Figure 4.4
Mean response vectors of 7 day old test larvae to varying amounts of
conditioning (l, 3, 5, and 7 days) done by 6, 9, and 12 day old
larvae.

A. CAT=(1,3,5,7),6,7
B. CAT=(1,3,5,7),9,7

C. CAT=(1,3,5,7),12,7

138

u

exec a“ oEHH

 

 

 

 

EL e e H e w H Ho
w.wH.wu.H<e "ewH.. x g.
w.wH.eu.H.<u "me. o .8
w.wH.eu.2u "me 111..
w.wH.Hu.H<u "me Ill...

.3.
.8
.8
O
\u .2

a» 5.\... s

/ \

a I .\\ row

0 u s a

{U/ \Axmkx o\\ [[6
\ \ .8
O O \\0
/\ l. H
9:

meme ca oEHH
CH 0 eh m m elm w H.
w.m.wu.H<u "eHH.. u
w.e.eue<u "me o .
weenie "we .11-...
H..e.HuH.<u mw .III:
. w
e
;

1H/(\\‘IIIIIIIIJ\\\\\ \

pood 'puog ut aenxeq %

H; 232

 

 

 

vow

rom

é>
~o

ton

vow

.om

 

.ooH

p005 'pUOD UT 08Al81 %

 

 

 

 

exec cw oEHH
. e e m e m H We 5
r 1 Ho
w.e.wu._.<u £on n ..
w.e.eu.H.<u .mw . . .oe
weenie me 1--....
w.e.Hu.H.<u m: I
.9.
/ .om
// \
)(ll‘II01l011II \ v00
17.x
All you
I x
ollé\l'll6_.\\\ \\
a %\\ ll!
‘\ I"‘\ v00
H02

0
6

JUAJBT

9005 'puoa ug

139

TMfle4£

Results of Hotelling's One(') and Two(*) Sample comparisons of the
reSponse of 11 day old test larvae to varying amounts of condition-
ing (1, 3, 5, and 7 days) done by 6, 9, and 12 day old larvae.

 

 

 

Comparisons Box's T2 Test for
EV. CV. F1 F2 Test Parallelism
13E vs. 16E 2.95* H P
13E vs. 19E 4.91** H P
13E vs. 22E 6.00** H P
16E vs. 19E 5.15"' 2.29 H P
16E vs. 22E 6.44"' 2.40 H P
19E vs. 22E 5.07"' 0.08 H P
14E vs. 17E 3.65" 1.19 H P
148 vs. 2013 f 2.52* H p
14E vs. 23E 4.19** H P
17E vs. 20E 8.11"' 1.34 H P
17E vs. 23E 9.94"' 1.52 H P
20E vs. 23E 2.61' 1.17 H P
1515 vs. 18E 3.81** H p
15E vs. 21E 4.12** H P
15E vs. 24E 4.36** H P
185 vs. 215 4.071! 1.86 H P
18E vs. 24E 7.02"' 2.26 H P
21E vs. 24E 3.29" 1.17 H P
F1(6,74)= 2.59 at P = .025. F2(6,153)= 2.51 at P = .025.
' F1(6,74)= 3.08 at P = .01. * F2(6,153)= 2.95 at P = .01.
"'F1(6,74)= 4.26 at P = .001. ***F2(6,153)= 4.03 at P = .001.
EV.= expected vector and CV.= covariance matrix for the One Sample

Test(').

HHeterogeneous variances

(P 5.05).

PMean response vectors are parallel (P 5.05) .

140

Figure 4.5
Mean response vectors of 11 day old test larvae to varying amounts of
conditioning (l, 3, 5, and 7 days) done by 6, 9, and 12 day old
larvae.
A. CAT=(1,3,5,7),6,11
B. CAT=(1,3,5,7),9,11

C. CAT=(1,3,5,7),12,11

141

.U

exec :H oEHH

Eeefeme

HH.wH.w1e<u "mew

HH.wH.eu.H.<u "me 61111.

HH.wH.euEu EH .1111.
HH.wH.H112u “meH Ill
0
w

0 w
x/o \ x

‘III‘

.05

vow

rom

 

HooH

e.e oaseHe

exec cH oEHH

cWOeeeMwmmH.

p005 'puog ut eenxeq %

HH.e.w1H<u "mew x1111lu
HH.m.e1H<e "mow 6111116
3.56196 “me r11-..
HH.e.H1H.<u um: 1111.11.

vow

uom

roe

wow

wow

vom

 

H85

<

exec :H oEHH

 

 

 

 

2 e 5 m e m H H w H
.H%
HH.e.wue<u Hmww a u 4.
HH.e.e1.H<u "mmH o o .8
HH.e.euH.<u HS .111...
HH.e.H1:a 52 full
.9.
0/0
n...
B
B
9
TL.
U v00
3
m
p 8‘\”
. 0|. 0”. VON.
1d
0
0
D1
. .oe
/ \ x
.8
H55

.‘1 f? A181 of,

'puog ug

Food

142

Table 4.6

Results of Hotelling's One(') and Two(*) Sample comparisons of the re-
5ponse of 13 day old test larvae to varying amounts of conditioning
(l, 3, S, and 7 days) done by 6, 9, and 12 day old larvae.

 

 

 

Comparisons Box's T2 Test for

EV. CV. F1 F2 Test Parallelism

25E vs. 28E 1.46 0.66 H P

25E vs. 31E 7.23*** P

25E vs. 34E 20.56*** H P

28E vs. 31E 4.08** H P

28E vs. 34E 12.97*** H P

31E vs. 34E 5.04*** H P

26E vs. 29E 3.29* H P

26E vs. 32E 4.50** H P

26E vs. 35E 9.37*** H P

29E vs. 32E 0.99 0.32 H P

29E vs. 36E 3.01* H P

32E vs. 35E l7.S3"' 1.73 H P

27E vs. 30E 2.33 0.92 H P

27E vs. 33E 4.39** H P

36E vs. 27E 1.25 P

30E vs. 33E 11.13"' 1.82 H P

30E vs. 36E 3.91" 2.13 H P

33E vs. 36E 3.85** H P

F1(4,76)= 2.96 at P = .025. F2(4,155)= 2.88 at P = .025.

" F1(4’76)= 5.59 at P = .01. * 52(4,155)= 5.47 at p = .01.
"F1(4,76)= 5.18 at P = .001. ***F2(4,155)= 4.94 at P = .001.

EV.= :xpeitid vector and CV.= covariance matrix for the One Sample

est .

HHeterogeneous variances (P!E.05).

PMean response vectors are parallel (P5305) .

143

Figure 4.6
Mean response vectors of 13 day old test larvae to varying amounts of
conditioning (l, 3, 5, and 7 days) done by 6, 9, and 12 day old larvae.
A. CAT=(1,3,5,7),6,13
B. CAT=(1,3,5,7),9,13

c. CAT=(1,3,5,7),12,13

144

e.e 6H5mHe

U m <

 

 

 

 

 

 

 

 

exec :H mEHH exec :H oEHH memo :H oEHH
pHemHemHewH SeekeeeeLwH areememwmmw
1Fo 1Fo .hw
2.2.520 nwee x a : 2.5.5.20 "me... 71.1.12 4.. 26.7.20 e3... 5. 1.
eH.wH.eu.H.<u "wee o o Hoe eH.w.euio ”eweo 6 5e eH.e.eu.H<u "mHe o1II1. .ee
eH.wH.e1.H<e "mom 1111 2551.26 ”mew 111..-... eH.e.eu.H.<u "mewiaal
eH.wH.Hue<u ”mww 1111111 eH.m.Hue<u "mew 1111111 eH.e.Hue<u "mew 1111111
.ov vow wow
"1 ml
8 8 f
.oe M .oe M oe
e e
T... T... \
.ee 6 .oe u \ .ee
6 e
W 1|I§II6IIO u le
D1 D. 91101
w .on . .on . .on
I In In
x x m S m
\ 8 p. x o .8 P \o .8
. a
\ I, .\ /.\.
0' \ o I‘
1s 0 o. x
\ .om / \\ .8 .8
9110 o n
I. .I \
. /$11: x/\\
9: Sea 8 wood”

 

 

o
6

aUAIBW

'puog UI

pooa

145

effect where any greater amount of conditioning beyond a certain level
does not elicit any greater response of the test larvae.

Although the curves in Figures 4.4, 4.5, and 4.6 exhibit a
trend to increasing response as days of conditioning increase for a
particular conditioning age, the F2-values for the Two—Sample test in
Tables 4.4, 4.5, and 4.6 demonstrate that not all such responses are
significantly different. However, within any group of responses at a
constant conditioning age, but at varying days of conditioning, we
would expect increasing responses as days of conditioning increase.
Examination of Figures 4.4, 4.5, and 4.6 A, B, and C shows this to be
the case and the F2—values confirm this hypothesis. If we look across
any one figure, i.e., 4.6 A, B, and C, a tendency for the curves to
bunch is shown. If the curves are grouping about a common response
level then there should be fewer significant F2-values in the tables
as age of conditioning larvae increases. This hypothesis holds for
all except Figure 4.5C where there are more significant differences
than would be predicted. This may be a result of the fact that
Figures 4.4C and 4.6C have a group behaving in an unpredictable
manner, namely 12E and 36E, whereas as the comparable group, 24E, in
Figure 4.5C is behaving normally.

Groups 12E and 36B did not respond as predicted. Group 12E,
Figure 4.4C, should have been the highest response vector yet it is
in the middle, and group 36E, Figure 4.6C, should have been the highest
but is the lowest response vector. The conditioning for these groups
was done by a 12 day old larva for 7 days. However, after 7 days an

initially 12 day old larva is 19 days old and has pupated. Observations

146

on the experimental dishes revealed that the silken tunnels normally
spun on the food had been chewed up and possibly re-absorbed as these
conditioning larvae pupated. Assuming response to conditioning to be
a response to these tunnels, then in groups 12E and 36E there would
be less conditioning to respond to. Other explanations are possible
as well. For example, wandering and pupating larvae may secrete sub-
stances that repel the test larvae. Group 24E Figure 4.5C, was also
conditioned by a 12 day old larva for 7 days, yet the response of 11
day old larvae to this conditioning was as predicted. There is no
apparent explanation for this except that the discriminating ability
of 11 day old larvae may be greater than either 7 or 13 day old test
larvae. However, the fact that 11 day old larvae respond as pre-
dicted tends to negate any hypothesis about wandering larvae secreting
some sort of repulsive substance, unless repulsion is age dependent.
If the Fl-values for the One-Sample test are examined we find
more rejections of the null hypothesis than with the Two—Sample test.
However, the trends indicated by the Two-Sample test are now signifi-
cant with the One-Sample test. This demonstrates that the reason for
not rejecting many of the null hypotheses is that there is hetero-
geneous variance between groups. In any comparisons which are not
significantly different with the One-Sample procedure we can be sure
that heterogeneous variances are not the reason for failure to reject

the null hypothesis.

147

B. Response of Test Larvae to Con-
ditioning Done by Varying Ages of
Conditioning Larvae at Constant
Days of Conditioning

Table 4.3 indicates that response of test larvae to condition-
ing done by varying ages of larvae at constant days of conditioning
increases as age of conditioning larvae increases. This probably
means that either older (bigger) larvae leave more conditioning than
younger (smaller) larvae in the same unit of time, or that different
ages of larvae secrete different substances. The F—values in
Table 4.3 exhibit an increasing tendency under such conditions. For
example, in the response of an 11 day old larva to 1 day of condition-
ing by a 6, 9, or 12 day old larva, the F-values are 0.49, 4.35, and
7.07, or the F-values for the response of a 7 day old larva to 3 days
of conditioning by a 6, 9, or 12 day old larva are 1.94, 8.50, and
23.97 respectively. To test the hypothesis that response to con-
ditioning increases as the age of the conditioning larva increases at
constant days of conditioning the comparisons in Tables 4.7, 4.8, and
4.9 were made and graphed in Figures 4.7, 4.8, and 4.9. All vari-
ances are heterogeneous, and except for comparisons 1E versus 3E and
16E versus 18E, all response vectors are parallel.

The results of these comparisons show the same trends in
response as for varying days of conditioning within any one age of
conditioning larvae. That is, as age of conditioning larva increases
the response vectors increase. For example, the following trends are

evident from Figure 4.7, and similar trends are found in Figures 4.8

and 4.9.

148

Tflfle4fl

Results of Hotelling's One(') and Two(*) Sample comparisons of the re-

response of 7 day old test

larvae to conditioning done by 6, 9, and

12 day old larvae as days of conditioning are held constant at 1, 3,

5, and 7 days.

 

 

 

Comparisons Box's T2 Test for
EV. CV. F1 F2 Test Parallelism __
1E vs. 2E 3.62 1.39 H P
1E vs. 3E 2.23* H
2B vs. SE 0.88 0.37 H P
4E vs. SE 3.66"' 1.27 H P
4E vs. 6E 3.07** H P
5E vs. 6E 3.01" 1.04 H P
7E vs. SE 1.57 0.58 H P
7E vs. 95 2.21* H P
8E vs. 9E 5.92"' 1.36 H P
10E vs. 11E 4.20"' 1.78 H P
105 vs. 12E 2.51' 1.46 H P
12E vs. 11E 2.48' 1.83 H P
" Fl(10’70)= 2.24 at P = .025. F2(10,149)= 2.15 at P = .025.
= = * = :-
Fl(10,70) 2.59 at P .01. F2(10,149) 2.46 at P .01.
11 = = *** = =
F1(10,70) 3.48 at P .001. F2(10’149) 3.24 at P .001.

EV.= expected vector and CV.= covariance matrix for the One Sample

Test(').

HHeterogeneous variances (Pé.05).

P
Mean response vectors are

parallel (PfE.05).

149

Figure 4.7
Mean response vectors of 7 day old test larvae to conditioning done
by 6, 9, and 12 day old larvae as days of conditioning are held con-
stant at l, 3, 5, or 7 days.

A. CAT=1,(6,9,12),7

B. CAT=3,(6,9,12),7

O

CAT=5,(6,9,12),7

D. CAT=7,(6,9,12),7

 

 

 

 

 

 
 

 

 

 

 

 

 

100‘ A 100' B
90- 90' 1...../“"
a As, ',.
80W 80: / 1” V \‘
'8 g "V \
1?. x~ "\ ”(3 "I \
, 70* xzx‘x/ " x , 70- /r \ 11
g ’rA\\ ’\ "é /
8 x I “- \ U60
60' ’ \ '
,5 /r’ .5
8 '/ 8
a 50 I i 50
N N
—J u—l
O fio
9 40' 40
--——-1 1E: CAT=1, 6,7 -—-——. 4E: CAT=3, 6,7
d -.---.. 213: CAT=1, 9,7 3 ..-__-. 515: CAT=3, 9,7
31. 11.-——. 515: CAT=1,12,7 fl 11——-—11 6E: CAT=3,12,7
{I 1 ' V v ' ' V V V T l I V V I I V I v fi
12545678910 12545678910
Time in Days
1001 C 1001 D
/"\ /’\
1 ~ '\ \
“"::—-” 7" 7‘ u ‘
80" \ 80
‘6’ \ E
u? \ u.
,70‘ .70‘
'U 'U
C C!
860‘ I 8601
.5 .5
d) O
(U
2.8 :50
g 3
9‘9 40‘ 6‘40‘
.._._. 7E: CAT=5, 6,7 —— 1013: CAT=7, 6,7
-»-~ 8E: CAT=5, 9,7 ...--...1113; CAT=7 9,7
50 11 11 9E: CAT=5,12,7 3 _ .1——-12£: CAT=7,12,7
CT j U V I V I V I U ‘ H V ' V V I I I I I I
12345678910 12345678910

Time in Days

Figure 4.7

Time in Days

151

Table 4.8

Results of Hotelling's One(') and Two(*) Sample comparisons of the re-
sponse of 11 day old test larvae to conditioning done by 6, 9, and 12
day old larvae as days of conditioning are held constant at 1, 3, 5,
or 7 days.

 

 

 

Comparisons Box's T2 Test for

EV. CV. F1 F2 Test Parallelism

13E vs. 14E 4.79"' 2.13 H P

13E vs. 15E 2.84* H P

14E vs. 15E 4.25" 2.33 H P

16E vs. 17E 5.28"' 2.48 H P

16E vs. 18E 4.18*** H

17E vs. 18E 8.11"' 1.84 H P

19E vs. 20E 1.79 0.68 H P

19E vs. 21E 7.49"' 1.37 H P

20E vs. 21E 9.71"' 1.90 H P

22E vs. 23E 1.49 0.59 H P

22E vs. 24E 7.44"' 2.34 H P

24E vs. 23E 1.90 1.09 H P

F1(6,74)= 2.59 at P = .025. F2(6,153)= 2.51 at P = .025.

' F1(6,74)= 3.08 at P = .01. * F2(6,153)= 2.95 at P = .01.
"F1(6,74)= 4.26 at P = .001. ***F2(6,153)= 4.03 at P = .001.

EV.= expected vector and CV.= covariance matrix for the One Sample

Test(').
HHeterogeneous variances (P5E.05).

PMean response vectors are parallel (PfE.05).

152

Figure 4.8
Mean response vectors of 11 day old test larvae to conditioning done
by 6, 9, and 12 day old larvae as days of conditioning are held con-
stant at l, 3, 5, or 7 days.
A. CAT=1,(6,9,12),11
B. CAT=3,(6,9,12),11
C. CAT=5,(6,9,12),11

D. CAT=7,(6,9,12),11

 

 

153

 

 

 

 

 

 

 

 

 

 

 

 

lOO'lA 1001 B
90" 90'
80‘ ’ 80'
5 m "4
11. 1‘ u.
, 70" ,2 , 70
'o a" '9
a ‘V’ c
o J' -,3
U
60" 60‘
E * .5
Q Q)
m m Oi
g; 50‘ Z 5
:3 :3
‘9 40' .5 40 '
...,_.._... 155: CAT=1, 6,11 .-———H 165: CAT=3, 6,11
~—---. 145: CAT=1, 9,11 Hm-.. 175: CAT=3, 9,11
501 11 11155: CAT=1,12,11 30 ,1 1. 18E: CAT=3,12,11
6T 1 (I . . . . . - - - ._.
10 12345678910
Time in Days
1001 100‘ D
M1?
901 90 /I \
I
>11
801 80* I
'U '9 I
O O
,2 12
.701 .70“
“U 'U
5 E
0601 U6
.5 51
o m
0"504 25
i a
,3 H1
«1 “’
401 4
.—-—.. 195: CAT=5, 6,11 .-———-. 225: CAT=7, 6,11
50 .......... 205: CAT=5, 9,11 3 o----- 255: CAT=7, 9,11
__ 11—-—-11 215: CAT=5,12,11 -ao-—-x 245: CAT=7,12,11
0T v v v u f v V I I v E v r I v v I v T v— I
12345678910 12345678910

Time in Days

Figure 4.8 Time

in Days

154

Table 4.9

Results of Hotelling's One(') and Two(*) Sample comparisons of the re-
sponse of 13 day old test larvae to conditioning done by 6, 9, and 12
day old larvae as days of conditioning are held constant at 1, 3, 5,
or 7 days.

 

 

 

Comparisons Box's T2 Test for
EV. CV. F1 F2 Test Parallelism
25E vs. 26E 4.931! 0.91 H P
25E vs. 27E 6.79"' 2.86 H P
26E vs. 27E 4.34" 1.78 H P
28E vs. 29E 3.35** H P
28E vs. 30E 3.46* H P
29E vs. 30E 1.33 0.31 H P
31E vs. 325 7.61"' 2.14 H P
31E vs. 33E 6.97 1.68 H P
32E vs. 33E 3.81" 1.12 H P
35E vs. 34E 0.63 0.32 H P
36E vs. 34E 8.52*** H P
36E vs. 358 6.75*** H P
F1(4,76)= 2.88 at P = .025. F2(4,155)= 2.88 at P = .025.
5 = = * = =
F1(4,76) 3.59 at P .01. F2(4,155) 3.47 at P .01.
H ._.._ = *1": , ._.
F1(4,76) 5.18 at P .001. F2(4,155) 4.94 at P .001.

EV.= expected vector and CV.
Test(').

covariance matrix for the One Sample

“Heterogeneous variances (P 5.05) .

l)Mean response vectors are parallel (P 5.05) .

155

Figure 4.9
Mean response vectors of 13 day old test larvae to conditioning done
by 6, 9, and 12 day old larvae as days of conditioning are held con-
stant at l, 3, 5, or 7 days.
A. CAT=1,(6,9,12),13
B. CAT=3,(6,9,12),13
C. CAT=5,(6,9,12),13

D. CAT=7,(6,9,12),13

 

 

100‘ A
90‘
80" H
"g /
1E )1
.701 /
'U *‘su‘
C‘.
O
U 1
1: 60
-1-1 ___.__./
8 1
g 50
t6
1.]
89 I
40
--—-. 255: CAT=1, 6,15
.----. 265: CAT=1, 9,15
501 5...... 275: CAT=1,12,13
01v.—~‘ ' v v v V "V V
1 2 5 4 5 6 7 8
Time in Days
10m

00 no
‘1 Q

\1
9

% Larvae in Cond. Food
U1 0‘
Q Q

4:.
Q

64
Q

 

L
0T

...—_. 515: CAT=5, 6,15
.---.. 525: CAT=5, 9,15
1...... 555: CAT=5,12,13

' V

12345678910

Time in Days

156

1001
901
801
70

601

SC

% Larvae in Cond. Food

40

 

9G

00
q

\1
Q

 

% Larvae inmCond. Food
EL

—— 285: CAT=3, 6,15
.--—- 295: CAT=3, 9,15

11....._..1 505: CAT=3,12,13

v v v

12545678910

Time in Days

1001 of...“

I
l

/

Y

I—-g\‘)

Figure 4.9

 

50
4
.———-.545: CAT=7, 6,15
H..-»- 555: CAT=7, 9,15
3 x 11565: CAT=7,12,13

 

545678910
Time in Days

1 2

157

1. Within any particular day of conditioning and age of test
larva, response to conditioning increases as the age of the
conditioning larva increases (Figure 4.7A, B, C, or D).

2. As days of conditioning increase (Figure 4.7A, B, C, and D)
the response to conditioning by a particular age of test larva
increases. That is, the response of a 7 day old larva is
greater to 7 days of conditioning by a 9 day old larva than
it is to 1 day of conditioning by a 9 day old larva.

3. As days of conditioning increase (Figure 4.7A, B, C, and D)
there is less discrimination in response to conditioning by
the 3 ages of test larvae. For example, the response vectors
in Figure 4.7D are at a higher level and more bunched than in
Figure 4.7A.

Even though these trends are evident in Figures 4.7, 4.8,
and 4.9, the F-values in Tables 4.7, 4.8, and 4.9 indicate that there
are only a few significant differences in response to conditioning
done by the three ages of conditioning larvae within a constant test
age and constant days of conditioning. This may be a function of the
ages of conditioning larvae chosen for these experiments, i.e., 6, 9,
and 12 days old. If the ages had been more spread out, i.e., 5, 10,
and 15 days old, the significant differences should have been more
numerous.

However, this lack of significant differences in response to
conditioning by different aged larva is more probably a function of
heterogeneous variances. The Fl-values in Tables 4.7, 4.8, and 4.9

exhibit many more significant differences among comparisons.

158

The problem with the response vectors for groups 12E, 24B,
and 36B is again evident.
III. (getween-Group Comparisons of the

Effects of Larval Age on Response
to Conditioning W

 

 

 

Table 4.3 indicates that there may be age differences relating
to larval response to homotypic conditioning. If such differences
exist, the F-values for the response to the same degree of conditioning
by the same aged conditioning larva should vary, i.e., increase, as
age of test larvae increases. For example, in Table 4.3 the F-values
for response of 7, 11, and 13 day old test larvae to 3 days of con-
ditioning done by a 9 day old conditioning larva (groups 5E, 17E, and
29E) are 8.59, 17.70, and 25.35, respectively. This indicates that
older larvae respond at a higher level than younger larvae to the
same degree of conditioning.

Since the response vectors for 7, 11, and 13 day old larvae
are not compatible time-wise, only the first 4 days of response were
used for each comparison. The results of these comparisons are shown
in Tables 4.10, 4.11, and 4.12 and the corresponding Figures 4.10,
4.11, and 4.12. All comparisons exhibit heterogeneous variances and
all comparisons are parallel.

With the Two—Sample comparison (F2-values) only significant
differences in response due to age show up when there are 7 days of
conditioning done by 6, 9, or 12 day old larvae. Groups 128, 24B,
and 365 have been ignored because they are atypical, as previously
discussed. Therefore, age differences prevail when CAT=7,6(7,13)

[comparison 10E versus 345, Table 4.10, Figure 4.10D], CAT=7,6(11,13)

159

Table 4.10

Results of Hotelling's One(') and Two(*) Sample comparisons for age
differences in response to conditioning. The responses of test lar-
vae (7, 11, and 13 days old) to 1, 3, 5, or 7 days of conditioning
done by a 6 day old larva are tested.

 

 

 

Comparisons Box's T2 Test for
EV. CV. F1 F2 Test Parallelism
135 vs. 15 8.33"' 1.26 H P
18 vs. 25E 12.19"' 1.82 H P
13E vs. 25E 1.55 0.72 H P
4E vs. 16E 3.00' 1.36 H P
4E vs. 28E 0.51 0.13 H P
28E vs. 16E 2.53 1.24 H P
7E vs. 195 3.29' 1.35 H P
7E vs. 31E 7.97"' 2.77 H P
195 vs. 31E S.85"' 2.32 H P
105 vs. 225 2.16 0.89 H P
108 vs. 34E 4.95** H P
22E vs. 345 6.35** H P
Fl(4 76): 2.96 at P = .025. F2(4 155): 2.88 at P = .025.
0 = = * = . =
F1(4,76) 3.59 at P .01. F2(4’155) 3.47 at P .01.
" = = *‘k‘k = :
F1(4,76) 5.18 at P .001. F2(4,155) 4.94 at P .001.

EV.= expected vector and CV.= covariance matrix for the One Sample
Test(').

HHeterogeneous variances (p E .05) .

pMean response vectors are parallel (P £305).

160

Figure 4.10
Mean response vectors of 7, 11, and 13 day old test larvae to l, 3, S,
or 7 days of conditioning done by a 6 day old larva.
A. CAT=1,6,(7,11,13)

5. CAT=3,6,(7,11,13)

O

CAT=5,6,(7,11,13)

C3

CAT=7,6,(7,11,13)

161

 

 

 

 

 

 

   

 

 

 

 

 

100' A 1001 B
90‘ 90‘
80‘ 80'
'U
“:3 8
LL. “\
u: 701 . 701 ‘~..._
5 '3 x
,0 8
J 60‘
.‘E. .5
C) a)
3 350‘
H 5
3 .1
«99 °‘°
4d
.———-. 15: CAT=1,6, 7 .._._. 45: CAT=3,6, 7
0, -.......155: CAT=1,6,11 30 o----ol6E: CAT=3,6,11
3.“ 11 112515: CAT=1,6,13 fl 111—12813: CAT=3,6,13
J,,--.--...J...-----._.
12345678910 12345678910
Time in Days Time in Days
lomc 100‘
go 90
1
801 1380
"U
E: E
.701 67d I
'O
8 5 J
U6d U6
.5. .5
Q) Q)
‘3 Cd
:50- - :35
:3 :3
o\° do
4d 4
r———. 75: CAT=5,6, 7 O--—-—o 10E: CAT=7,6, 7
O-c---o 195: CAT=5,6,11 3 .......... 225: CAT=7,6,11
30 5—1 515: CAT=S,6,13 x 8545: CAT=7,6,13
db
OLV V V U 1" V V U I V 6f V I V V V T V I V 1
12345678910 1234567891

Time in Days Time in Days

Figure 4.10

162

Table 4.11

Results of Hotelling's One(') and Two(*) Sample comparisons for age
differences in response to conditioning. The responses of test lar-
vae (7, 11, and 13 days old) to l, 3, S, or 7 days of conditioning
by a 9 day old larva are tested.

 

 

 

Comparisons Box's T2 Test for
EV. CV. F1 F2 Test Parallelism
2E vs. 14E 4.06" 1.72 H P
25 V5. 265 6.52"' 1.90 H P
265 V5. 14 4.71" 1.19 H P
55 V5. 175 5.82"' 2.71 H P
55 V5. 295 5.24"' 1.96 H P
175 V5. 195 5.96"' 1.39 H P
85 V5. 205 1.70 0.62 H P
85 V5. 325 6.15"' 1.96 H P
205 VS. 325 2.64 1.57 H P
115 V3. 235 6.52"' 2.05 H P
115 V5. 355 5.47*** H P
235 V5. 355 9.59'1' 1.30
F1(4,76)= 2.96 at P = .025. 52(4’155)= 2.88 at P = .025.
' F1(4,76)= 3.59 at P = .01. ** 52(4’155)= 3.47 at P - .01.
"F1(4,76)= 5.18 at P = .001. ***52(4’155)= 4.94 at P = .001.

EV.= expected vector and CV.= covariance matrix for the One Sample
Test(').

HHeterogeneous variances (P $.05).

PMean response vectors are parallel (P 3.05) .

163

Figure 4.11
Mean response vectors of 7, 11, and 13 day old test larvae to l, 3, 5,
or 7 days of conditioning done by a 9 day old larva.
A. CAT=1,9,(7,11,13)
B. CAT=3,9,(7,11,13)
C. CAT=5,9,(7,11,13)

D. CAT=7,9,(7,11,13)

164

 

 

 

 

 

 

 

   

 

 

 

 

 

 

100) A 100‘ B
90‘ 90'
,2. >9
80' 80' / V
v p
O O
o 0 7
u. y LL
_ 70 , 70
v v
c c
o 0
U ‘ U i
c 60 c 60
".4 -H
3 . 8
a 50 i 50'
m m
5—1 1—1
o\° o\°
40' 40‘
—. 2E: CAT=1,9, 7 .-———-o 5E: CAT=3,9, 7
3 ....._..14E: CAT=1,9,11 30. ~——-.—.17E: CAT=3,9,11
4L; ‘26E: CAT=1,9,13 __x x29E: CAT=3,9,13
nT-..-.....-J..-..,--
1234567891 12345678910
Time in Days Time in Days
10% C 1001
It
90' 4/ \§
"
80‘
'U '0
o O
o O
U- u.
_70‘ .
5 5
8 o
u ‘ u
c 60 c
-r-4 "'1
550' {.00
c!) (11
.J -J
°°44r "40
——-———~ 8E: CAT=5,9, 7 ._————. 11E: CAT=7,9, 7
-----20E: CAT=5,9,11 30‘ ~------- 23E: CAT=7,9,11
3%: 4525: CAT=5,9,13 "__ v: x 555: CAT=7,9,13
CL I I I I ' V I I V ()T—' I I I U I I I I '
12345678910 12345678910

Time 1n Days Figure 4.11 111116 1n Days

165

Table 4.12

Results of Hotelling's One(') and Two(*) Sample comparisons for age
differences in reSponse to conditioning. The responses of test lar-
vae (7, 11, and 13 days old) to 1, 3, 5, and 7 days of conditioning
by a 12 day old larva are tested. ‘

 

 

 

Comparisons Box's T2 Test for
EV. CV. F1 F2 Test Parallelism
3E vs. 15E 0.63 0.33 H P
3E vs. 27E 2.44 1.20 H P
lSE vs. 27E 1.73 0.73 H P
6E vs. 18E 9.46"' 2.20 H P
6E vs. 30E 4.04" 1.36 H P
30E vs. 18E 6.47"' 2.34 H P
9E vs. 21E 3.69" 0.97 H P
9E vs. 33E 7.33"' 2.77 H P
21E vs. 335 5.76"' 1.81 H P
12E vs. 24E 5.30*** H P
12E vs. 36E 2.43 0.72 H P
36E vs. 24E 3.95** H P
F1(4,76)= 2.96 at P = .025. F2(4,155)= 2.88 at P = .025.
' = = * = =
F1(4,76) 3.59 at P .01. F2(4,155) 3.47 at P .01.
'1 -_- = *** = :
F1(4,76) 5.18 at P .001. F2(4,155) 4.94 at P .001.

EV.= expected vector and CV.= covariance matrix for the One Sample
Test(').

HHeterogeneous variances (PS.05).

PMean response vectors are parallel (P:E.05).

166

Figure 4.12
Mean response vectors of 7, 11, and 13 day old test larvae to l, 3, 5,
or 7 days of conditioning done by a 12 day old larva.
A. CAT=1,12,(7,11,13)
B. CAT=3,12,(7,11,13)
C. CAT=5,12,(7,11,13)

0. CAT=7,12,(7,11,13)

 

 

167

 

 

 

 

 

 

 

 

 

 

 

100‘ A 100' B
’1'
90' 90- /
i‘
(x "
30' 80‘ /
'U H 'U I
O O
o 1” 0
m . ’ m J
, 70 /__,i . 70'
U Y I '2
S I o
D . I U
c 60 I 1: 60‘
.H .H
8 8
a 50‘ :1 50‘
:11 (U
.—1 .4
o\° I o\°
40 40
-———-. 55: CAT=1,12, 7 —-——. 65: CAT=3,12, 7
-_-..155: CAT=1,12,11 ..-....185: CAT=5,12,11
30j .1 112713: CAT=1,12,13 3" .. 5505: CAT=3,12,13
{I_" _ "fitLW
12345678910 12545678910
Time in Days Time in Days
100‘ 100' D
I \
90' / V’
l
l
80‘
v F r..
E E ..
. _70
p 'U -
6
.2 .5 C1
8 8
>50 >50
8 8
.—J .J
°\° 4o- 6940‘
..__. 95: CAT=5,12, 7 .___. 125; CAT=7,12, 7
.---.. 215: CAT=5,12,11 ......... 245: CAT=7,12,11
3%.: x505: CAT=5,12,13 3‘: x 11565: CAT=7,12,13
CT 7 V V I ' V V ‘U I ' J I I’ V v V I I I I 1
12345678910 12545678910

Time in Days

Figure 4.12

Time in Days

168

[comparison 255 versus 34E, Table 4.10, Figure 4.100], and CAT=7,9
(7,13) [comparison 11E versus 35E, Table 4.11, Figure 4.11D].

However, examination of Figures 4.10, 4.11, and 4.12 reveals
some apparent age differences in response which the Two—Sample analy-
sis is not picking up. For example, in Figure 4.11A curves 2E and
26B appear very different but the Two-Sample procedure says they are
the same. Again we have the problem of heterogeneity of variances.
When all of these comparisons were rerun with the One-Sample approach
(Fl-values in Tables 4.10, 4.11, and 4.12) many more age differences
become apparent, and not just at high levels of conditioning. Rather
now the trend is a consistent one of age differences in response
being at all levels of conditioning, which biologically makes more
sense. Comparisons of Figures 4.10, 4.11, and 4.12 reveals that 7 day
old larvae consistently respond at a lower level to the same degree
of conditioning than 11 or 13 day old larvae and 11 day old larvae are
variable, being intermediate between 7 and 13 day old larvae or not
different from 13 day old larvae. Figure 4.120 shows the "atypical"
groups (12E, 24B, and 36E) as earlier discussed.

If these differences are age-related they should disappear
when comparing segments of the response vectors where ages are com-
patible. For example, comparing the first 4 days of the 13 day old
response vector with the last 4 days of the 7 or 11 day old vectors,
would result in comparing each test age group from the time the larvae
are 13 days old until they are 16 days old. To get some indication if
the responses in Tables 4.10, 4.11 and 4.12 are age-related these com-

parisons were only made for those groups shown to be different for

169

the first 4 days by the Two-Sample test. The results are shown in
Table 4.13.

The results in Table 4.13 show that, except for comparison 22E
versus 348, the response vectors remain significantly different even
for comparable segments of the vectors. It will also be noted that
the two significant comparisons with the Two-Sample procedure exhibit
non-parallel responses, which means that the larvae may now be
exhibiting different behaviors, rather than different degrees of the
same behavior. However, we must be very careful about interpreting
these results because of the confounding variables, early experience
in the colony dishes and length of time in the experimental situation.

These will be discussed later.

Table 4.13

Results of Hotelling's One (') and Two (*) Sample Tests on comparable
4 day segments of the response vectors shown to exhibit age differ-
ences by a Two-Sample comparison in Tables 4.10, 4.11, and 4.12.

 

 

 

Comparisons F1 F2 Box's T2 Test for
EV. CV. Test Parallelism
10E vs. 34E S.70** H

22E vs. 34E 20.09" 2.79 H P

11E vs. 35E 5.13** H

0 = _

F1(4,76) 2.96 at P - .025.

= 5.18 at P = .001.

H
F2(4,76)

HHeterogeneous variances (P_: .05).

PMean response vectors are parallel (P §_.OS).
= 2.88 at P = .025.
= 3.47 at P .001.

*F
2(4,155)

*5
2(4,155)

*

170

Since the groups in Figure 4.2 were run at different times
and with larvae from different populations the question arose as to
whether behavior is constant over time and from group to group. As a
test of this question, replications were periodically run and their
mean response vectors compared to the original experimental groups.
Nineteen such replications (Figure 4.13) were performed and the
results are shown in Table 4.14. The replicates (R) are numbered as
their corresponding experimental groups, the only differences being
that the replicates were run at some time after the experimental
groups and with different larval hatches. In every case, the F1-
and F2-values show no difference, but in 13 of 19 comparisons the
variances are heterogeneous, which means that behavior is fluctuating
from population to population, although the response examined is con-

stant.

IV. A Brief Analysis Relating to Initial
Preference to_Conditioned Food andv

the Amount of Larval Movement as
Conditioning Increases

 

 

A. Does Larval Initial Preference to
Conditioned Food Change as Degree
of Conditioning Increases?

It can be seen from examination of Figures 4.4, 4.5, and 4.6
that most initial preferences, i.e., the first point of each response
mean vector are probably not good predictors of maximum response to
X degrees of conditioning. However, a comparison of initial prefer-
ence to low versus high conditioning will give us some idea if dif-
ferences do exist. The comparisons made deal with CAT=1,9(7,11,13) vs.

CAT=7,9(7,11,13). The analysis is by means of a 2 x 2 contingency

171

.N.v madman a“ cmfimou Hancoswuomxo on» scum museum wouooaom xfisovcmu mo msowumowamom

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

nH.v oasufim
mA\NH\N wh\mfl\m
«we
- p , zmv UV.
2 9
A\HH\NH mA\NH\N «A\m\NH VA\NH\N A\~fi\~fl an.
mmm- men «mm- mam mom. raw
I,
m...
mn\mfi\n - - —— .\a
mtoin 2:... «5&2 mimic {EH23 I.
V5\HN\HH

«mm mHN «AH mo“ mm” q.-
1 w w v
eh\~H\m un-
S M
m5\mfl\m Ln .4.

Naofl mu\o\v m5\0\¢ ¢A\NH\m

<5\V\NH
Hmofi an «m an
E o o S o S o o a. o o 48..
k m _ 49:91

 

 

 

 

 

OZ.ZO_.:DZOU *0 mmumvmo

172

Amonmumu moucmanwfi mfioocomonouozm

.o>fim=Hocoo:O mm: osfim>uo

 

 

 

 

 

O
A Q
.OO. n O OO ON.N u OO OOO ON.O u OO .ONO. n O OO OO.N uOOOO OOONO. .ONO. n O OOO.N.N "OON OOOOO.
OOO. ONO. OOO. OOO. OOO. OOO. OOO. OOO. OOO. NON. OOOOOOO
ONO. ONO. ONO. OOO. OOO. OOO. OOO. ONO. OOO. NON. OOONO OOOO
: OOO. ONO. OOO. OOO. OOO. OOO. OOO. OOO. NOO. NON. OOMO .O>
.OO.O OOO. ONO. ONO. OOO. NOO. OOO. NOO. NOO. ONO. ONN. OOOON OOO
OOO.O OO O OO O NON. ONO. ONO. OOO. NOO. ONO. NOO. NOO. OOO. ONN. OOON.
NON. OOO. OOO. OOO. OOO. NOO. OOO. ONO. OOO. OON. OOOOOMO
ON
ON.O NOO. NON. OON. ONN. NON. OOO. ONN. NON. NON. OOO. ONO .O>
ONO.O OO O NON. ONN. ONN. NON. NON. NOO. OON. NON. OON» ONO. ON“. ON
OON. OON. OOO. OOO. NOO. OOO. OOO. OOO. NON. OON. OOOOOMO
OO
OOO.O ONO. ONO. OOO. ONN. ONO. NOO. ONN. NON. ONN. ONO. OON .O>
OOm.O, NO O NOO. NON. NOO. OOO. OOO. OOO. OON. NON. OON. NOO. OMMI, OO
NON. NON. OON. OOO. OON. OON. ONN. ONN. OON. NON. OOOOOMO
OO
OOO.O OO.O ONO. NOO. ONN. OON. ONN. NON. NOO. NOO. OOO. NOO. OOx .O>
OOO.O ONO. NON. NON. NON. NOO. OON. ONO. OOO. ONO. NOO. OON. OO
OOOe O NO OO OO O O N O O O O N O OOOOOO
Opmom mxmo :O oeflh

 

.goum: cu nova:

Scum bum mafia uo>o wcfimcmgo m“ How>mson Hugues: mo pump a mm mcofiumofiOmon Hausa an“: ~.v ohswflm
scum mmsopm OmucoEwuomxo nepooOom xOEowamH mo mnemfiummaoo oOmamm mOUose paw n.vo:o m.m:fiHHouoz

vH.v oHan

173

.0>Hm3HUr—OUCH v.63 OSHN>IQ

. hmo. w Ac Omocmfihm> Ozoocomoyouoxr

 

 

 

 

 

O
A A
.OO. n O OO ON.N u OO OOO ON.O u OO .ONO. n O NO OO.N «OOOO OOONOO .ONO. n O OO ON.N "OON OOOOO.
OON. OON. NON. OON. NON. NON. OON. OON. OON. OON. OOOOOMO
ONO
OOO.O ONO. ONO. NOO. NOO. OOO. ONO. NOO. ONO. ONO. NOO. ONOO .O>
OOO.O OO O ONO. OOO. NOO. OOO“ OOO. NOO. NOOO.NOO. OOO. NOO. ONMO ONO
OOO. OOO. OOO. OOO. OOO. OOO. OOO. OOO. ONO. OON. OOOOOMO
NNO. OOO. OOO. OOO. OOO. OOO. OOO. OOO. OOO. NON. NOOMO NOOO
O ONO. ONO. ONO. OOO. OOO. OOO. OOO. ONO. OOO. NON. OOOWO .O>
OON.O NON. NOO. NOO. OOO. OOO. NOO. NOO. ONO. NOO. NON. NOOOO OOOO
OOO.O OO O OO.O OOO. ONO. ONO. OOO. NOO. OOO.MOOO. NOO. ONO. ONO” OOOOM
OOO. OOO. OOO. OOO. OOO. OOO. OOO. OOO. NOO. OON. OOOOOOO
N
NNO. OOO. OOO. OOO. OOO. OOO. OOO. OOO. OOO. NON. NOOWO NOOO
: OOO. ONO. OOO. OOO. OOO. OOO. OOO. OOO. NOO. NON. OONO .O>
OON.O NON. NOO. NOO. OOO. OOO. NOO. NOO. ONO. NOO. NON. NOOOO OOO
. ON.O ON.O . . . . . . . . . . OOO
OOO O , NON ONO ONO OOO NOO ONO NOO NOO OOO ONN O
OOOO O NO OO OO O O N O O O O N O OOOONO
m .xom 995 Car 25L»

 

O.O.O:ouO OO.O OOOOO

. mmo.VOV $2829, Ozoocomoufiozx

 

 

174

 

 

 

 

 

.o>OO:Oocoo:O mm: osOm>ua u H .mo. u m um OO.N u an vcw nw.H u 40
Q Q
.ONO. n O OO OO.N "OOOO OONO. .ONO. n O OO OO.N "OON OOOO.
OOO. OOO. ONO. ONO. OOO. NON. OOOOOMO it»
ONO. OOO. OOO. ONO. OOO. NON. OOOmO OOON
O ONO. OOO. OOO. ONO. OOO. NNN. OOWO .O>
OOO.N . ONO. NOO. NOO. ONO. NOO. ONN. OOONm OON
OOO.N OO O ON O OOO. ONO. NOO. ONO. NOO. ONO. OOmm1
NOO. NOO. ONO. OOO. ONO. OOO. OOOOOOO
N
0 O C . C O “NH
NOO ONO ONO OOO ONO OOO NO ONO
O NOO. OOO. ONO. NOO. ONO. OOO. ONMO .O>
OOO.O OOO. OOO. NOO. ONO. NOO. NON. ONOM ONO
OOO.O OO O OO O OOO. NOO. OOO. OOO. OOO. NON. mNmu
ONN. ONN. NON. OON. NON. OON. OOOOOOO
OOm OOO
OOO N ONO. ONO. OON. ONN. NON. NOO. m .O>
ONO.O NN O NOO. NOO. ONO. NOO. NON. ONN. OOMO OOO
OOO. OOO. OOO. OOO. NON. OON. OOOOOMO
. . . . . . OOO
O NNO OOO NNO NON OOO OON NO OOO
OOO. NNO. NON. OON. OON. NON. OOMO .O>
ONO.O ONN. NOO. ONN. ONN. NON. NOO. OOO“ OOO
mmwOO ON O OO O NON. ONN. ONN. NOO. NOO. ONO. OOww
OOOO O NO OO O O O O N O OOOOOO
m.xom mxmo cw mags

 

 

 

O.O.u:ooO OO.O OOOOO

175

.nmo.W& 3053.2; Osoocomoacpozz

 

 

 

 

.®>Mm3HUCOUCfi WN3 mDHN>IQ N H .mo. " am PM OHoN " DO “:6 MWOH " JO
.ONO. n O OO OO.N "OOOO OONOO .ONO. n O OO OO.N "OON OOOO.

OOO. OOO. ONO. OOO. NOO. NON. OOOOOOO

OOO. ONO. ONO. OOO. OOO. OON. ON O
O ONO. ONN
ONO. OOO. ONO. OOO. ONO. OON. NO .O>
ONO.O ONO. NOO. NOO. OOO. NOO. OOO. ONNm ONN

OOO.N NO O OO O NOO. ONO. NOOmLmOO. OOO..mOO. ONOO

ONO. NOO. NOO. OOO. OOO. NON. OOOOOMO
ONO. ONO. ONO. OOO. OOO. OON. NOOMO NOON
O ONO. OOO. OOO. ONO. OOO. NON. OOOmO .O>
OOO.N ONO. ONO. ONO. NOO. OOO. OON. NOON“ OOON

OOO.N OO O OO O ONO. NOO. NOO. ONO. NOO. ONN. OOOOO

OOO. OOO. OOO. OOO. OOO. NON. OOOOOOO

N

ONO. ONO. ONO. OOO. OOO. OON. NOOWO NOON
O ONO. OOO. OOO. ONO. OOO. NNN. OOmO .O>
OOO.N ONO. ONO. ONO. NOO. OOO. OON. NOO m OON

OOO.N OO O OO O OOO. ONO. NOO. ONO. NOO. ONO. OONx
OOOO O NO OO O O O O N O OOOOOO

O.xom Oxmo 5 2:3.

 

O.O.OOOOO O0.0 OOOOO

176

. 30. W E moocmwhm> msoocomohoumm

 

 

 

 

 

 

 

 

O
.o>OO:Oo:oo:w Om: 03Hm>uo u H .mo. n a an OO.N u no paw mm.O u 4o
a a
.ONO. n O OO OO.N "OOOO OONO. .ONO. n O OO OO.N "OON OOOO
illi—
ONO. ONO. OOO. ONO. OOOOOMO
ONO. NOO. NOO. OOO. OOWO
O OOO OOO
ONO. NOO. ONO. ONO. NO .O>
NN.N ONO. NOO. NOO. NOO. OOOx OOO
OO.N NO O ON N ONO. NOO. ONO. ONO. OOO“
ONO. OOO. OOO. ONO. OOOOOMO
ONO. OOO. OOO. ONO. OOMO OOO
O ONO. OOO. OOO. ONO. OOMO .O>
NO.N ONO. OOO. OOO. OOO. OOOM OOO
ON.N OO O ON O ONO. NOO. OOO. OOO. OOmm,
NOO. ONO. OOO. OON. OOOOOOO
OOO. OOO. NNO. OOO. ONNO
O ONN ONN
NNO. NNO. OOO. ONN. NO .O>
OO.N NOO. NON. ONN. OON. ONNM ONN
OO.N OO O OO O ONN. ONN. NON. NOO. ONO“
OON. ONN. ONN. ONN. OOOOOO
NO OON
. . OON .
ON N OOO. ONO. NOO. NOO x O>
. .O -
ON N NN OOO. OOO. NOO. NOO. mem, OON
quH. 9 mm, OO O m N O Omsono
O.xom Oman cw cage

 

0.0.0couO OO.O OOOOO

177

.mmo.WO: Ooocmwnm> Osoocomououomf

 

 

 

 

.o>OO:Ho:oo:O Ow: osfim>ua u H .mo. u a pm OO.N u :9 OOO OO.O u an
Q Q
.ONO. n O OO OO.N "OOOO OONOO .ONO. n O OO OO.N "OON OOOO.
OON. NON. OON. NON. OOOOOOO
xwm mwv
ON.N NOO. OOO. OOO. ONO. O .O>
OOO.O NO O ONO. NOO. NOO. OOO. OOOMI OOO
NON. NON. OON. NON. OOOOOOO
OOW OOO
OOO.N NOO. NOO. NOO. NOO. O .O>
NO.N OO O ONO. NOO. ONO. NOO. OOOM OOO
NOO. NOO. OOO. ONO. OOOOOMO
NOO. NOO. NOO. ONO. OOMO
O OOO OOO
NOO. NOO. ONO. OOO. NO .O>
ON.N NOO. NOO. NOO. OOO. OOOO OOO
NO.N OO O NO O NOO. NOO. ONO. ONO. OON“
OOOe O NO OO O O N O OOOOOO
m.xom mxmo cw oEOH

 

O.O.OOOO OO.O OOOOO

178

table with a Chi—square analysis and the results are shown in
Table 4.15.

The results in Table 4.15 indicate that, at least initially,
7 and 11 day old larvae do not discriminate between low and high
degrees of conditioning. However, when the total mean response
vectors for 2E versus 11B and 14B versus 23B are compared (Tables 4.4
and 4.5) differences in response are detected. Therefore, initial
preference is not a good predictor for long-term preference. The 13
day old test larvae, on the other hand, does discriminate in its
initial response to low versus high conditioning. There would there-
fore appear to be an age difference in the ability to discriminate

between varying degrees of conditioning. However, if the initial

Table 4.15

Chi-square analysis of initial larval preference to the extremes of
conditioning within the three test ages.

 

 

 

Age of Test Com arisons Number in Number in Calculated
Larvae p Cond. food Plain food Xz-value
Low Cond.(2E) 39 41

7 vs. 3.0636
High Cond.(llE) 50 30
Low Cond.(l4E) 51 29
11 vs. 2.9760
High Cond.(23E) 61 19
Low Cond.(26E) 53 27
13 vs. 7.4295**
High Cond.(3SE) 7O 10
2 -
*X = 3.81 at P = .05.
(1)
**X2 = 6.635 at P = .01.

(1)

179

response of 7 day old larvae to high conditioning versus low con-
ditioning (50 versus 30) is compared, by means of a 2 x 2 contingency
table with Chi-square analysis, with the initial response of a 13

day old larva to high versus low conditioning (70 versus 10) a Xz-value
of 13.33 results. This is significant at the .005 level. Therefore,
not only do 13 day old larvae show a higher discriminating between

low versus high conditioning but their initial preference to high

conditioning is greater than for that of a 7 day old larva.

B. Does Amount of Larval Movement Vary
in Different Amounts of Conditioning?

Experiments up to this point have shown that larvae prefer
conditioned food over non-conditioned food and that as conditioning
increases the frequency of larvae found in conditioned food increases.
This would indicate that larvae prefer to settle in conditioned food
and that the number of movements from plain food to conditioned food
should increase as conditioning increases. As a partial test of this
hypothesis it was determined how many times larvae moved to conditioned
food and vice versa, for each of the test ages in a low-conditioned
situation (groups 2E, 14E, and 26E) and a highly-conditioned situation
(groups 11E, 23B, and 35E). The results are shown in Table 4.16. To
equalize number of days spent in the experimental situation, only
counts for the first 4 days were used in this analysis.

The results in Table 4.16 show that, regardless of age, sig-
nificantly more larvae move to the highly conditioned food than move
away from it. At low conditioning, however, only 7 day old larvae show

a significant difference. Examination of the observed frequencies

180

.mo.o u a um va.m u

OOO
Nx

 

 

OOONO.O M MO mewm MW wwwmu OOO OO
OONNO.O MUM MO mewm MW wwwwm OON OO OOOO
OONNO.O MHWM MN mewm MW mw»mm OOO N
OOOO.O MN MW mewm MW wwwmu OON OO
NO.N MHMM mm mewm MW wwwwm OOO OO OOO
OONNO.O MHWM MN meww MW ww»mm ON N
OOOOO. .OOOO .OOOOO OOO.O:

 

.cmO: no 30H nonuflo OO mcficofipflucoo can: mmaaO doom :OOOQ cam wocofluflw
-coo coozpos vm>oe om>OmO OOEOO mo Hones: may no OOOxOmcw paw mo OOocwoom mumscm-O:o a mo OOOSOOO

O0.0 manmh

181

also shows that, regardless of the degree of conditioning, the number
of movements of 7 or 11 day old larvae (younger larvae) are greater
than for older 13 day old larvae. For example, in high conditioning,
7 day old larvae moved a total of 29 times in 4 days as opposed to
only 14 times for 13 day old larvae. In a chi—square goodness of

fit test this yields a XZ-value of 5.2324 which is significant at the
0.025 level. The number of movements by an 11 day old larvae however,
is not different from either 7 or 13 day old movements. No differ-

ences exist between total movements in low conditioning for any of the

three test ages.

Control for Food Limitation

Before making conclusions about the effects of homotypic con-
ditijning on the distribution of isolate larvae, it is first necessary
to demonstrate that food is not a limiting factor. To demonstrate
thiis, an experiment was run in which developmental time on a maximally
conditioned food lump was compared to developmental time on a plain
food lump. A six-group experiment was set up with N=20 per group. In
three groups each dish had a single food lump conditioned by a 9 day
old larva for 7 days. After conditioning the 9 day old larvae were
removed and isolate 7, 11, or 13 day old larvae reared until pupation.
The Other three groups were identical except the food was not con-
ditioned. If developmental time (pupation time) does not differ among

Azfifiese 351)< groups, then food was not a limiting factor in my experiments.
141/1e res’ults of an analysis of variance on this data is shown in
r’g—ble 4~17 and it can be seen that all pupation times were the same.

Tpe means and standard deviations for the 7, 11, and 13 day old

182

Table 4.17

Analysis of variance on developmental times of 7, 11, and 13 day old
larvae reared on single food lumps which were either maximally con-
ditioned or non-conditioned.

 

 

 

 

 

 

 

 

 

 

 

Source of Variation d.f. S.S. M.S. F ratio
Between Group 5 0.7416 0.1483 - 0.2372
Within Group 114 71.2500 0.625
Total 119 71.9916

F(5’119) = 2.2914 at P = .05.

isolates reared on a single unconditioned food lump were 18.35 :_0.55,
18.50 :_0.68, and 18.55 :_0.57, respectively. The means and standard
deviations for the 7, 11, and 13 day old isolates reared on a single
conditioned food lump were 18.55 :_0.78, 18.60 i 0.56, and 18.50 :_0.57,

respectively.

V. Can Response to Conditioning Be
Overridden by Varying the Food Source?

Park (1948) stated that there is probably some kind of inter-
action between conditioning and food source involved in the response
of Tribolium to conditioned food. With Galleria larvae, I was inter-
ested in finding out if larval initial preferences distinguish between
conditioned food sources which are either poor in nutrients or rich in
nutrients. In order to make such a test, a diet (deficient diet) was
needed on which development is delayed. The following diet was used:
Pablum = 43.16g, yeast = 0g, honey = 7.5g, glycerin = 31.16g, water =
18.16g, and beeswax = 0g, whereas the normal diet consists of

pablum = 32.98g, yeast = 9.7g, honey = 24.25g, glycerin = 21.34g,

183

water = 8.73g, and beeswax = 3g. Sixteen 11 day old larvae were
reared on single 4 gram lumps of deficient diet and sixteen on single
lumps of normal diet (the normal diet being that used in all previous
experiments). The ANOVA in Table 4.18 demonstrates that deveIOpment
was slower on the deficient diet, it being 26.12 :_0.52 days on

deficient and 18.37 :_0.3 days on normal food.

Table 4.18

Analysis of variance on 11 day old larvae reared on normal and
deficient diets.

 

 

 

 

 

 

 

 

 

 

Source of Variance d.f. S.S. M.S. F ratio1
Between Group 1 480.5 480.5 168.5964*
Within Group 30 85.5 2.85
Total 31 560.00

*F(l,31) = 13.2222 at P = 0.001.

To determine if larvae exhibit a preference for normal over
deficient food, 16 dishes were set up, each with one normal and one
deficient 4-gram food lump on which no conditioning was allowed. One
11 day old larva was tested for preference on each dish. Data was
collected for initial preference (the preference 5 minutes after being
placed in the experimental situation) and preference 24 hours later.
The results were that all 16 larvae preferred the normal food lump at
both data points.

I next set up an experiment consisting of two groups of 16

dishes each. In one group each dish had two normal food lumps one of

 

184

which was conditioned. In the other group, each dish had one normal
food lump and one deficient conditioned lump. Tests were made with
one 11 day old larva per dish and data collected for initial and 24
hour preferences. The data being analyzed is the number of times
larvae that initially selected the deficient conditioned lump or the
normal conditioned lump switched (flipped) to the normal plain lump
in 24 hours. The results are shown in Table 4.19, where DC =

deficient-conditioned lump and NC = normal-conditioned lump.

Table 4.19

A 2 x 2 contingency table analyzed by chi-square of larval responses
to normal-conditioned and deficient-conditioned food.

 

 

0c NC
Flip 8 0 8
No Flip 2 16 18
10 16 26

 

Calculated X2 = 18.5287*

*x2 = 7.879 at p = 0.005.

There is therefore, an interaction between food source and
conditioning. Initial preference is to conditioning but food prefer-
ences will over—ride the initial preference in the deficient food
situation. It should be noted, however, that there is also a dif-
ference in initial preference for deficient-conditioned versus normal-
conditioned food. Only 10 larvae initially selected the deficient con-
ditioned lump and then 8 of the 10 flipped, whereas all 16 larvae

selected the normal-conditioned lump with no flips.

185
VI. Do Adult Galleria Exhibit Prefer-
ences to Conditioning Done by Larvae?

Adult Galleria (metamorphosed) do not eat and are not normally
found in the larval food source, and conditioning would probably have
very little attraction for them from a feeding point of view. Paddock
(1913) and Oertel (1962) both indicate that adult females exhibit pref-
erences for oviposition sites and Oertel (1962) further demonstrated
that adults are attracted to bee hives which have been larval con-
ditioned. I therefore decided to ask if female adults prefer to lay
their eggs in conditioned food.

Two groups of 20 dishes were set up. In one group each dish
had 2, 4-gram normal food lumps, one of which was conditioned by a
9 day old larva for 3 days (low conditioned). In the other group one
of the lumps was conditioned by a 9 day old larva for 7 days. The
conditioning larvae were then removed and one female adult was placed
in each dish. The females were taken from the breeding tank. After
5 days the eggs laid on each food lump were counted and each group
analyzed by a chi—square goodness of fit test as shown in Table 4.20.

Female adults exhibit no distinction between low-conditioned
food and non-conditioned food for egg laying. However, they prefer

non-conditioned food to highly conditioned food for egg-laying.

Discussion

 

Studies with invertebrates indicate that biological condition-
ing affects spatial distribution. These studies, however, deal with
whole populations of organisms and are confounded by animal-animal

interactions, sexual differences, and various environmental parameters.

186

Table 4.20

Chirsquare goodness of fit test on female adult preference for egg-
laying in low and high larval conditioned food.

 

Observed Expected

 

 

 

 

Comparison Number Number Calculated
of eggs of eggs X -va1ues
Low Cond. Food 5705 5785.5
vs. 1.3980
Plain Food 5832 5768.5
High Cond. Food 3742 5019
vs. 649.8222*
Plain Food 6296 5019
*X2 = 7 879 at P = 0 005
(1) . . .

No sfiystematic attempt has been made to elucidate the effect of con-
ditioning on individual behavior. Park's work (1934, 1935), for
example, is confounded by the presence of every developmental stage
in his culture dishes, and Naylor‘s (1959) studies were unable to
definitively separate conditioning effects from sexual or density
factors. Wellington (1957) developed a bioassay for individual
behavioral response of the tent caterpillar to light, but not to con-
ditioning, and some of the responses of his Type II (less active)
larvae could be interpreted in terms of lack of proper stimuli from
homotypic conditioning. Surtees (1963a, l964e) completely ignored the
effects that biological conditioning may have been having on popu—
lations of Tribolium and various grain beetles, and it is not clear

from Long's (1955) data if aggregation formation and feeding behavior

187

in butterfly larvae are a function of biological conditioning, animal-
animal interactions, or both.

The hypothesis for my studies in this paper has been that the
spatial distribution of isolate Galleria larvae is a function of
larval conditioning and that preference for conditioned medium is a
function of the degree of conditioning and age of responding larvae.

Larval Responses to Increasing
Degrees of Conditioning

The between-group comparisons of larval responses to con-
ditioning showed that response increases as days of conditioning
increase, as age of conditioning larvae increase, or as a function of
the two. Older larvae may be conditioning the medium more than
younger larvae or they may be differentially conditioning the medium.
In every case examined, individual responses to conditioning are either
random or attraction. Avoidance was never seen, but a response
plateau was indicated. There may be a threshold effect in which con-
ditioning beyond a certain level will not elicit a greater response.
This requires further investigation particularly in reference to
whether conditioning beyond that used in my experiments may elicit
avoidance reactions from the test larvae. Such experiments, however,
would need to carefully control for avoidance to food limitation. It
may be that repulsion to conditioning is not an appropriate behavior
'unless the conditioned medium is also food limited. Park (1948) noted
that Tribolium was repelled by excessive conditioning, but his system

was feed limited in that the medium used was two year old colony food.

188

Age Differences in Response to Con-
ditioningand Possible Mechanisms

 

 

The data on age differences in response to the same degree of
conditioning are not as clear. Significant age differences in response
were evident only at the highest degrees of conditioning when the Two-
Sample comparison was utilized and at all degrees of conditioning
when the One-Sample test was used. There is also every indication
(Table 4.13) that these differences do not disappear by comparing
identical segments of the mean response vectors.

Two-Sample comparisons in Table 4.13 show that not only do
age differences not disappear by comparing equal segments of the
response curves, but that non—parallelism tends to result as well,
indicating the presence of interactions other than age or that
behavior is qualitatively different. The age analyses may be con-
founded by differences in larval size, activity, sensory capabilities,

and early experiences, both in the colony and the experimental apparatus.

Size of Responding Larvae

The size of different aged larvae may have been a factor in
their ability to negotiate the distance between food lumps. As shown
in Figure 4.1, the size difference between a 7 and a 13 day old larva
is great. The distance between food lumps was 35 mm across the center
of the dish. Therefore, 7 day old larvae had a greater relative dis-
tance to travel from one food lump to the other than did 13 or 11 day
old larvae. If, for example, a 13 day old larva initially chose the
plain food lump, it could extend part of its body out of that lump

and almost encounter the conditioned lump. A 7 day old larva would

189

have to completely leave the plain food and travel a relatively large

distance before encountering the conditioned lump.

Activity Differences

Galleria larvae may exhibit activity differences dependent
upon age. Activity is being defined as amount of movement between
food lumps, not within food lumps. Wellington (1957) showed that tent
caterpillars are of several types as related to activity levels, some
being active and others being sluggish and less investigatory. How-
ever, these types are genetically determined not age dependent,
although Wellington's data is not very convincing. Long (1955), on
the other hand, demonstrated that butterfly larvae exhibit different
activity rhythms as development progresses. The older larvae became
more active and tended to exhibit many investigatory excursions, and
this developmental trend may be common to Lepidoptera. If 7 day old
Galleria are initially less active and if response to conditioning is
tactual rather than olfactory, they would tend to randomly select a
food llUflp and remain there for a time. As they age, however, their
activity pattern changes to investigatory and they move out to explore
the surroundings. Assuming conditioning to be a preferred stimulus,
when they encounter the conditioned food they stop moving and settle.
This behavior would account for the low initial preferences of 7 day
old test larvae and the gradual initial rise in their response vectors
as they find the conditioned food. The more active 13 day old larvae
investigate their surroundings immediately. Upon encountering con-
ditioned food their preferred stimulus lessens their activity and they

settle. This behavior is reflected in the higher initial preferences

190

shown by 13 day old larvae to high degrees of conditioning. Defini-
tive experimental support for this hypothesis is lacking, but obser-
vational data and the mobility data in section IVA tend to support it.
The mobility data shows more total_movements between food lumps by

7 day old larvae than for 13 day old larvae, which indicates that 13
day old larvae find and settle in the conditioned food early, whereas
7 day old larvae do more exploratory movements.

However, the crucial point is not total activity of the dif-
ferent ages, but when the activity is taking place. If, for example,
we re-examine Table 4.16, we find that 7 day old larvae moved a total
of 29 times to only 14 times for 13 day old larva. This indicates
that 7 day old larvae are more active than 13 day old larvae. However,
the data sheets reveal that 7 day old larvae move a total of 6 times
in the first two days of the experiment and 13 day old larvae moved
a total of 7 times. This means that only 20.68% of the 7 day old
movements occurred in the first two days to 50% for 13 day old larvae.
Therefore, 7 day old larvae are, at least initially, relatively
immobile and are not exploring their environment and finding the con-
ditioned lump as soon as are older more active larvae. Even though
the total mobility data shows 7 day old larvae to be more active than
13 day old larvae, this activity is occurring 3 or 4 days into the
experiment at a time when the initially 7 day old larvae are older
and apparently their activity has increased.

If this hypothesis is true, we would expect the response
‘vectors for all degrees of conditioning to eventually reach the same

level because all larvae will eventually find the conditioned food.

191

This does not happen. When a larva initially selects a plain food
lump, it begins conditioning it. When investigatory movements set in,
the larva may sample the initially conditioned lump but find that it

is not as highly conditioned as the lump it just left and therefore
return to its original choice. For example, the initial response of a
7 day old larva to conditioning done by a 9 day old larva for one day
is low; that is, many 7 day old larvae initially chose the plain

lump. Because their activity is low at this time, they remain in the
plain lump and begin conditioning it. When they are older and their
activity increases they begin to investigate their surroundings and
encounter the initially conditioned lump. However, the conditioning
they did for 3 or 4 days is greater than that done by the 9 day old
larva for 1 day and they return to the "preferred" stimulus and settle.
If the initially conditioned lump had more conditioning than the lump
they left they will settle in it. Thus, the mean response vectors

for low conditioning remain at a lower level than for high conditioning.
These behaviors would necessitate some ability on the part of the
larvae to discriminate between different levels of conditioning.

Although there are indications of this ability in my studies it needs

further testing.

Sensory Differences Between Ages

There may be sensory differences between ages, in which 13 day
old larvaerhave better discriminatory powers for conditioning than do
7 day old larvae. The initial preference analysis in section IVA shows
that, at least initially, 7 and 11 day old larvae do not distinguish

between high and low conditioning, whereas 13 day old larvae do.

 

192

Whatever the mechanism(s), response to conditioning is more highly
developed in older larvae. The analysis shows that there is a dif-
ference in initial preference of 7 or 13 day old larvae to highly
conditioned food which tends to support a hypothesis relating to
differential sensory ability. However, this initial preference data
may also be interpreted in terms of activity differences. The signifi-
cant initial preference of 13 day old larvae for conditioned food over
plain food may be a reflection of their greater activity and early
exploration of the experimental situation, whereas non-significance

in 7 day old initial preferences for conditioned over plain food could
be a reflection of their relative inactivity once any food lump is

encountered.

Early Experience Factors

Early experience factors may also play a role in the age dif-
ferences seen in my experiments. Seven day old test larvae are taken
from the 7 day old colony dish which is less highly conditioned than
the 11 or 13 day old colony dishes. Therefore, the older the test
larva, the more early experience it has had with conditioning and with
higher degrees of conditioning. This may account for a 13 day old
larva's response to a high degree of conditioning being greater than
a 7 day old's, and would be extremely difficult to control. One
approach would be to raise larvae, for testing, as isolates and change
their food every day, but this would not control for self-conditioning.
Even if’the food is changed daily a 13 day old larva would be experiO

encing;its own conditioning which is greater than a 7 day old larva's.

193

Another approach would be to raise all larvae in mass cultures and
base larval age for experimentation on larval size.

Finally, there may be an interaction between age and time
spent in the experimental situation, 7 day old larvae spending more
time in the dishes than 13 day old larvae. This would be difficult
to control and still be able to follow total response mean vectors
for each group. However, time spent in the experimental dishes is
probably not a factor causing the differences in age. The analysis
for age differences was initially with the first four days of the
experimental period, regardless of age, and the differences were
demonstrated.

Response to Biological Conditioning
as a Bioassay of Behavior

The response of larvae to various degrees of homotypic con-
ditioning is a good bioassay of individual behavior. The behavioral
responses to graded series of conditioning has been shown to be con-
stant for any given age and degree of conditioning. The replications
of randomly selected groups in Figure 4.13 and Table 4.14 demonstrate
that the response mean vectors do not change over time or from popu-
lation to population of larvae, i.e., from hatch to hatch. However,
these comparisons show that the variances between populations of
larvae are heterogeneous. The behavioral response to conditioning may
have become evolutionarily fixed with allowances for variability in
how the response is achieved. That is, there are various strategies

by which the same response is achieved.

194

If the response itself is of major interest, as it was in my
studies, and not the variability in response from population to
population, then the heterogeneous variances indicate the desirability
of selectively breeding a "standard" Galleria stock for future experi-
mentation. Response to conditioning could be used as a bioassay of
the behavior being selected for.

As repeatedly mentioned and demonstrated throughout these
Studies the variances are heterogeneous. This has profound conse-
quences on the power of the test to discriminate between a true and a
false null hypothesis. Coupled with the positive serial correlation
in my data two responses would have to be very different before being
detected with the Two-Sample test, even with such large sample sizes.
The power of the test was noticeably increased when the One-Sample
test was employed, mainly because this procedure utilized the lower
covariance matrix as the "best estimate" of the variances. Utilizing
both procedures has enabled some insight into the behavior or Galleria
larvae in responding to biological conditioning.

Aside from being heterogeneous, the variance vectors in
Table 4.3 show a trend of decreasing magnitude as degree of condition-
ing increases. In other words, as conditioning increases the behavior
of the larva tends to become more fixed. They select the conditioned
food lump and show very little variation in response. If, however,
conditioning is low the response is variable, with a good deal of
movement between the conditioned and plain food. One might think of
:response in a low-conditioned situation as an equilibrium state in

vflrich there is a great deal of oscillation between food lumps, but

195

about a predictable level. As conditioning increases, however, these
oscillations decrease in number as more and more larvae select the
conditioned food lump and remain there. This trend in decreasing
variance is also seen in the response of the 3 test ages to the same
degree of conditioning. As larvae get older they tend to select and
stay in conditioned food. This is probably related to the mobility
differences discussed earlier.

There are any number of sources for variation in such experi-
ments. Some of it may be due to environmental variability. Although
light, temperature, and humidity were closely controlled there may
have been undetected differences from experiment to experiment.
However, these differences were not great enough to affect the results
of larval selection as indicated by the replication data in Table 4.14.

Genetic variability may be the main contributor to the hetero-
geneous variances. Many experiments were run with different popu-
lations of larvae. Although controls and replications indicate this
did not affect the behavior being looked at, the genetics may have
‘been changing enough to account for the variability. This is impos-
sible to determine at this time since the genetics of the larvae was
‘not closely regulated or monitored.

An interesting fact from all my tables, however, is that even
‘though the variance of any two groups is heterogeneous, the response
(nirves are parallel. Parallelism indicates that, although at different
levels, the response curves were achieved by similar behaviors. The
heterogeneity may be a result of the degree to which each behavior is

exhibited under varying degrees of conditioning. In selecting low or

 

196

high conditioned food, a larva utilizes its preference behavior for
conditioning but this behavior increases and is less fluctuating as

conditioning increases.

Possible Functions of Biological
Conditioning

 

My studies have demonstrated that the spatial location of
Galleria larvae may be largely determined by their homotypic con-
ditioning. Individual larvae are attracted to conditioning and this
attraction is a function of both the degree of conditioning and age of
the larvae. These studies will be pursued further in the next chapter
in relation to animal-animal interactions and resulting effects on
individual preferences to conditioning.

Aside from its function in spatial distribution, biological
conditioning may be important for other reasons. Some possibilities
are that conditioning (1) serves as protection for the larvae, (2) pro-
‘vides a preferred consistency of the medium, (3) facilitates mobility,

(4) signals the presence of other larvae, and (5) is related to feed-

ing behavior.

Protection

Paddock (1913) hypothesized that spinning behavior in Galleria
larvae serves as a scaffolding to prevent the hive combs from collap-
siJu; as the larvae tunnel throughout the hive and literally decimate it.
Although this is an interesting hypothesis, there is no support for it.

Paddock (1913) also postulated that silk tunnels protect the
larvae from the bees. The only support for this hypothesis is circum-

stantial. Free (1961) noted that bees sting dark colors, rough

197

surfaces, and rapidly moving objects. Galleria larvae, particularly
older larvae, tend to be darkly colored and move in rapid, jerky
movements. In their whitish, smooth, silken tunnels, however, their
color and movements may go undetected. However, the interior of a

bee hive is dark and vision on the part of the bees may not be utilized
in detecting Galleria.

A possible test for both these hypotheses would be to arti-
ficially select for a population of Galleria that do not spin these
tunnels. Comparisons could then be made to see if any selective
advantage accrues to the spinning population.

Biological Conditioning as a
Preferred Consistency

Larvae may prefer the consistency or the "feel" of biologically
conditioned medium over non-conditioned medium. Although not a very
tenable hypothesis, it does have minor support. Beck (1960) found
that larvae preferred certain medium over others. Since he could not
find any adaptive advantage to such a preference, he concluded that
larvae like the consistency. Other workers, however, found a better
explanation. Young (1961) found that the preferred medium enhanced
larval growth, and Chase (1921) earlier noted that larvae raised on
poor diets transform into smaller and less viable adults. However,
consistency may well be a factor in their preferences if, for example,

medjinn is easier to tunnel through.

Mobility

Mobility may be easier or quicker in conditioned than in non-

conditiomed food. This may be related to the silk itself or to the

198

fact that the tunnels provide ready access through the food. Stanley
(1949), for example, demonstrated with Tribolium that tunnel-making
in the food enhances mobility and that the beetles preferred using
existing tunnels rather than constructing new ones.
A Signal for the Presence of
Other-Larvae

Biological conditioning may indicate the presence of other
larvae. If individual larvae are responding to conditioning because
it signals the presence of other larvae it would first be necessary
to demonstrate that a larva is at least willing to tolerate the presence
of other larvae. If it is not, then one would predict an avoidance to
conditioning rather than an attraction. A partial test of this hypoth-
esis is provided in Chapter V.
Biological Conditioning and
Feeding Behavior

A final, and more probable, possibility is that preference
for biological conditioning is related to feeding behavior. Brindley
(1910) and Edwards (1910) demonstrated a relationship between silk
threads spun by the processionary caterpillars and feeding excursions.
Wellington (1957) showed that even relatively inactive larvae will
follow silk threads to a food source, and Rathcke and Poole (1975)

found that butterfly larvae (Mechantis isthmis) cooperate to spin a

 

silk scaffolding over their food supply, enabling exploitation of an
otherwise unavailable food source. Rathcke and Poole further postu-
lated that such a massive silk gallery is energetically feasible only

if several larvae pool their resources and share the benefits. Most

199

larvae spin some silk even if only during pupation and the evolution
of a feeding web would entail only an elaboration of an already
present ability. However, such behavior would require cooperation

and possibly some form of social behavior. Long (1955) also demon—
strated a relationship between biological conditioning and feeding
behavior in larvae of the large white butterfly. Newly hatched larvae
initially locate a food supply by following silk trails. Larvae also
feed in masses, constantly spinning and enlarging a silk mat around
the food supply.

Biological conditioning may signal an exploitable food supply
to Galleria larvae. The higher the degree of conditioning means the
better the food source, and so the greater the response to it, since
the food supply would have to be nutritious enough for a larva to
spend enough time on it to highly condition it. However, very highly
conditioned food might also indicate a food depleted situation which
might not affect larval initial preference for the conditioning.
Haydak (1936), for example, showed that Galleria larvae reared on
deficient diets produce less silk than those reared on normal diets.
In my experiments with deficient diets, I had to use more larvae to
condition a deficient food lump to the same degree as one larva's
cxnuiitioning on normal food, as measured by response of a test larva.
These experiments also show that, although initial preference is to
cxnuiitioning, if the conditioned food source is deficient the larvae

quickly move to a better non-conditioned source.

 

200

Biological Conditioning and
Population Density

 

 

Biological conditioning may function as an index of larval
population density. If this is true, then at some high degree of
conditioning, larvae will avoid the conditioned situation. This was
not observed in my experiments, probably because conditioning was not
high enough, or food not limited as a result of the conditioning.
However, a non-preference for highly conditioned food was seen on the
part of adults to conditioning.

In 1959, Naylor stated that gravid female Tribolium tend to
select, for egg laying, niche units of low adult and larval occupancy
or avoid those of high occupancy. Chiang and Hodson (1950) had earlier

demonstrated that larval Drosophila melanogaster inhibit oviposition

 

of the adults and that oviposition is density-dependent on the larvae.
My studies with egg laying preferences of female Galleria show that
they avoid highly larval conditioned food for ovipositing, but
exhibit a random response if food is only slightly conditioned. The
biological conditioning may be functioning as an index of population
density, and the survival value of the adult females' behavioral
response is obvious. By laying her eggs in low or non-conditioned
niches she ensures that hatching larvae will not be in extreme com-
petition for the available resources which enhances their chances of

survival to breeding age.

CHAPTER V

THE RELATIONSHIP BETWEEN BIOLOGICAL CONDITIONING AND
ANIMAL-ANIMAL INTERACTION IN THE SPATIAL DIS-
TRIBUTION OF GALLERIA MELLONELLA (L.) LARVAE

Introduction

 

In Chapter IV it was demonstrated that isolate Galleria larvae
exhibit a preference for biologically conditioned food, that this
preference increases as the degree of conditioning increases, and that
possibly an age variable enters into the selection of conditioning.
However, biological conditioning is a function of the organisms them-
selves and, as such, may be closely linked with interactions between
,other conspecifics. It is the purpose of this paper to look at how
animal-animal interactions affect the preferences for conditioning
established in Chapter IV.

Animal-animal interactions are known to affect the spatial
distribution of organisms. Surtees (1963a, b, and c; 1964a, b, c,

e and f) found in grain weevils that as density increased, the number
lof weevils appearing on the surface of the medium increased, but he
failed to account for the extra conditioning as a result of increased
densityx A similar oversight was made by Legay and Chase (1964) who
‘hypothesized that dispersal in Tribolium is dependent on exploration
and population pressure due to increasing density. They attempted to
explain their results purely on the basis of animal-animal interactions.

201

202

For example, they found the mean duration of stay of individual
beetles in a partial enclosure to increase with increasing density.
They interpret this as a result of accommodation to tactile stimuli
among the beetles which had not discovered the way out of the enclo-
sure. However, as density rises so does conditioning, and the
"accommodation" to tactile stimuli may have been an increasing pref—
erence for increasing degrees of conditioning.

Naylor (1959 and 1965) found that female Tribolium confusum
usually exhibit a uniform distribution but will form very loose aggre-
gations in conditioning. Males on the other hand, aggregate in con-
ditioning and with other males. His general conclusion was that spatial
distribution in Tribolium is dependent upon density, but if sex and
conditioning are added as variables the spatial patterns become
altered. In 1963, Ghent contrasted the behavioral responses of

Tribolium confusum and I, castaneum in mixed populations and demon-

 

strated that the distributions of these populations depend upon animal-
animal interactions, response to homotypic and heterotypic conditioning,
and differences in behavior between the sexes.

Much of the work with Tribolium and other beetle populations
is somewhat confounded by density and the effects of sexual behavior
(”1 response to biological conditioning. The effects of biological
conditioning and animal-animal interactions may be more clearly seen
511 larval populations free from adult—young and sexual interactions.
Galleria (larva) is such a model but has not yet been worked out.
Edwards (1910) showed that larvae of the processionary caterpillar

stay in feeding excursion lines as a result of silk strands and

 

203

head-to—tail larval contact, the latter being of predominant impor-
tance.

Several larval forms of butterflies exhibit a relationship
between biological conditioning and animal-animal interactions (Long
1955). Silk threads are used to find food but larval—larval inter-
actions serve to integrate the cyclic feeding behavior. Long also
discovered that varying activity rhythms of different aged larvae
affect the ability of individuals to join pre-established feeding
groups. Similar kinds of behavior are noted by Wellington (1957 and
1960) with tent caterpillars. Different larval types can be iden-
tified on the basis of activity. Type I larvae are capable of inde—
pendent, oriented movements, whereas Type II larvae need a silk trail
or animal-animal contact for directed movements. Movements are even
more directed if both stimuli are present.

Although demonstrated that biological conditioning and animal-
animal interactions affect spatial patterns, the two variables are
often confounded at the population level since there is little if any
control over the amount of conditioning populations are doing. There
have been only minor attempts at elucidating the effects of these
'variables on individual behavior. Knowing the relationship between
individual preference for conditioning and degrees of conditioning
(Chapter IV), it is the purpose of this paper to investigate whether
these individual preferences are altered by the presence of another

conspecific in the larvae of Galleria mellonella. The general hypoth-

 

esis being tested is that individual preference for conditioning is

unaffected by the presence of a conspecific and the resulting response

204

is a function of whether the conspecific is a resident in the con-

ditioned food or introduced with the test larva.

Materials and Methods

Husbandry

All experimental larvae were reared as described in Chapter II
and drawn from colony dishes maintained in darkness in incubators at
32 :_1°C and 75% R.H. The experimental larvae used in these experi-

ments are pictured in Figure 4.1 and their weights given in Table 4.1.

Apparatus

The experimental apparatus consisted of 100 x 25 mm plastic
petri dishes with two 4 :_0.05 gram food lumps in each dish. Within
any one dish the food lumps were placed on opposite sides of the dish

so that test larvae could be introduced between them and allowed to

make a choice.

Experimental Procedures

The experimental procedures for all experiments in this paper
arerthe same as described for the experiments in Chapter IV, with
respect to how food is prepared and put in the experimental dishes,
Inna conditioning is done and how test larvae are introduced and data
(Hallected. The only difference between the two procedures is that in
these experiments two test larvae are added to each dish rather than
one. These test larvae are of varying ages depending on the test
sijniation. Both larvae were placed at the center of the dish and

allowed to choose a food lump. Data was then collected on the

205

distribution of both larvae over time, although, in some situations,
only the response of one of the larvae was monitored.

There are several different experimental designs in this
paper and each will be described at the proper time. However, there
are some nomenclature differences from Chapter IV. Only two different
degrees of conditioning are utilized in these experiments, namely
CAT=1,12,(7,11,13) and CAT=7,6,(7,ll,13) which will be referred to as
LOW and HIGH conditioning respectively, where C = days of conditioning,
A = age of conditioning larvae, and T = age of test larvae.

Each experimental group has been assigned an identification
code for ease in referring to the various comparisons I will be making.
For example, codes such as SOP(l-3)7+7 or 64PH(l-S)7+13 will be fre-
quently encountered. The first two numbers (50 or 64) are identifi-
cation numbers for each experimental group and the rest of the nomen-
clature conveys the experiment that was performed. The P stands for
"pairs" of larvae, meaning that two test larvae were used in the
experiment, whereas an I stands for "isolate" in experiments when a
single test larva was utilized (i.e., 50P(l-3)7+7 or 591(l)7).

The letter following the P or the I will be either an L or
sun H referring to "low" and "high" degrees of conditioning. If
:neither an H or I appears in the identification number this means that
when the experimental group was tested, at least initially, neither of
the food lumps had been previously conditioned (50P(l—3)7+7). The
two numbers on the end of the identification number (7+7 or 7+13)

iJuiicate the ages of the test larvae at the start of the experiment.

206

The numbers in parentheses refer to how the data for each
experimental group is being analyzed. When two larvae are used in
the experiments there are five possible ways in which to analyze the
data:

1. total number of organisms on the conditioned food lump,
regardless of whether paired or single;

2. total number of pairs on the conditioned food lump only;

3. total number of pairs regardless of the food lump on which
they occur;

4. the number of older larvae of the pair of test larvae on the
conditioned food lump, i.e., response of 13 day old larva when
a 7 day old larva is also present;

5. the number of younger larvae of the pair of test larvae on
the conditioned food lump, i.e., response of 7 day old larvae
when a 13 day old larva is also present.

Larvae used in these experiments were not marked for identi-
fication purposes. Because I used such large sample sizes, marking
them would have been prohibitive, and I would have had to mark them
several times during the course of the experiment since they shed
their cuticles several times throughout development. This would have
:necessitated disturbing the experimental procedures to remark them.
'Therefore, when two test larvae are the same age (i.e., 7+7, ll+ll,
or 13»13), individuals cannot be identified. Because of the size
(distribution of larvae, however, if two larvae are of different ages
(i.e., 7+ll, 7+l3, or ll+13) individuals can easily be identified

for analysis. Therefore, when both larvae are the same age, only the

207

first three of the above analyses are possible, whereas when the
larvae are different ages all five of the analyses are possible.

In one experiment in this paper, neither food lump was pre-
conditioned at the start of the experiment. In such a case the five
possible analyses do not refer to the conditioned lump. However, one
of the food lumps received an identification mark so that it could be
distinguished from the other lump, making the five analyses possible
as well. The following are a few examples of how to interpret the
identification numbers in this paper:

SOP(l-3)7+7: Experimental group #50, with a pair of test
larvae with no initial conditioning, and in
which the ages of the two larvae were 7+7.
The possible analyses are (1—3) since both
larvae are the same age.

6OI(1)lS : Experimental group #60 with an isolate test
larva of 13 days of age. When an isolate is
used only the first analysis is ever utilized

(1).

S7PL(4)7+13: Experimental group #57 with a pair of test
larvae tested with Low conditioning. The
larvae are 7 and 13 days old and the analysis

is on the 13 day old larva in conditioned
food (4).

64PH(5)7+13: Experimental group #64 with a pair of test
larvae that are 7 and 13 days old and High
conditioning was employed. Analysis is on the
response of the 7 day old larva to conditioned
food (5).

68PH(S)7+1§; Experimental group #68 with a pair of test
larvae that are 7 and 13 days old and High
conditioning was employEd. The fact that the
l§_is underlined indicates that this was a
resident larva in the conditioned food and not
introduced when the 7 day old test larva was.
If neither age is underlined then both larvae
were introduced together. Analysis is on the

response of the 7 day old larva to conditioned
food (5).

208

When two test larvae are of different ages, the length of the
mean response vectors is dictated by the age of the older larva, the
experiments being terminated when the older larva reaches its wander-
ing phase. For example, if the two larvae are 7+13 days of age, only
four experimental days will be analyzed. After 4 days the 13 day old
larva enters the wandering phase. The 7 day old larva will not enter
the wandering phase until 5 days after the 13 day old larva. However,
I was interested only in animal-animal interactions during the "larval
phase." Therefore, when the older larva enters the wandering phase,
data analysis ends. There is one exception to this procedure which

will be noted at the proper time.

Analysis

The dependent variable chosen for analysis within each experi-
mental group is the mean response to conditioned food over time for
pairs or either member of the pair of test larvae. When conditioned
food was not employed, the dependent variable becomes the number of
pairs, regardless of food lump, over time.

Each experimental group has 80 dishes and, in most cases, each
«dish has two test larvae. If both larvae are in the conditioned lump
:1 score of 2 is recorded, if neither is in that lump a O is recorded.
If only one larva is in the conditioned food a l is recorded, and in
the case of different aged test larvae the l is identified as to
whixfli larva it represents. It was desirable to be able to compare two-
larvae situations with isolate situations. To do this the sample

sizes had to be adjusted. The following procedure was used.

 

209

If, for a two-larval test situation, it was desirable to
analyze the number of pairs or the total number of larvae on a
particular food lump in all 80 dishes over time, then the matrices
would have entries of all two's or zeros for the paired situation and
two's, one's, and zeros for total larvae on a particular food lump.
These matrices were divided by the number of test larvae per dish,
in this case 2, which reduces the entries to either one's, 0.5'5,
or zeros. These matrices are now comparable to matrices generated
from only one test larva and the sample size is always 80 (the number
of dishes) and not 160 and 80. If, on the other hand, the response
of only one of the two test larvae is analyzed, the entries in its
data matrix are already one's and zeros and need not be divided by 2.

Statistical analysis is by means of Hotelling's One- and Two—
Sample procedures, as discussed in Chapters III and IV, although some
Chi-square and ANOVA analyses are employed. All hypotheses are two-
tailed and it was a_priori determined to use an a-level of .025 because
of heterogeneous variances, positive serial correlation, and multiple

comparisons. These considerations were all discussed in Chapter IV.

Results
The following description of the results has been divided into
tflrree experiments, all of which relate to whether individual preference
ier biological conditioning is altered by the presence of a conspecific.
Experiment I: This experiment examines the question of whether,
in the absence of initial conditioning, Galleria
larvae aggregate and whether their distribution

is a function of the age of the other member of
the pair.

210

Experiment 11: This experiment examines individual preferences
for a conditioned food lump when a conspecific
is also present and asks whether preference is
a function of the conspecific, the age of the
conspecific, or whether the conspecific is a
resident in the conditioned food.

Experiment III: Experiment 11 will demonstrate that when 7+13 day
old larvae are tested with high conditioning the
7 day old larva's response to conditioning is
significantly lowered initially but increases
with time. Experiment III deals with several
test relating to whether this behavior is related

to the behavior of the 13 day old larva, the
conditioning, or an interaction between the two.

Experiment I: Do Galleria Larvae Group
in the Absence of Initial Conditioning
and Is their Distribution in Such a
Situation a Function of the Presence
of a Conspecific?

Before asking whether larval preferences for conditioning are
affected by animal-animal interactions, it was decided to determine
if larval-larval interactions affect their distribution when neither
food lump is previously conditioned. The only valid measure to answer
this question is the initial preference (5 minutes after the start of
the experiment) of the larvae for being with a conspecific of varying
ages or of remaining isolated. The larvae are constantly conditioning
their environment so that very shortly after the experiment begins the
food becomes conditioned and it would no longer be possible to deter-
lnine if response is to the conditioning or the other larva. However,
tn! looking at certain facets of the whole experiment, we can get some
.Ldea.of which variable may be responsible for the resultant distrib—
utions.

The experimental design is shown in Figure 5.1. The six

experimental groups are shown along with replications (R) of those

AGE OF TEST LARVAE

AGE OF TEST LARVAE

211

 

7

11

13

 

SOP(1-3)7+7

9/11/74

50R(l—3)7+7
4/17/75

SIPCI15)7+11
9/11/74

SlR(1-5)7+11
4/17/75

52P(1-S]7+13
9/11/74

52R(1-S)7+13
4/17/75

 

539(1—3)11+11

549(1—5)11+13

 

 

 

 

 

 

9/11/74 9/11/74
11
53R(l-3)ll+ll S4R(l-5)ll+13
4/17/75 4/17/75
55P(1-3)13+l3
9/11/74
13
SSR(l-3)l3+l3
4/17/75
Figure 5.1

Experimental design for asking whether larvae will

group (pair) in the absence of homotypic conditioning.

 

212

groups and the dates when each was run. All experimental groups were
run at the same time and all replications were run together, but
several months later. Each identification code indicates the larval
combinations and the possible analysis and each group has a sample
size of 80.

The question being asked is whether larvae will group (pair)
in the absence of initial conditioning and when the age combinations
of pairs is varied. The analyses utilized are in relation to pairing
regardless of food lump, since both food lumps were "equal" at the
start of the experiment.

Table 5.1 shows the results of comparing each experimental
group with its replication by both the One- and Two-Sample procedures.
The results indicate that, although variances are heterogeneous,
pairing behavior in the absence of initial conditioning does not
change from population to population of larvae.

Figure 5.2 (A+B) shows the mean response vectors for percent
pairs of larvae on both food lumps when the food lumps were initially
unconditioned. The only valid comparison for whether larvae prefer
to pair in the absence of conditioning is for the first points on
these curves. The results of this comparison may be found as Chi-
square values in the first boxes of each comparison in Table 5.5.
'This table will be explained shortly, but for now the important point
is that regardless of age combinations, pairs of larvae initially dis-
'tribute themselves randomly. This indicates that, at least initially,

the larvae are not reacting to each other. If such reactions were

213

 

 

 

 

.mo. u a 3m ©3.~ u 3cm.ovaa ecu mm.3 u 30w.oe3o .mmo. u a pm 3m.N u 3mm3.ov~a. use mm.~ u 3vn.oV3a.
.mo. n a 3m w~.N «3om.O3U=a an“ 35.3 "30w.O3U3n .mwo. u a pa m3.~ n3me3.03vma cam «N.N "3on.03e3m
$3. 3m. «8. SN. 33. SN. #3338? L
mo3. om3. New. Now. m3m. NmN. mmmm
: mom. m3m. mom. ONN. M33. mmm. ammm 33+33mmwmmm
3cm.3 Ame. Ame. mNA. mmn. Awe. New. «emu 33+333mvamm
A3.N ao.o wN.3 N35. con. man. Noe. con. mmm. ammm
NV3. aw3. ~3N. 0mm. omm. mew. eo3oomm
em3. N03. “om. 0mm. sew. mew. m3mm
: @33. n33. @33. 0mm. 3mm. New. a3mm 33+Ammwm3m
330.3 ~3w. cow. N35. cmo. mom. New. 13mm 33+A3mva3m
3mw.3 mm.o mm.3 0mm. cos. “we. owe. omm. mum. a3mm
wmm. mNN. "m3. ONN. mam. new. omN. mmN. mmm. omN. eo3oomm
emm. oem. m3m. ovw. new. new. mmm. mmm. mmm. mmu. momm
m mam. NNN. eww. eww. mam. new. mmN. mmm. mmm. new. aomm A+Ammwmom
.om.3 omo. Ame. N30. ewe. ~30. mhm. mam. ~3m. mNm. N3m. «emu n+53mvaom
.m¢.3 Nw.o 33.3 5mm. who. ene. Ame. Awm. mum. Ame. mne. mew. mam. aomm
ones a «a 3a 03 m m A o m e m m 3 masons
m.xom mama cw osfie

 

.mnfiom mo hopes: Hope» we mfimon one so owes one meanwhomaou

.cfi ouo: xonp mesa poem coax: mo mmofiuhowon

.3.m onsmfim scum anm:0wuoUAHmoh

oz» msmuo> Amummsonw Houcoafinomxo mo mcomwuomaou oagaom mawdze van n.uo:o m.m:«33ouo: mo muHSmoz

1m 93.3.

214

A

 

 

 

 

.mo. u a 3m 03.3 n so was mm.3 u a
.mNO. n a we mm.~ «3mm3.vVNal ecu om.~u3oe.v33a.
mam. mmm. mmm. mmm. eu3ooam
mew. 3mm. nmm. mmN. mmmm
.: mmm. mmm. mmm. mmm. ammo m3+m3mmwmmm
3m.m Awm. omm. mmm. mmm. ammm M3+m33mVamm
om.~ oo.o me.~ com. com. 5mm. oom. ammm
w3N. mmN. ANN. mmm. em3ooam
om3. ONN. w3N. mmN. memm
: mam. mam. 5mm. mmm. avmm m3+33mmwmem
3m.N owe. Noe. Ame. ~3m. mvmm m3+333mvaem
mm.~ NN.3 en.~ 5mm. New. mNo. Awe. new“
mMN. mmm. meN. mmm. eo3ooam
m3m. GNN. 5mm. mmm. ammm
: 3mm. 3mm. mmN. mmm. ammm M3+Ammwmmm
A3.N “me. New. mmo. N3m. mmmu m3+n3mvawm
m3.~ em.o 3o.N 5mm. 0mm. N3m. mew. ammm
amok. a N a 3 m e n N H mmzopu
m Fxob who: :3 anr

 

3.e.3coov 3.m 033me

215

taking place we would expect either significant attraction or repul-
sion, when in fact, a random distribution is exhibited.

If we examine Figure 5.2 (A G B), however, there is a trend
for increased pairing with time. Table 5.2 shows the results of
Hotelling's One-Sample comparison of each of these preference curves
with an expected vector whose entries are all 0.5. The F-values show
that in only three groups does aggregation become significantly
different from random over time, namely in the 7+7, 7+ll, and ll+ll
situations. Whenever a 13 day old larva was a member of the pair
grouping never reaches significance. This may be an indication that
some sort of animal-animal interaction occurs over time. However,
the preference vectors in Figure 5.2 and compared in Table 5.2 are
unequal in time. Therefore, an analysis was performed using the
first four days and the last four days of the experimental periods,
the two being the same for 13 day old response vectors. The results
of this analysis are presented in Table 5.3. It can be seen that
the same groups show significant pairing except that the 7+7 group
does not show significant pairing until the end of the experimental
period, that is until the larvae have been in the experimental situ-
ation for at least four days.

One might contend that these differences are a function of
time in the experimental situation. That is, 13 day old larvae do
not show grouping behavior because they only have four days in the
experimental apparatus, whereas a 7 day old larva has 10 days, and an
11 day old larva has 6 days. This may be true of the 7+7 situation,

but it will be noted that 7+ll and 11+ll groups significantly pair

216

Figure 5.2

Mean response vectors for percent pairs of larvae on any food lump
when the food lumps were initially unconditioned.

A. Pairs composed of larvae of the same age.

B. Pairs composed of larvae of different ages.

217

A.m m3:m3a

m <

 

 

 

 

 

 

whom a“ oefib mxoo :fi oeflb
03 m w A o m e m A 03 m m A o m e m A 3
Dr I D D D P P P I D O P b P r P P P P DI L l O
H H
2.333333 3. x 4.0... 2.725333 x a 3.0m
m3+AEANm r.....l 33+33$Emm 3..-...
33+Afimva3m cuuull. A+A3mvaom .unnun.
cow 0/0 IO?
.d
9
T...
.om m .om
O
u
.oo .m .00
.d
m
.OA 9. .0A
A...
m
.ow .ow
.om .om
.o03 30o3

 

dmnq pood fine no sated %

218

Table 5.2

Results of Hotelling's One Sample comparison of each ex-
perimental (P) group in Figure 5.1 with an expected vec-
tor whose entries are all 0.5 to see if grouping (pair-
ing), regardless of food lump, is different from ran-
dom in the absence of homotypic conditioning. Response
mean vectors(X), estimated variance vectors(Sz), and D-
values for serial correlation are found in Table 5.1.

 

 

 

Groups F

509(3)7+7 2.81**

519(3)7+11 13.32***

529(3)7+13 0.57

539(3)11+11 4.19**

54P(3)ll+13 2.01

559(3313+13 0.20
**F(10,70)=2.59, F(6,74)=3.08, F(4,76)=3.59 at P=.Ol.
***F(10,70)=3.48, F(6,74)=4.26, F(4,76)=5.18 at P=.001.

219

Table 5.3

Results of Hotelling's One Sample comparison of pairs, re-
gardless of food lump, versus an expected vector whose en-
tries are all 0.5. These comparisons are shown for the
first four days and the last four days of each experimental
group to see if time in the experimental situation affects
grouping behavior.

 

 

 

First 4 Days Last 4 Days
Groups F F
SOP(3)7+7 1.07 2.99*
51P(3)7+ll 3.53* 19.64***
52P(3)7+13 0.57 0.57
53P(3)ll+ll S.83*** 6.02***
S4P(3)ll+l3 2.01 2.01
55P(3)13+13 0.20 0.20
*
F(4,76) = 2.96 at P - .025.
***
= 5.18 at P = .001.

IV4.76)

220

when just the first four days are considered which is the same length
of time 13 day old larvae have.

In Table 5.4 are presented the results of Hotelling's One- and
Two-Sample comparisons for all possible comparisons of the experi-
mental groups in Figure 5.1 for the first and last four days of the
experimental periods. The F2-values show only one significant dif—
ference for the first four days, namely that ll+ll pairs are more
frequently found than are 7+7 pairs. For the last four days, however,
there are many more differences, the greatest pairing occurring when
an 11 day old larva is a member of the pair, and the greatest differ-
ence is the 7+ll versus the l3+l3 situation. Again illustrated is
the fact that pairing is not significant when at least one member of
the pair is a 13 day old larva. The Fl-values in Table 5.4 point out
some other differences, probably as a result of the heterogeneous
variances in the Two-Sample case.

There are four possible ways to explain why aggregation may be
occurring over time: (1) there may have been discrepancies between the
two plain food lumps due to handling or nutritional differences,

(2) the larvae may be responding to conditioning of other larvae,
(3) some sort of larval-larval interaction may develop, and (4) com-
binations of 1-3 are possible.

We have already seen that initial preference for the food
lumps is random, indication that the food lumps were equal in all
respects. However, if the food lumps are actually not different, then
pairing with respect to one food lump should be random, since the

larval distribution should not be a function of food. To determine

221

Table 5.4

Results of Hotelling's One(') and Two(*) Sample comparisons among the
experimental groups in Figure 5.1. Comparisons are made for the first

4 days and the last 4 days to see if time in the experimental situation
affects grouping, and all comparisons are for pairs of larvae regard-
less of food lump. For the One Sample(') Test the groups used for the
expected vectors are in the column headed BV. and those whose covariance
matrix was used are in the column headed CV.

 

FIRST FOUR DAYS

 

 

 

comparisons Box's T2 Test for
EV. CV. F1 F2 Test Parallelism
50P(3)7+7 vs. 51P(3)7+ll 3.76" 1.94 H P
52P(#)7+13 vs. 50P(3)7+7 2.65 1.14 H P
50P(3)7+7 vs. 53P(3)ll+ll 3.67** H
50P(3)7+7 vs. S4P(3)ll+13 4.17" 2.22 H P
50P(3)7+7 vs. 55P(3)13+13 0.69 P
52P(3)7+13 vs. SlP(3)7+ll 2.71 1.42 H P
51P(3)7+ll vs. 53P(3)ll+ll 1.66 P
51P(3)7+ll vs. S4P(3)ll+13 2.23 P
55P(3)13+l3 vs. SlP(3)7+ll 4.06" 2.09 H P
52P(3)7+13 vs. 53P(3)ll+ll 5.34"' 2.23 H P
52P(3)7+l3 vs. 54P(3)ll+l3 1.33 P
52P(3)7+l3 vs. 55P(3)13+l3 0.66 0.45 H P
54P(3)11+13 vs. 53P(3)11+11 1.95 0.84 H P
53P(3)ll+ll vs. 55P(3)13+l3 2.44 P
SSP(3)13+13 vs. S4P(3)ll+13 1.05 0.49 H P
' F1(4,76)= 2.96 and * F2(4,155)= 2.88 at P = .025.
v. **
F1(4,76)= 3.59 and F2(4’155)= 3.47 at P = .01.
111 ***
F1(4,76)=‘5’18 and F2(4,155)= 4.94 at P = .001.

HHeterogeneous variances (P 5.05).

PMean response vectors are parallel.

222

Table 5.4 (cont'd.)

 

LAST FOUR DAYS

 

 

 

comparisons Box's T2 Test for
EV. CV. F1 F2 Test Parallelism

50P(3)7+7 vs. SlP(3)7+11 5.97*** H
52P(3)7+13 vs. 50P(3)7+7 1.23 P

50P(3)7+7 vs. 53P(3)ll+ll 4.47" 1.50 H P

50P(3)7+7 vs. S4P(3)1l+13 2.07 P
55P(3)l3+l3 vs. 50P(3)7+7 3.41" 1.86 H P
52P(3)7+13 vs. 51P(3)7+ll 5.93*** H P
SlP(3)7+ll vs. 53P(3)ll+ll 2.91* H P
SlP(3)7+1l vs. 54P(3)11+13 5.50*** H
55P(3)13+13 vs. 51P(3)7+ll 7.87*** H
52P(3)7+13 vs. 53P(3)ll+ll 4.59" 2.34 H P
52P(3)7+13 vs. 54P(3)ll+13 1.33 P
55P(3)l3+13 vs. 52P(3)7+13 0.66 0.45 H P
53P(3)ll+ll vs. 54P(3)ll+13 3.10' 1.39 H P
55P(3)13+13 vs. 53P(3)ll+1l 4.82" 2.16 H P
55P(3)13+13 vs. S4P(3)ll+13 1.05 0.49 H P
I *

F1(4,76)= 2.96 and F2(4,155)= 2.88 at P = .025.
c. **

F1(4,76)= 3.59 and F2(4,155)= 3.47 at P = .01.
"'F1(4,76)= 5.18 and ***F2(4,155)= 4.94 at p = .001.

HHeterogeneous variances (PfE.05).

PMean response vectors are parallel.

223

if larvae are pairing with respect to only food several contingency
tables with Chi-square analyses were set up in which pairs on one

food lump are compared with pairs on the other lump at each time point
as well as for the total experimental period. This enables parcelling
out the contribution of each time point in respect to pairing. If
significance at any one or more time points results or if the Chi-
square values are significant then the indication is that the two food
lumps were different, even initially. This analysis is presented in
Table 5.5.

It can be seen from these results that there are no signifi-
cant Chi-square values either in each box or the totals. It may be
concluded from this that pairing over time is not due to any dis-
crepancies between food lumps. However, once the larvae select a
food lump, they begin conditioning it and this cOnditioning may
attract the other larva and thus pairing increases. For example, in a
7+l3 situation the 13 day old larva may be doing more conditioning
than the 7 and this attracts the 7 to where the 13 is located. We
have already seen that whenever a 13 day old larva is present no
significant pairing occurs (Tables 5.3 and 5.4) and examination of
Figure 5.2 shows the response curve to be random. However, the 7+7,
ll+ll, and 7+11 situations show significant pairing with time. In the
7+7 and ll+ll situations we would expect each larva to be doing an
equal amount of conditioning and thus excess conditioning by one
larva which attracts the other larva is probably not the mechanism.

It was also noted (Tables 5.3 and 5.4) that 7+7 day old larvae take

longer to exhibit pairing than do the 7+11 or ll+ll situations and

 

 

224

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.35. u 5.33 35.33 u 3533 p53 Ao.33 u 3333 .mo. - A 33 35.3 n 3333
N N a. N .33
3mAm.o 30505.5 om3o.o eoNo.o noooo.o 3535.5 Nm3m.o
NNm Am em mm mm em Ne
Nmm3.o moooo.o NO3o.o GO3o.o oooo.o No3o.o N5N3.o
333 .mm. NN mN 3N NN on, N 33553
ANNN.o mooo.o 5500.5 3500.5 moooo.o emoo.o omm3.o 3 5553 33.33
533 5N NN 5N AN NN MN
33 m3 53 33 N3 33
NONN.3 eAeN.c mooo.o NNNN.o memo.o 5033.5 5350.5 3A3o.o AAmo.o 3503.0 mNmN.o
mme A5 em 3m 3m A5 35 mm om mm ow
Ao3m.o 3350.5 3ooo.o 5303.5 OONo.o 53oo.o mANo.o oooo.o OONo.o A3vo.o 5533.5 N 5553
ANN 5N em em mm 33 on mN NN mN AN
\+N.
3AeA.o mem3.o ONoo.o NAA3.o memo.o omo3.o OAmo.o 3330.5 AAmo.o 5055.01 @533.o 3 5553 r
553 53 ON A3 N3 e3 33 33 53 m3 m3
533 53 N3 A3 03 m3 33 N3 N3 33

one xen Aeoo ennunz

 

 

 

 

 

 

 

 

 

 

 

.m9553 poem anon pew mannem oEnn HHe no>e o=35> onoevmunno Hence osn pee compennnu
-ceo onoSGmunnu nmonu pee e353 poem ne venom oanu Aeoo no unwed Honey one one arenm em3<
-nsu.3ospn>np:n nnonu pee venom oEHp sumo no @533 poem neoo nem oo>noH me mnnom me nonase “open one azegm

.mo33o> onoacm

.Houeu one on 93:3 poem nooo nem venom oanu sooo me :enusnnnueeo oap seam ea ponenu

ununom onoecmungo 3epeu one can: 3.m onsunm an moon» :eoo nem mnmnnoe< onoadmnnnu nuns moneoe noaownueeu
m.m 023,—.

225

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 5.5 (cont'd.)
t1 12 t3 t4 t5 t6
Lum 23 21 24 27fi’ 27: 32 154
7+11 P 0.0400 0.0004 0.0359 0.0155 .0007 0.0117 0.1042
Lum 23 23 28 28 29 :36 167'
P 0.0361 0.0005 0.0333 0.0130 .0005 0.0112 0.0947
46 44 52 55 56 68 321
0.0761 0.0009 0.0692 0.0285 .0013 0.0229 0.1989
t1 t2 t3 t4
Lump 20 19 15 15 69
13+13 0.5566 0.3620 0.2200 0.2200 1.3586
L 20 24 25 25 94
“PP 0.4060 0.0157 0.1632 0.1632 0.7481
40 43 40 40 163"'
0.9626 0.3777 0.3832 0.3832 2.1067
t1 t2 t3 t4
Lum 20 19 *23' 21 83
7+13 P 0.0526 0.1097 0.0454 0.0116 0.2229
Lum 18 22 21 22 83
P 0.0526 0.1097 0.0454 0.0116 0.2229
38 41 44 43 166
0.1052 0.2194 0.0908 0.0232 0.4458
t1 t2 t3 t4
Lump 12 18 16 16 62
11+13 0.1365 0.0452 0.0225 0.0005 0.2047
Lum 27 32 29 31 119
P 0.0721 0.0230 0.0113 0.0003 0.1067
39‘ 50 4s 47 181
0.2086 0.0682 0.0338 0.0008 0.3114
*x2 - 3 84 x2 = 7 81 and x2 = 11 07 at p = 05
(1) ° ’ (3) ° ’ (5) ° ' °

226

that the 7+ll group exhibits the greatest amount of pairing. Since
initial preference in relation to food lump is random, we would not
expect Table 5.5 to shed any light on why pairing is occurring over
time, even if the larvae are changing those food lumps and making
them more "attractive" to the other member of the pair. However, we
can approach the question of whether something one larva did to one
food lump is attracting the other larva in another way.

Let us hypothesize that the older larvae do more conditioning
than the younger larvae over time and that the younger larvae are
attracted to that conditioning. We can test this hypothesis using
the 7+ll, ll+l3, and 7+l3 situations. In these groups every dish in
which the older larva initially selected one of the plain food lumps
and remained there for the whole experiment was selected for analysis.
There were 60 such dishes in the 7+ll group, 68 in the 7+l3 group, and
65 in the ll+l3 group. Those food lumps are now designated as the
conditioned lump and we can ask how often the younger larva is attracted
to that lump, either because of the presence of the other larva or
the conditioning. The results of this analysis are in Table 5.6 and
graphed in Figure 5.3.

The results demonstrate that the 7 day old larva is homing in
on the lump with the 11 day old larva in it, but that neither a 7 or
11 day old larva are attracted to the lump with the 13 day old larva
in it. It will also be noted that the response vectors in Figure 5.3
are very similar to those in Figure 5.2 B in which pairs over time

were initially graphed.

227

Table 5.6

Results of Hotelling's One-Sample analysis on the preference of the
younger larva of a pair for the food lump in which the older larva
initially selected and remained in for the entire experimental period.
Comparison is made against an expected vector whose entries are all 0.5.

 

 

 

Group Fl
54P(5)11+13 2.10
52P(5)7+13 2.26
5.P(5)7+ll 10.27"
' = = = :
F1(4,65) 3.00, F1(4,68) 2.99, and F1(4,6l) 2.63 at P .025.
H = :-
F1(6,6l) 4.37 at P .001.

Of the 12 dishes in group 52P(S)7+l3 in which the 13 did not
initially select and remain in a food lump, in 5 dishes the 7 moved
away when the 13 moved to its lump and in 3 dishes the 13 moved away
when the 7 moved to its lump. In group 54P(5)1l+13, of the 19 dishes
in which the 11 did not stay in its initial choice, the 7 moved away
3 times when the 11 moved to its lump and the 11 moved away 4 times
when a 7 moved to it. In group 54P(5)ll+13 of the 15 dishes in
which the 13 did not choose and stay in a lump, in 5 dishes the 11
moved away when the 13 moved to its lump and in 4 dishes the 13 moved
away when the 11 moved to its lump.

From the results of Experiment I it is possible to develop
the following hypothesis. Only pairs 7+7, 11+11, and 7+ll are sig-
nificant over time and whenever a 13 day old larva is one of the
test larvae, pairing does not occur, regardless of the age of the

other larva. In the cases where significant grouping is found it is

1001
90-
80J
70‘
60~
sol
404

30‘

% Younger Larvae of Each Pair in "Cond." Food

 

228

3....
I,”

v—-—-** 51P(S)7+11
X—fl SZP (5)7+l3
e... ..... 54P(5)11+13

 

V r U I I wt r V T V j T V

3 4 5 6 8 9 10 ll 12 13 14

Time in Days

Figure 5.3

Mean response vectors of the younger larvae in a pair to a
particular "plain" feed lump that was initially selected by
the older member of the pair and in which the older larva
remained in for the entire experimental period. That food
lump is considered the "conditioned" lump for purposes of
these curves.

229

not because of some difference between the two food lumps, rather it
appears to be due to the presence of another larva or to something
that larva does to the food. When it was possible to distinguish
individuals, it was found that 7 day old larvae would move to where
an 11 day old larva was and stay there. There are two possibilities
for this occurrence: either the younger larva prefers to be with the
other larva, or the other larva, being older and bigger, is more highly
conditioning its food lump than is the 7 day old larva and this
attracts the 7. However, there would also appear to be an animal-
animal interaction in this behavior since neither a 7 or 11 is so
attracted to where a 13 day old larva's lump is the the 13 should be
doing the most conditioning. It is possible that the 13 day old
larva is doing a different kind of conditioning that younger larvae do
not find as attractive. However, the response of 7 and 11 day old
isolates, Chapter IV, to similar degrees of conditioning are very
strong, indicating there is not a qualitative conditioning difference.
In situations where both larvae are the same age, 7+7, ll+1l,
and l3+l3 it is not possible to definitively determine what is going
on. The initial 7+7 and ll+ll situations exhibit significant grouping
with time, the l3+l3 does not. It would appear that a 13 day old
larva will not tolerate the presence of any other larva, but younger
larvae will. However, both 7 day old larvae should be equally con-
ditioning their environments as should both 11 day old larvae. One
possible explanation for grouping in these situations is on the basis
of animal-animal interactions not conditioning. That is, the presence

of older larvae overrides the preference for conditioning by younger

230

larvae, whereas the presence of a younger larva does not. We will
come back to these points in the discussion section.

To be sure that any one food lump was sufficient for develop-
ment of two larvae of any age and that the distributions observed in
my experiments are not due to food limitation, the following was done.
One hundred and twenty dishes were set up with one food lump in each
dish. In 60 of these dishes the lump was highly conditioned (the
maximum conditioning of any experiment in this paper) and in 60
dishes it was plain. Two 7 day old larvae, two 11 day old larvae,
and two 13 day old larvae were raised in 20 dishes in each situation
and their pupation time measured. If the developmental time of these
larvae is not different among the groups then food was not a limiting
factor in any of the experiments presented in this paper. The
results in Table 5.7 show that development was not different from

group to group.

Table 5.7

Analysis of Variance on developmental times of two 7, two 11, and two
13 day old larvae reared on single conditioned or plain food lumps.

 

 

 

 

 

 

Source of Variation d.f. S.S. M.S. F ratio
Between Group 5 3.5416 0.7083 0.9723
Within Group 114 83.05 0.7285
Total 119 86.5916

 

 

 

 

 

F(S,119) = 2.2914 at P = .05.

231

Experiment 11: When Initial Condition-
ipgfis Present, is Preference for Con-
ditioning Affected by the Presence of
a Conspecific, the Age of the Con-
specific, and Whether the Con-
specific is a Resident in the
Conditioned Food?

 

 

 

 

 

 

 

These experiments examine whether the preference of a larva
for conditioning is affected by the presence of a conspecific when
that conspecific is an established resident in the conditioned lump
or when it is introduced with the other larva. The experimental
designs for these experiments are shown in Figures 5.4 and 5.5. Each
experimental group has a sample size of 80 and the identification
numbers were explained earlier.

Only two ages of test larvae are used, 7+l3 days old, and
only two degrees of conditioning, high and low. Isolate larvae, of
all ages, are attracted to both degrees of conditioning (see
Chapter IV), but at a lower level to the low conditioning than to the
high conditioning. These two degrees of conditioning, namely LOW:
l,12,(7,l3) and HIGH=7,6,(7,13) were chosen because the conditioning
larva is 13 days old at the end of the conditioning period in both
groups and may then be left in the conditioned food as a resident
larva for some of the experiments. To maximize the probability of
finding animal-animal interactions only 7 and 13 day old larvae are
utilized throughout the remainder of these studies.

The experimental procedure for Figure 5.4 is as for all
previous experiments. When the conditioning period is over, for each
degree of conditioning, the conditioning larvae are removed and a pair

of test larvae added, the pairs being either 7+7, 7+l3, or 13+13. In

232

 

LOW CONDITIONING

HIGH CONDITIONING

 

Age of Test Larva

Age of Test Larva

 

 

 

 

 

 

 

 

 

 

 

 

7 13 7 13
56PL(1-3)7+7 57PL(1-5)7+13 63PH(1-3)7+7 64PH(l-S)7+13
o
E 7
.4 2/5/75 2/5/75 2/5/75 2/5/75
‘2
Q)
n-._
44
° 58PL(l-3)13+l3 65PH(1-3)13+l3
3°
'13
2/5/75 2/5/75
591L(1)7 6OIL(1)13 66IH(1)7 67IH(1)13
2/5/75 2/5/75 2/5/75 2/5/75
Figure 5.4

Experimental design (top) for animal-animal interactions, at two
levels of conditioning, when the conditioning larva is removed
and both test larvae are new to the experimental situation.

 

LOW COND.

HIGH COND.

 

,Age of Test Larva

 

7

13

 

2/5/75

61PL(1-S)7fl§.

2/5/75

68PH(1-5)7+l§

 

13

Age of Test Larva

2/5/75

 

 

 

62PL(1-3)13f§§

2/5/75

 

69PH(1-3)13+l§_

larva.

 

J

 

Figure 5.5

EXperimental design (left)
for the effect of a resi-
dent larva on the prefer-
ence of a test larva for
biological conditioning.
The resident larva is the
original conditioning

233

Figure 5.5, however, at the end of the conditioning period the con-
ditioning larva is left in the conditioned lump and a single 7 or 13
day old larva added. Both experiments utilize two test larvae, but
in one, both are new to the situation and in the other, one is a
resident (always a 13 day old larva) and the other is new to the
situation (either a 7 or 13 day old larva).

Table 5.8 lists each group from Figures 5.4 and 5.5 which have
been analyzed, their mean response (preference), their estimated
variance (82), and the D-values for serial correlation. It must be
remembered that there are several ways in which to analyze any one
experimental group based on (1) total larvae in the conditioned lump,
(2) pairs in the conditioned lump, (3) pairs regardless of lump,

(4) only the 13 day old larva in the conditioned lump and (5) only
the 7 day old larva in the conditioned lump. The groups are listed
as such in Table 5.8. However, not all of the comparisons were made
because they were not necessary to answer the questions being asked
and some comparisons would be meaningless.

To find out if larvae in pairs are attracted to low and high
conditioning, each group was compared to an expected vector whose
entries are all 0.5. This analysis used the total number of larvae
on conditioned food to determine if two larvae distribute themselves
randomly or are attracted to the conditioned food.

The results in Table 5.9 demonstrate that, regardless of
degree of conditioning or age combination of larvae, all groups are
significantly different from random and are attracted to conditioned

food as can be seen by examining their mean response vectors in

234

Table 5.8
This table presents the response mean vectors(X), estimated variance vec-
tors(82), and D-values for serial correlation for the groups in Figures

5.3 and 5.4 to be used in future cppparisons.

 

 

LEW CONDITIONING

 

PTime in Days

 

 

Groups 1 2 3 4 S 6 7 8 9 10 D
’x .637 .643 .650 .643 .643 .712 .781 .743:.781 .681 1.46*
56PL(1)7*7 82 .120 .134 .142 .134 .147 .119 .113 .139 .132 .115
x .412 .450 .475 .450 .475 .537 .662 .637 .575 .375 1.46*
56PL(2)7*7 52 .245 .251 .253 .251 .253 .252 .226 .234 .247 .237
i .562 .675 .700 .681 2.20
57PL(1)7*13 82 .123 .115 .099 .109
ssz(1)7+13 x .562 .675 .700 .681 .718 .612 .531 .443 .431 .375 1.60*
Wandering 52 .123 .115 .099 .109 .119 .133 .091 .076 .068 .092
x .312 .450 .475 .462 2.20
57PLC2)7*13 52 .218 .251 .253 .252
X .650 .725 .775 .762 2.11
57PLC4)7+13 52 .230 .202 .177 .183
X .475 .625 .625 .600 .687 .750 .775 .750 .750 .625 1.40*
57PL(5)7*13 82 .253 .237 .237 .243 .218 .190 .177 .190 .190 .237
x .837 .856 .856 .825 2.39
58PL(1)13+13 52 .075 .071 .058 .077
x .737 .750 .725 .687 2.25
58PL(2)13+13 s2 .196 .190 .202 .218
591L(1)7 x .537 .550 .660 .662 .737 .725 .750 .725 .687 .525 1.39*
52 .252 .251 .226 .226 .196 .202 .190 .202 .218 .253
1 750 775 787 837 2 31
601L 1 13 - - - - -
C 3 s2 .190 .177 .169 .138
5 IL 1 7 + _
631LE1;13/2 x .643 .662 .725 .750 .781 .706 .562 .437 .418 .337
x .843 806 787 798 2.24
PL 7+13 : - -
61 (1) -—- 52 .054 .060 .062 .061
X .687 .612 .575 .587 2.24
61PL(Z)7+1§- 2 .218 .240 .247 .245
x .700 .637 .600 .625 .675 .775 .812 .850 .800 .612 1.43*
4.
61PL(5)7 52- $2 .213 .234 .243 .237 .222 .177 .154 .129 .162 .240
x .862 .893 .900 .875 1.961
62le“HYPE-s .063 .049 .047 .073
x 750 812 812 800 2 061
L 2 13+13 - - - - -
62p ( ) -—-s2 .190 .154 .154 .162
* - ..
DL(10,80) . 1.71 and DU(10,80) _ 2.28 at P - .05. 1 . .
D-value inconclusive,
* = = =
DL(4,80) 1.89 and DU(4,80) 2.10 at P .05.

235

Table 5.8 (cont'd.)

 

HIGH CONDITIONING

 

Time in Days
Groups 1 2 3 4 s 6 7 8 9 10 0
.600 .693 .743 .743 .762 .750 .762 .768 .787 .768 1.26*
63P”(1)7+7 82 .129 .092 .082 .082 .082 .101 .120 .120 .100 .120

x .375 .450 .525 .525 .562 .575 .687 .700 .662 .650 1.30*
63P"(2)7*7 52 .237 .251 .253 .253 .249 .247 .218 .213 .226 .230

x .743 .706 .718 .712 2.031
64PHC1)7*13 52 .108 .099 .100 .119

64PH(1)7+13 X .743 .706 .718 .712 .750 .737 .506 .506 .525 .425 1.27*
Wandering S .108 .099 .100 .119 .108 .114 .092 .066 .063 .032

64PH(2)7+13 X .575 .487 .512 .537 1.981
82 .247 .253 .253 .252

x .862 .812 .837 .800 2.24
64PHC437+13 52 .120 .154 .138 .162

x .625 .600 .612 .637 .725 .787 .762 .825 .875 .850 1.27*
64P“(537*13 52 .237 .243 .240 .234 .202 .169 .183 .146 .111 .129

1 .650 .750 .806 .806 2.12
65P“(1)13*1382 .135 .108 .079 .079

x .462 .600 .650 .650 2.071
65P"(2)13*1352 .252 .243 .230 .230

66IH 1 7 x .737 .837 .887 .862 .875 .900 .900 .900 .925 .800 1.49*
C 3 52 .196 .138 .101 .120 .111 .091 .091 .091 .070 .162

 

67mm 22351323232323;
ggIHEI313+ X .850 .900 .931 .918 .900 .868 .650 .581 .593 .531
68mm :2 :23 :32; :32; :38
W__ :52 :32: :23: :22: :33;
688527513. £52 :23 :13?) :33 :31? 1333 :13; :‘iii :33? :332 :33} ”5*
69PH9>13+13§2 2333 :31? :313 :323 “7
69P"(2313+.1_3’§g:3§i :33? 1332 :33 “1
:ELCIO'SO): 1.2: ::::U(10’80): :.i: :: : : .::. ID—value inconclusive.
L(4,80) ' "(4,30) . . .

Hotellings One Sample test on the eXperimental groups in Figures 5.4
Total larvae in the conditioned food (paired, singly, or
none) is compared with an expected vector whose entries are all 0. 5.
Response mean vectors(X), estimated variance vectors(S2 ), and D-values

and S. S.

236

Table 5.9

for serial correlation are found in Table 5. 8.

 

LOW CONDITIONING

HIGH CONDITIONING

 

 

 

 

Group F Group F
56PL(1)7+7 6.02* 63PH(1)7+7 9.SO*
S7PL(1)7+13 8.23* 64PH(1)7+13 12.98*
58PL(1)13+13 Sl.89* 6SPH(1)13+13 23.83*
591L(1)7 4.56* 661H(1)7 23.75*
60IL(1)13 l8.66* 67PIH(1)13 792.07*
61PL(1)7fL§ 63.53* 68PH(1)7+1§_ 83.70*
62PL(1)13+1§_ $25.50* 69PH(1)13+1§_ 238.89*
*

F(10,70) a 3.48 and F(4,76) = 5.18 at P = .001.

237

Table 5.8. It must be remembered, however, that when the resident
groups are so analyzed, half of the larvae are already in the con-
ditioned food. This will be sorted out shortly. It will be noted
that the F-values for high conditioning are generally higher than for
low conditioning indicating greater attraction to high than low con-
ditioning as would have been predicted on the basis of response to
conditioning alone.

Table 5.10 presents the results of the analysis on whether
the number of pairs in conditioned food is significantly different
from random. In every case aggregation is occurring in the conditioned
food and more so in highly conditioned food. Therefore, not only
are total larvae in conditioned food significantly different from
random, but of those larvae there are more pairs than would be pre-
dicted on the basis of chance alone.

Neither of these analyses says anything about whether an
individual's preference for the conditioned food has been altered by
the presence of a conspecific. That is, are the interactions between
the larvae affecting individual behavior? This may be answered by
treating each pair of larvae as a unit and comparing it to a similar
unit from the isolated situation or by comparing only one member of
a pair with an isolate's behavior of the same age. For example, if
animal-animal interaction has no effect on an individual's preference
for conditioning, then the total number of 13 day old larvae, in the
situations where two 13 day old larvae are present, in conditioned
food, should not be different from the response vector of isolated 13

day old larvae; or the response vector for the 7+l3 situation should

238

Table 5.10

Hotelling's One Sample test on the experimental groups in Figures 5.4
and 5.5. Pairs of larvae on conditioned food are compared with an
eXpected vector whose entries are all 0.25. Response mean vectors(X),
estimated variance vectors(Sz), and D-values for serial correlation
are found in Table 5.8.

 

 

 

 

 

Low CONDITIONING HIGH CONDITIONING
Group F Group F
56PL(2)7+7 7.04** 63PH(2)7+7 8.74**
57PL(2)7+13 4.28* 64PH(2)7+13 11.74**
58PL(2)13+13 34.16** 65PH(2)13+13 13.77**
61PL(2)7+13 24.23** 68PH(2)7+1§_ 49.44**
629L(2)13+1§_ 49.41** 69PH(2)13+1§_ l45.96**

*F(10,70) 2.86 and F(4’76) = 4.11 at P = .005.

**F(lo,7o) = 3.48 and F(4’76) = 5.18 at P = .001.

239

not be different from the average response of a 7 day old isolate

and a 13 day old isolate; or the response of a 7 day old larvae when
with a 13 day old larva should not be different from that of a 7 day
old isolate's response. Such comparisons have been made in Table 5.11
and graphed in Figures 5.6 and 5.7. Using Hotelling's Two-Sample
approach it can be seen from Table 5.11 that there are no significant
differences in response on the part of larvae, when grouped or iso-
lated, to low conditioning, although the One-Sample method shows some
differences. Examination of Figure 5.6 shows all the response curves
for respective comparisons to be about the same. However, in every
case the comparisons when high conditioning is used demonstrate that
response of individual larvae when with a conspecific, regardless of
age, is lower than when a similar aged larva is isolated. This is
graphically demonstrated in Figure 5.7. It will also be noted that
the greatest effect of animal-animal interaction is when at least

one of the members of the pair is 13 days old. Therefore, animal-
animal interactions lower individual preferences for conditioned food
only when that food is highly conditioned. We will come back to
these considerations later.

In Table 5.12 are found comparisons examining resident versus
non-resident effects on individual preferences for low and high con-
ditioning, and the mean vectors are graphed in Figures 5.8 and 5.9.
Hotelling's Two-Sample test (F2-values in Table 5.12) shows that in
low conditioning, whether one of the larvae is a 13_day old established
resident in the conditioned food, or both larvae are new to the situ-

ation, it has no effect on individual preferences for conditioning.

240

.nm03m.mv HoHHonom ohm mnouoo> oncommoh :mo2m

A3.“ 5 moocownoc, msoocomonouoxz

.mcomwnomEoo onaom one on» you xwupoa ooamfino>oo u.>u can napoo> popoomxo u.>m

 

 

.Hoo. u a pm em.v nfimmfi.evmo 6am .vm.m ufimqfi.ofivmm... 6:6 .mH.m unea.vvfio on“ .wv.m "mon.oHVHm...
.Ho. u a on av.m unmmfi.qvma 6:6 .66.N "havfi.ofivmo .. 6:8 .mm.m "ASN.6VHG 6:8 .mm.~ "floa.oHVHm ..
.mmo. u a on mm.~ anmfi.¢vmm 6:8 .mH.N unaqfi.ovao . 6:“ .oa.~ "no“.vVHm 6:6..VN.N "moa.oflvfio .
: gm.~ ...HH.6H mHfiHUquo .m> mH+mHfiHVgamm
a : ON.O .om.~ noHVgHmm .m> mH+anmvganm
a : mH.H .am.~ mHflHVOHoo .m> mfl+anevqoam
mm.H mH+nfi~Vgaem .m> «\mHAHVquo + amfingom
a : w6.H ...mm.m a+nflfivgaom .m> hhHUgHmm
amfifiofiflwowa 5869 No Hm .>u .>m
pom pooh H m.xom mcomflhomaou

 

N oszoHeHazoo zoo

 

.omo ouofinmoammo o:u mo oo>poH one mo oncommou oau mo oaom on» on pasonm .omo :fi Honcoaa
ohm oo>HmH umou ozu opp con: .om>HmH nonuwo How Hoooo> omcommon ocp om~< .omm oeom on» mo o>hmH oco mo
omcommon onp mo oamm onp on stosm omm fiasco mo oo>hoH 03H pom muouoo> omcommou ocu .cowpoououcfi Hmawco

-Hoe«:o o: mafiESmm< .¢.m oHSMMm a“ museum on» macaw mcomflnmmaoo ofimaom hivoze paw n.voco m.m:wHHouo=

HH.m oanuh

241

.fimo.w.mu HoHHonom ono mn0p0o> oncommon coo:

a

. Go...“ as moononnoc» mascocomonoaflhI

.mnomnnomaou ofimamm oco onu nom xnnuoa ooconno>oo u.>u woo nouoo> wouoomxo u.>m

 

 

 

a u
.noo. u a on 66.6 n Ammn won“ and eH.w u amen envmo...
q a
.no. u a pm nv.m u nmmn coma 6:8 66.N u amen enema ..
a a q
.Hoo. u a pm w6.m u non Onvnm... .mmo. u a pm mw.~ n mmmn evmm 6:5 mn.~ u amen onvmo .
a : ...mw.m mnnnvmnno .ms mn+mnnnvmamo
: ..wo.m nfinvznoo .m> mn+nflmvmaeo
a m .m~.m mannumnno .m> mn+nflvvmavo
...mn.6 mH+nflnvzavo .m> «\mnnnvznne + nnnvznoo
a : .mm.~ nfinvznoo .m> n+nflnvmamo
amfinonnonmd pooh Nu Hm .>u .>m
now noon NH m.xom mcomnnomeOu

UZHZOHHHQZOU ~1on

 

fl.c.ocooV nn.m ensue

242

Figure 5.6
Mean response vectors for comparisons (Table 5.11) of two larvae
situations with one larva situations to find if individual preference
for low conditioned food is affected by the presence of a conspecific.

A. 591L(1)7 vs. 56PL(1)7+7

B. 6OIL(l)l3 vs. 58PL(1)13+13

 

C. 57PL(1)7+13 vs. 591L(1)7 + 6OIL(1)13
2

D. 591L(l)7 vs. S7PL(S)7+13

E. 57PL(4)7+13 vs. 6OIL(1)13

 

 

243

U

whom an oEnH

Emmwmpw

 

N
22:86 + 2333 Fill.

a 3.23.53 {Illa

 

Lro

rom

.ov

.00

.on

 

SS

mxmo an oEnH

cromrwmwmmw

'PUOD UT GBAJB'] 96

P003

2313.53 win}... 3

233% [Ill

“Ho

Tom

vow

tom

 

rOOH

'puog uI 88A181 %

9005

o.m onswww

<

whoa an oEnb

n o m w m m

P D b '

7:31:18 c.1113.

2333 III.

 

.om

fiooH

 

o
6

'puo3 u: eenxeq

POOd

244

mxmo an oanh

0H m w u o m v m

III I I P I .P P

23.38 gill...

2:93.53 tllll

N

vom

.0?

.00

.om

 

.OOH

Pooa 'puoo at 984191 910 £90 91 %

fl.o.0:663 o.m unsun5

a

mxmo an oanb

0H m w n o m w

P b b I ’ P P I |

21.33.53 III...

23.5% 9111...“.

 

.0w

'0
\O
pooa 'puoo HI aeAze1 910 £90 L %

.om

 

fiooH

245

Figure 5.7

Mean response vectors for comparisons (Table 5.11) of two larvae situ-
ations with one larva situations to find if individual preference for
high conditioned food is affected by the presence of a conspecific.

A.

B.

66IH(1)7 vs. 63PH(1)7+7
67IH(1)13 vs. 65PH(1)13+13

64PH(1)7+13 vs. 66IH(1)7 + 67IH(1)13
2

 

661H(l)7 vs. 64PH(5)7+13

67IH(1)13 vs. 64PH(4)7+13

 

n.m unsun5

U m <

 

 

 

 

mxmo an oEnH mxmo an oenb mxmo an oEnn
pwun .m h m m m m, m H 02 m m m m m w M 1m 2 omirm m m w m N m m1 w
Po Ho F0
N .. . ..
2A32$ + 2328 III: .0... 2.223252 :1... 1.. Ho... $2353 I! 1.. .om
2+235371§ 232$ .11.... 2328 III-1
.05 .05 .os
o/0 0/0
"I. n1
B B
.om M .0m M .om
B B
a 9
I. TL.
.00 u .00 u .00
3 3
m m...
p. x s.
.2 . x .0... . .2.
m . m
D. \\x 0.
flow r.....» .8 .8
.05 .om .om

 

 

 

 

.03 .03 .2:

venxeq %

'puog u;

9009

247

2.0.00003 5.0 onsmno

m n.

 

 

 

whom an osnh mxmo an oEnn
02 0 0 n 0 m 6 m N 2 02 0 0 n 0 m 6 m N 2
III! b P D D D P I| P O P b b r r I P L b3 P o
H H
2.263.560 x... I I a 4. 2.23560 0...... .. 1x -1
.00 .6 .00
022230250 6:111:11 m“ 5223:200 .1111114
.06 .m .06
K
0
I
.00 o. .00
"I
H
m
V00 9 too
I. \KIIK
u
u
m.
.00 m_ .00
.00 .00
.\\u111\\x .002 .002

 

 

poog 'puog up aeAJeq p10 Keg L g

248

.HoHHonmm ono mnopoo> omcommon :0o25

.mmo.mvmv mooconnm> mooooowonopo:25

.mcomnnomsoo oHQEom oco onu nom xnnuoa ooconno>oo u.>o poo nouoo> oouoomxo n.>m

 

 

Q a
a A a A
.5N0. u 5 26 00.N "2552 63N5 6:6 .52.N "2062 023N5 . 606 .00.N u 205 6325 606 .6N.N "205 02325 .
5 05.0 .mw+m22N325N0 .m> 02.522N32505
5 5 0N.2 .05.N 02.52532555 .m> 52232205
5 5 20.N ...00.N .mm+52532520 .m> 52232200
5 5 50.2 ...00.0 52.52532555 .m>.MM+N2532520
05226226205 2565 N5 25 .>0 .>5
now noon H m.xom moomnnomsou

 

N UZHZOHHHQZOU 304

 

.o>n02 omoommon ogp can: mesoonounaanm oooooonucn o>no2 nonuooo mo uoommo onu mo oaom onu on
m>n02 omcommon on“ no o>nm2 ucoonmon o mo uoommo oSH Mn mooconoMMno now on no: omHo onsosm ononh
.moomnnomEoo omonu on mooconommfio moo on no: oaoonm onosp mconuoonoucn HoannoIHosnco o: mana:mm<
.mcncononoooo HoonmoHonn now ooconomonm :o muoommo ucoonmonncoc msmno> unoonmon ocnanouoo on m.m

woo 6.m monsmnm on masonm Houcoannomxo onp macaw mcomnnomsoo oamaom 083039 paw n.3oco m2mcn22opo:

NH.m oHan

249

.Howwonom ono mnouoo> omcommon coo:n2

. 30. m2: 082262.262, mooocomonouozr

.mcomwnoaeoo oHQsom oco ocu now xnnpoe oocownm>oo u.>u woo nouoo> oouoomxo u.>m

 

 

 

.200. u 5 26 60.6 "2002 63N5 6:6 .6N.0 "2062 02355... 656 .02.0 6265 6325 666 .06.0 6205 02325...
.20. u 5 66 56.0 "2002 63N5 666 .06.N "2062 023N5 .. 606 .00.5 "265 6325 606 .00.N "205 02325 ..
.0N0. u 5 66 00.5 «2002 63N5 6:6 .02.N "2062 023N5 . 6:6 .60.N "205 6325 666 .6N.N 6205 02325 .
5 ...05.N2 02+022N3=506 .6>.MM+M22N35506
: .00.0 52235206 .6> 02.52035566
: N0.N ...06.0 52235206 .6>.mm+5203:500
: .06.N 02.52035566 .6>.mm+5203:506

a226226.265 2665 N5 25 .>0 .>5

now pooh me m.xom mcomnnomaou
022202520200 :02:

 

2.6.60663 N2.5 62065

250

Figure 5.8

Mean response vectors for comparisons (Table 5.12) of resident versus
non-resident effects on larval preference for low conditioning.

A. 61PL(5)7+13_vs. 57PL(5)7+13
B. 61PL(5)7+13_vs. 591L(l)7
C. 57PL(5)7+13 vs. 59IL(1)7

D. 62PL(2)13+1§_vs. 58PL(2)13+13

 

 

100'

90<

80'I

70'

601

% 7 Day Old Larvae in Cond. Food

251

I A

-—__.. 61PL(S)7+_1_3_

 

 

 

 

 

30 x— --x 57PL(5)7+13
45
J2 . . - - - - .02.
l 2 3 4 S 6 7 8 9 10
Time in Days
100‘ c
901
'0
o
o
u.
6'
a
o
L)
:
-H
o
o
E
o
.—'l
'o
H
c:
>.
g 40
h
69 '-—--v 57PL(5)7+13
30
" x---l 591L(l)7
0T v v v I 1 v v u ' v
1 2 3 4 5 6 7 8 9 10

Time in Days

 

10m

90'

40

3 7 Day Old Larvae in Cond. Food

30

 

l
r

 

% Paired in Cond. Food

 

3(

 

-——-—- 61PL(5)7+_1_3_

1— --'! 591L(l)7

f v v v v

23356789

Time in Days

H:

-—-——-—-o 62PL(2)13+1§_

{ -- - -¥ 58PL(2)13+13

 

J

Figure 5.8

1 2 3 4 S 6 7 8 9 10
Time in Days

252

Figure 5.9

Mean response vectors for comparisons (Table 5.12) of resident versus
non—resident effects on larval preference for high conditioning.

A.

B.

68PH(5)7+l§_vs. 64PH(5)7+13
68PH(5)7+1§_vs. 661H(1)7
64PH(5)7+13 vs. 66IH(1)7

69PH(2)13+1§_vs. 6SPH(2)13+13

 

 

253

100'

0
O
A

m
C
A

\1
‘2

 

Ox
‘2

U1
C2

.1:
Ca

~—-————. 68PH(S)7+_1_3

% 7 Day Old Larvae in Cond. Food

k» - --x 64PH(5)7+13

 

 

30‘
.J.
01:, , . , . . f . ,fi
1 2 3 4 5 6 7 8 9 10
Time in Days
100‘(:
’1
¢*‘*‘f \
90' f‘f’fi’ \

\1
9.

 

ox
Cl

01
Q

4:.
Q

‘—-—-o 64PH(S)7+13

% 7 Day Old Larvae in Cond. Food

x- - --r 66IH(1)7

 

u
'LQ

 

°—+

3 4 S 6 7 8 9 10
Time in Days

1 5

Figure 5.9

100‘

 

 

 

 

 

90
'U
8 .
m 80
'u'
8
U 70
I:
.fi
3 60‘
E
t6
.—1
T, s
H
0
>4
8 4 ~———-‘ 689H(5)7+_1_3_
l\
a? 3 *- '"" '1 66IH(1)7

- Y i 5 5 s 6 7 8 53—16
Time in Days

100‘ D

80'
'6
O
O
D-

.70
72 2""
'3 «
560i '

I
““350 ’
-H
g 7
6"4
o——-———< 69PH(2)13fl§
3( X“- - -l 65PH(2)13+13

I5 41 E5 (3 77 1% S) l ()

Time in Days

1 2

254

All larvae behave as if isolated. The One-Sample Fl-values, however,
indicate some differences possibly due to the difference in initial
preferences as can be seen in Figure 5.8, although these differences
are probably due to heterogeneous variances. The comparison 64PH(S)7+13
versus 66IH(1)7 is depicted for comparative purposes, as is comparison
64PH(S)7+13 versus 66IH(1)7. However, there are individual behavioral
differences when conditioning is high, Table 5.12, and Figure 5.9. A
7 day old larva's preference for high conditioning is lowered when the
13 day old larva is introduced with it, but is not affected when the
13 day old larva is the resident in the conditioned food lump. The
preference curve for the 7 day old larva when the resident is present
is the same as the 7 day old isolate curve. This indicates that the
significant animal-animal interaction is taking place when the 7 and
13 are initially placed in the center of the dish and not in the food
lump. I will come back to this point in the discussion section. The
two 13 day old larvae cannot be distinguished so the comparison was
based on the number of pairs in conditioned food and it is noted that
a 13 day old larva with a 13 day old resident is not affected but it
is if both 13 day old larvae are introduced together.

In the next experiments I have selected only the situation
where the 7 and 13 day old larvae are introduced together into a
situation with high conditioning. These experiments will show that
there is an interaction between biological conditioning and animal—

animal interactions in the responses we have been examining.

255

Experiment III: Are the Lowered Responses
of Younger, 7 Day Old Larvae, When a

13 Day Old Larva is Simultaneously
Introduced, a Result of the Changing
Biological Conditioning, Interaction
with the 13 Day Old Larva, or Both?

 

 

 

 

 

 

In the last experiment we saw that a 7 day old larva was rela-
tively unaffected by a 13 day old larva if the 13 day old larva was
a resident in the highly conditioned food, but its preference for
conditioned food was significantly lowered if the 13 day old larva
was introduced at the same time. However, re-examination of the
response vectors (see Table 5.12 and Figure 5.10) shows that the 7
day old response gradually rises to a level where it would have been
if the 13 day old larva had not been present. In Figure 5.10, I
have also graphed the response vector for the 13 day old larva
including its wandering phase at which time it leaves the conditioned
food and moves about the dish until a suitable pupation site is
located and pupation occurs. Notice that as the 13 day old curve
falls the 7 day old curve rises, the two crossing at about day S of
the experiment, or when the 13 day old larva is 17 1/2 days old and the
7 day old is 11 1/2 days old. This suggests that when the 13 day
old larva enters the wandering phase and vacates the conditioned food
lump, the 7 day old larva moves to that lump, or that the 13 day old
larva moves over to the 7 day old larva's lump, and in some way inter-
acts with it causing the 7 day old larva to shift food lumps. There
may, of course, be an interaction between the two. Since it is the
older larva that seems to be determining the behavior of the 7 day old
larva, older larva, i.e., 13 day old, will be used for manipulative

purposes and only the 7 day old response will be monitored.

% Larvae in Cond. Food

1001

90'

80J

70‘

40‘
30'
20'

10‘

256

 

fl

\

)I—x—x

x- - - ——x 64PH(4)7+13

‘-'--'0 64PH(S)7+13

 

 

I I I I I r T j I v I r v

2 3 4 S 6 7 8 9 10 ll 12 13 14

Time in Days

Figure 5.10

Mean response vectors of a 7 day old larva and a 13 day
old larva with the wandering phase of the 13 day old larva
included. These are from experimental group 64PH(l-5)7+13.

257

If the rise in the 7 day old response curve is a function of
some interaction due to the wandering phase of the 13 day old larva,
we should be able to shift the 7 day old curve two days earlier (to
the left) or two days later (to the right) by pairing 7 day old larvae
with 15 or 11 day old larvae, respectively. To test this hypothesis,
the following 4-group experiment was designed. In one group, a 7 day
old and a 15 day old larva were introduced, in another, a 7 and 13, and
in a third a 7 and 11 were used. The fourth group is a 7 day old
isolate larva with high conditioning. All experimental procedures
are as previously described and one of the two food lumps was highly
conditioned. For identification purposes the 4 groups are:

70PH(1-S)7+15

71PH(l-S)7+13

72PH(l-5)7+11

73IH(1)7

The mean response vectors (i), estimated variance vectors (82),
and D-values for each 7 and the ll, 13, and 15 day old larvae from
these 4 groups are presented in Table 5.13 and graphed in Figure 5.11.
However, only comparisons among the 7 day old response curves are of
interest, but the response of the older larvae are shown for compara-
tive purposes. Table 5.14 shows the results of the One- and Two-Sample
comparisons among the three 7 day old preference curves when in the
presence of high conditioning and either a 15, 13, or 11 day old con-
specific.

It can be seen from the results of this experiment that the

response to high conditioning of a 7 day old larva when paired with

2

Table 5.13

58

This table presents the response mean vectors(X), estimated variance
vectors(S2 ), and D- values for serial correlation for the groups to be
analyzed in Table 5.14.

 

HIGH CONDITIONING

 

Time in Days

 

 

Group 1 2 3 4 s 6 7 8 9 10 D

x .587 .612 .725 .837 .912 .887 .900 .900 .925 .837 1.41*
7°?"(5)7*15 $2 .245 .240 .202 .138 .081 .101 .091 .091 .070 .138

x .600 .625 .587 .625 .725 .712 .825 .850 .875 .850 1.38*
71PH(5)7+13 82 .243 .237 .245 .237 .202 .202 .146 .129 .111 .129

X .625 .587 .600 .612 .675 .700 .687 .837 .900 .812 1.55*
72P“(5)7+11 $2 .237 .245 .243 .246 .222 .213 .218 .138 .091 .154
73IH(1)7 i .712 .887 .875 .887 .925 .900 .900 .912 .900 .850 1.62*

82 .207 .101 .111 .101 .070 .091 .091 .081 .091 .129

2 .850 .900 .912 .812 .800 .425 1.831
7OPH(4)7*15 32 .129 .091 .081 .154 .162 .247

x .850 .837 .850 .850 .787 .612 .337 .087 1.39*
71PH(4)7+13 32 .129 .138 .129 .129 .169 .240 .226 .081

X .837 .825 .862 .887 .837 .837 .875 .700 .250 .050 1.55*
72PHC437+11 $2 .138 .146 .120 .101 .138 .138 .111 .213 .190 .048
*
DL(6,80) — 1.83 and DU(6,8O) = 2.16 at = .05.
*
DL(8,80) = 1.77 and DU(8,8O) 2 2.22 at = .05.
*
DL(10,80)= 1.71 and DU(10,80)= 2.28 at = .05.

I . .
D-value was inconclu51ve.

259

Figure 5.11

Mean response vectors of 7 day old larvae to high conditioning when a
15, 13, or 11 day old conspecific is also present. The mean response
vectors for the 15, 13, and 11 day old larvae are shown, with their
wandering phases, for comparative purposes.

Illustrated are the mean response vectors for:

A.

B.

7 and 15 day old larvae.
7 and 13 day old larvae.
7 and 11 day old larvae.
All the 7 day old response vectors from A, B, and C, as well

as the response vector for a 7 day old isolate to the same
initial degree of conditioning.

260

100'

90‘

on
0
LL

\1
Ci

60q

 

l--“*~-~

% Larvae in Cond. Food

0'- -' -- -¢ 70PH(4)7+15

 

' '_'-—-‘ 70PH(5)7+15

J"

j

153586789.

Time in Days
100

90

00
q

\I
Q

Larvae in Cond. Food

 

o.
’o

A
9

w- - -- w 72PHC4J7+11

3EL —— 72PH(S)7+11

0T

 

V U ‘

123415678910

Time in Days

.Figure 5.11

 

 

 

    
   

1001 B
901
80‘
-u
0
0
LL.
, 70
'o
a
8
:60‘
-H ‘
8
> 5 \
’3 1
.—l
o\° ‘
4 x
l
.........71PH(4)7+13 \
\
-—--—~71PH(S)7+13 ‘L‘m‘
' i E 3 i 2'; 6 7 8 876
Time in Days
lOO‘lE)
,“'-’¢O"
90‘ ’
I
801 [I
l
l

\I
Q

   

if".

Larvae in’Cond. Food
—SL-

 

 

'v ’/
1 ”)4
so
°“°4 .—-————. 7OPH(5)7+15
x 1: 71PH(S)7+13
3 o... N.-. 72PH(5)7+11

-----~ 73IH(1)7

12345678910
Time in Days

 

261

Table 5.14

Hotelling's One(') and Two(*) Sample comparisons of the 7 day old
mean response vectors when paired with an ll, 13, or 15 day old
larva or when an isolate, all with initial high conditioning on

one food lump.

 

 

 

Comparisons Box's T2 Test for

EV. CV. F1 F2 Test Parallelism
71PH(5)7+13 vs. 70PH(5)7+15 2.24* H
71PH(5)7+13 vs. 72PH(S)7+11 2.16 0.90 H P
72PH(5)7+11 vs. 70PH(5)7+15 2.87* H

73IH(1)7 vs. 70PH(S)7+15 7.08"' 1.96 H

73IH(1)7 vs. 71PH(S)7+13 2.60* H

73IH(1)7 vs. 72PH(5)7+11 3.19* H
' Fl(10,70)= 2.24 and F2(10,149)= 2.15 at P = .05.
" 81(10,70)=2.59 and ** F2(10,149)= 3.59 at p - .01.
"'Fl(10’70)=3.48 and ***F2(10,149)= 5.18 at P = .001.
EV.= expected vector and CV.= covariance matrix for the One Sample

comparisons.

HHeterogeneous variances (P

PMean response vectors are parallel (P‘&.

E .05).

05).

262

a 15 day old larva is significantly higher than its response when
paired with a 13 or an 11 day old larva. However, the 7 day old
response when paired with a 13 day old larva is not different from
when paired with an 11 day old larva. In Figure 5.11, it can be seen
that as either the 15 or 13 day old larvae enter the wandering phase,
their response curves begin to fall and that of the 7 day old response
to conditioning begins to increase. In the 7+ll situation, however,
the 7 day old response is on the rise before the 11 day old larva
enters the wandering phase. It can be seen also that the response of
a 7 day old larva to high conditioning is significantly lower when

an 11 or 13 day old larva is present than when the 7 day old larva is
alone. The 7 day old response when with a 15 day old larva, however,
is not different from the isolated condition, although it would be

at the .05 level, and is different with the One-Sample test, indicating
a problem due to heterogeneous variances.

All comparisons, except one, yield non-parallel responses
indicating the possibility that the 7 day old larvae are utilizing
different behaviors in their response when a conspecific is present,
than when isolated. These responses also appear to be different with
different aged conspecifics, except in the 7+ll versus 7+l3 situation.

It is very tempting to say that the response of’7 day old
larvae to conditioning is lower because it has been repelled by the
older larva. However, in Table 5.15 are presented the results of
comparing the first three days of each 7 day old response mean vector
to a vector whose entries are all 0.5 to see if repulsion has occurred.

These results show (in combination with Figure 5.11) that the initial

263

Table 5.15

Hotelling's One Sample (') comparison of the first 3 days of the 7 day
old response vectors when with various aged conspecifics and when
isolated, with an expected vector whose entries are all 0.5.

 

 

 

Comparisons Fl
7OPH(5)7+15 7.12"
71PH(5)7+13 1.74
72PH(5)7+11 1.95
73IH(1)7 43.15"

I = =

F(3’77) 3.29 at P .025.

u = =

F(3’77) 6.02 at P .001.

responses, although lower than if isolated, are all either not dif-
ferent from random or significantly higher than random. Therefore,
although the presence of an older conspecific significantly lowers the
response of a 7 day old larva to conditioning, the term repulsion is
not appropriate since the responses are not significantly lower than
random.

These results are suggestive that the response of a 7 day old
larva to conditioning when paired with an older conspecific may be the
result of an interaction between biological conditioning and animal-
animal interaction. The initial lower response in each case being
due to some interaction between the 7 day old larva and the conspecific
in the center of the dish before either larva has sampled the food
lumps. The rise in the 7 day old response with time, however, is more
difficult to explain. Except when paired with an 11 day old con—

specific, the 7 day old response seems to coincide with the 13 or 15

264

day old wandering phase. It may be that when the older larva leaves
the conditioned lump it goes over to the 7 day old's lump and a
second interaction occurs causing the 7 day old larva to move to a
new food lump. However, it must be remembered that when the older
larva chooses the originally conditioned lump, it spends at least a
day, and in the case of the 13 day old larva, 3 days, adding its own
conditioning to that already present. Therefore, when the older
larva leaves this lump, the 7 day old larva may be responding to the
extra conditioning. There may also be an interaction between these
two events. The following experiment was designed to look at these
possibilities, using only the 7+13 situation since the greateSt
effects on the 7 have been with a 13 day old larva.

The next hypothesis to be examined is that the response of the
7 day old larva to conditioning when a 13 day old larva is present is
due to an interaction between three variables: (1) the initial
encounter between the two larvae in the middle of the dishes, (2) the
extra conditioning that the 13 day old larva adds to the already con-
ditioned lump, and (3) a second interaction between the two larvae
when the 13 begins wandering. These events will henceforth be
referred to as l, C, or 2 respectively. The following experimental
groups were set up with various combinations of the above variables:

74IH(l)7(O-O-O) : This is the 7 isolate situation having
none of the above variables operating.

7SPH(S)7+13(1-C-2): This is the situation in which a 7 and
13 day old larva are placed in the
experimental situation and the 7 is
monitored for the total experimental
period.

265

76PH(5)7+13(l-0-0): In this group the 7 and 13 day old larvae
are placed in the experimental situation
and in 12 hours the 13 is removed. There-
fore, the 7 received the initial encounter
but not the second encounter and there was
no extra conditioning being done by the 13,
or at least only a very small amount in 12
hours.

77PH(5)7+13(l-C-0): The 7 and 13 were placed in the apparatus
and when the 13 day old larva was 16 days
old, just before wandering, it was removed
from the dish. Thus, the 7 received the
initial encounter and the extra conditioning,
but not the second encounter with the 13 day
old larva.

78PH(5)7+13(l-O-2): The 7 and 13 were placed in the dishes and
12 hours later the 13 was removed. Four
days later, or when the 13 would have
begun wandering, a wandering larva from
the colony was placed in the dishes. There-
fore, the 7 day old larva received the
initial encounter and second encounter but
no extra conditioning had been done by the
13 day old larva.

Each of these 5 groups received different manipulations through-
out the experimental period, and as a control for this the following
was done. Whenever the 13 day old larva was removed from the dishes
in the respective groups, the dishes in the groups not receiving this
treatment were opened and the food lumps gently tapped with a pair of
forceps. When a wandering larva was added to group 78PH(5)7+13(l—O-2)
all other dishes were opened and the center floor of the dish gently
tapped with forceps.

The mean vectors (X), estimated variance vectors (82) and
D-values for serial correlation are shown in Table 5.16 for the 7 day
old test larvae in each group. The results of Hotelling's One- and

Two-Sample comparisons between all possible combinations of 7 day

response vectors from Table 5.16 are shown in Table 5.17. The

266

wom.owva wom.0wvmo

a.

.mo. u m an wm.mn Q cam H5.H u

 

mew. omw. emw. emw. emw. wmw. oew. mew. mew. mew. wm

.eo.w wmm. ome. wmm. wmm. wme. mwm. wHe. wmm. wmm. wmm. w fio-o-wvmw+wwmvma6w
mww. wow. 6mm. mww. wow. www. oew. mew. mew. mew. wm

.mm.w oow. mww. wmw. oow. www. wmm. wwm. com. wmm. wmm. w wo-u-wvmw+wwmvzaww
wow. mmw. mew. wow. mew. mew. oew. oew. wew. wmw. wm

lwm.w mww. wow. wmw. com. www. wmw. wwm. wwm. mwm. mwm. m Aw-o-wvmw+wwmvzwww
wmo. HHH. mew. mmw. omw. wow. wmw. mew. mew. mew. wm

.mw.w mwm. mwm. mwm. wow. omw. www. mwm. wmm. com. com. m ww-u-wvmw+wflmvzamw

mwm. owo. owo. Hmo. How. owe. Hem. oww. mmm. wow. wm
.mm.w omm. mwm. mwm. com. wmm. mwm. wmm. wmm. wmm. mww. x wo-o-ovwfiwvmwew

 

nammwmmemwm mango
mw<a cm 65mm

 

.omeuommcou vHo mew ma m gum: Houcsooao vacuum

may cam .mchOwuwwcoo mnuxo .Houcsooco Hmwuwcw mo muommmo any wow powwow mam
-mn masonw Hancoswnomxo on» Scum om>hmH vac mew w on» we comuawonhoo Hawnom pom
monam>ua use .flmmeHOpoo> oocmwum> woumawumo .Amvmnouoo> omcommou cams mo manmb

eH.w ofinmh

267

Table 5.17

Results of Hotelling's One(') and Two(*) Sample comparisons among the
7 day old response vectors in Table 5.16.

 

 

 

Comparisons Box's T2 Test for

EV. CV. F1 F2 Test Parallelism
(l-C-Z) vs. (0—0-0) 2.17* H P
(1-0-2) vs (0-0-0) 2.93** H P
(l-C-O) vs. (0-0-0) 3.28** H P
(1-0-0) vs (0-0-0) 3.73*** H P
(1-0-2) vs (l-C-Z) 2.31' 0.85 H P
(l-C-O) vs. (l-C-2) 2.19 0.95 H P
(1-0-0) vs (1-C-2) 4.35"' 1.58 H P
(1-C-O) vs (1-0-2) 0.97 0.44 H P
(1-0-0) vs. (1-0-2) 2.11 0.99 H P
(1-0-0) vs (l-C-O) 0.68 0.35 H P
O *

F1(10,70) - 2.24 and F2(10,149) = 2.15 at P = .025.
II **

F1(10,7O) I 2.59 and F2(10,149) = 2.46 at P = .01.
It! ***

F1(10,70) - 3.48 and F2(10’149) = 3.24 at P = .001.

EV.= expected vector and CV.= covariance matrix for the One Sample Test.
HHeterogeneous variances (Pf .05).

1)Mean response vectors are parallel (P:’: .05).

268

response mean vectors for the 7 day old larvae from all groups are
graphed in Figure 5.12. (For the purposes of identifying these

groups in the tables and figures, I have used only the three numbers
for the variables in question, i.e., l-C-2, l-C-O, 1-0-0, 1-0-2, or
O-O-O. Since all groups deal with only the response of the 7 day old
larva this system makes it clearer which variables are being referred
to.) It can be seen from the Two-Sample comparisons (Table 5.17) that
the 7 day old isolate response to conditioning is significantly higher
than all the others, but that there are no other significant differ-
ences. The One-Sample test elucidates only two other differences.

All responses are parallel, however, indicating similar behaviors but
at different levels.

These results were a bit surprising in light of the separation
of response curves in Figure 5.12. However, examination of the data
indicates some possible reasons for the results obtained. The range
in which the responses have in which to separate themselves out is a
relatively narrow one during the time in which the responses begin to
exhibit differences; from about 60% to 85%. The variances are also
heterogeneous, as attested to by Box's test and the results of the
One-Sample test.

Throughout this chapter, as well as Chapter IV, I have alluded
to this fact. If we examine the variance vectors for these groups,
and in fact for all groups in this Chapter, an interesting trend will
be noted. As response to conditioning increases the variances
decrease. Therefore, the variance vectors may be a good mirror for

larval activity. For example, examine the variance vectors for groups

o
6

Larvae in Cond. Food

269

 

 

 

 

 

 

1001
904
80*
70‘
604
50*
404
30' O-~~ C-I-a-a-O (O'O'O)
:- fi (l-C-Z)
20' o . (l-O-O)
I 1 (1-0-2)
10‘ ‘-------o (l-C-O)
0 V V f 1 t I 1 I j v v v r fl
1 3 4 S 6 7 8 9 10 11 12. 13 14

Time in Days

Figure 5.12

Response mean vectors for the 7 day old larvae from all
experimental groups in Tables 5.16 and 5.17.

270

l-C-Z and 1-0-0 and notice that in Figure 5.2, as the response vector
for l-C-2 rises, the variances get smaller, perhaps indicating a more
stable activity of the larva as they select and stay in the conditioned
food. The variances for group l-O-O, however, remain large, and as

can be seen, the response vector for this group never rises much.

That is, this group does not get to conditioned food and their

activity is high or fluctuating. This is the most probable explanation
for the lack of significant results in the comparisons in Table 5.17.

Because of the heterogeneous variances, it was decided to
analyze the last three data points for each of the response vectors.

It can be seen from Table 5.16 and Figure 5.12 that the variances have
stabilized, although still large in some groups, and the response
vectors have relatively stabilized at this time. We would not expect
to get some of the same differences in this comparison since some of
the responses are now at the same level. For example, groups l-C-2
and 0-0-0 are no longer different because l-C-2 is at the same level
as 0—0—0 in the latter stages of the experiment. Table 5.18 shows

the results of these comparisons. The only new difference now
detected is that l-O-O is a lower response than l-C-2. Tables 5.17
and 5.18, therefore, indicate the following.

All manipulations produced significantly lower responses on the
part of the 7 day old larvae than for the response 0f isolated 7 day
old larvae. That is, the presence of an older larva, even if only
for 12 hours, is a sufficient stimulus to prevent the response of 7 day
old larvae to conditioning from reaching the level of an isolate's

response. This is also a sufficient stimulus for preventing 7 day

271

Table 5.18

Results of Hotelling's One(') and Two(") Sample comparisons among
the 7 day old response vectors in Table 5.16 for the last 3 days
of each mean response vector.

 

 

 

Comparisons Box's T2 Test for
EV. CV. F1 F2 Test Parallelism
(l-C-2) vs. (0-0-0) 4.96" 1.58 H P
(1-0-2) vs. (0-0-0) 10.ll"' 2.81 H
(l-C-O) vs. (0-0-0) 3.95* H P
(1-0-0) vs. (0-0-0) 7.66*** H P
(1-0-2) vs. (l—C-Z) 5.69" 2.01 H P
(l-C-O) vs. (l-C-2) 6.87"' 2.77 H P
(1-0-0) vs. (l-C-Z) 4.40** H P
(1-0-0) vs. (l-C-O) 0.94 P
(1-0-2) vs. (1-0-0) 3.84' 1.67 H P
(l-C-O) vs. (1-0-2) 0.20 P
' 32 d *
Fl(3’77) - . 9 an F2(3,162) = 3.22 at P = .025.
t! *1:
Fl(3,77) = 4.05 and F2(3,l62) = 3.94 at P = .01.
'Il ***
Fl(3,77) - 6.01 and F2(3,162) - 5.78 at P = .001.
EV.= expected vector and CV.= covariance matrix for the One Sample Test.

HHeterogeneous variances (P $.05).

pMean vectors are parallel (p $.05) .

272

old larvae from reaching the same response level as a 7 day old larva
that received the two encounters with the older larva and the extra
conditioning. The other two groups are intermediate. That is, the
groups that received both encounters but no extra conditioning or only
the first encounter with extra conditioning are statistically not
different from the group that received only the first encounter or
the group receiving all three stimuli. However, the trend for the
response vectors of these two groups is that they lie between the
1-0-0 and l-C-2 groups. It would also appear that the l-C-O group

is closer to the l-C-Z group. This may indicate that the extra con-
ditioning is not as necessary to overcome the initial encounter
(i.e., l-C-O) as is the second encounter without the conditioning
(i.e., 1-0-2). The F2-values also indicate this trend. Although
suggestive, these "trends" are not significantly different from one
another. It would also appear that these experiments have not yet
made it possible to make definitive statements about certain aspects
of the interaction between biological conditioning and animal-animal
interactions, although there are some strong indications. This will
be discussed in the discussion section.

A point made earlier in this chapter is worth repeating at
this time. Even though the 7 day old response vectors are signifi-
cantly lowered by the presence of an older conspecific, they are not
being repulsed. If repulsion were the mechanism, then we would
expect the response vectors to be significantly below random, i.e.,
below 509. Table 5.19 shows the results of comparing the first 5

days of response for each experimental group, since it is during the

273

Table 5.19

Results of Hotelling's One Sample comparison of the first 5 data points
of each group from Table 5.16 with an expected vector whose entries
are all 0.5.

 

 

 

Group F1
0-0-0 277.62"'
l-C-2 4.22"
1-0-2 13l.42"'
l-C-O 2.78'
1-0-0 1.31
' = =
F(5’75) 2.74 at P .025.
H = =
F(5,75) 3.28 at P .01.
m = =
F(5,7S) 4.64 at P .001.

first half of the experimental period that all responses are lowest.
The results show that all groups are either random or significantly
above random.

These differences would not be as great if fewer data points
had been used, but the point is that all groups, except group l-O-O,
are significantly different from random in the direction of attraction
to conditioned food and therefore not repulsed in the strict sense
of the word. Group l-O-O is also not repulsed but is randomly
responding. This may indicate the need for a second stimulus to
counteract the effect of the initial encounter with the 13 day old
larva.

Throughout this chapter, several groups have been replicated

for various experiments. In particular, the 7 day old isolate group

274

and the 7 day old group when paired with a 13 day old larva have each
been replicated three times. Table 5.20 presents the results of com-
paring these replicates to see if they differ from one experiment to
the next.

Table 5.20 demonstrates that all comparisons are not signifi-
cantly different with the Two-Sample test, but that one is different
with the One-Sample test. This is the first hint that populations of

larvae may be genetically changing with time.

Discussion

 

Invertebrate studies indicate that spatial distribution is a
function of an interaction between biological conditioning and animal-
animal interactions. These studies, however, are population studies
in which it is impossible to separate individual behavior from group
behavior, and there has been no systematic attempt to do so. In

Park's work (1934, 1935) the distribution of Tribolium confusum was

 

concluded to be heavily dependent on the amount of homotypic con-
ditioning. However, as population size increases, not only does con-
ditioning increase, but the number of adults and larvae increases.
What appear to be responses to increased conditioning may well be
responses to increased density or an interaction between the two.
Naylor (1959) attempted to separate these variables and was partially
successful. His results indicate that spatial patterns in adult
Tribolium are primarily dependent upon interactions between adults as
density increases but that these patterns are modifiable when sexual
relationships and biological conditioning are introduced as variables.

However, Naylor was not able to determine the relationship between

275

. Goiwn: 3:33 one muouoo> omcommou cmozd

. amo. w my moocomuo> msoocowowouoxz

.pmoe owmaom oco on» pom xwwuos ooaomuo>oo u.>o one wouoo> ooooomxo u.>m

 

 

q q
.wo. . a we me.w u Amew cwvww.. use mm.w u how cwvwm..
.mwo. u a we mw.w u nmew vawm . ea“ ew.w u how oHVHm .
a m Hm.o ew.w mw+wfimvzaem .m> mw+whmvzwmw
a : wm.o ew.w mw+wfimvmaww .m> fiw-u-wvmw+wfimvzamw
a x wm.o mm.o mw+wnmvzwem .m> Aw-u-wvmm+wflmv:wmw
a z wm.o Hw.w wo-o-onw+wnszHew .m> wflwvzmmm
a : om.o ww.o wo-o-onw+wwwV:Hew .m> wmmvmmmw
a m we 8 .mm w wflssmwem ..> wwscmwmw
smwwomwewew umow wm Hm .>o .>m
Ham pooh Na m.xom mnemmnomaou

 

.uoumono
was» unocwsonnu powwowamou mmsoum maowno> mo mconwuomaoo onsom awvozh one m.vo:o m.m:«HHouo: mo mHHSmom

ew.w ofinmh

276

biological conditioning and adult interactions partly because he did
not utilize a graded series of conditionings. Wellington's (1957)

work with tent caterpillars was free of the adult-young interactions

of Park's (1935) work, but he did not determine whether the preferences
of these larvae for homotypic conditioning was affected by the presence
of conspecifics. In 1963a and l964e, Surtees concluded from his
studies on grain weevils that spatial patterns in high densities of
weevils are a result of changes in individual behavior as a result of
stimulation from conspecifics. However, his studies are confounded

by the presence of high biological conditioning which he neither con-
sidered nor controlled for.

A previous paper (Chapter IV) demonstrated that the distribution
of isolate larvae is a function of larval conditioning and that pref-
erence for conditioned medium is a function of the degree of condi-
tioning and the age of the responding larva. The rationale for my
studies in this chapter has been that the preference of individual
Galleria larvae for homotypic conditioning is a function of the degree
of conditioning, the presence of a conspecific, and the age and loca-
tion of the conspecific.

Individual Responses to a Conspecific
and Biological Conditioning.

 

 

When there is no previously conditioned food available, larvae
tend to distribute themselves randomly when isolated or when a con-
specific is present. There is, initially, neither an attraction or a
repulsion between pairs of larvae regardless of age. With time,

however, larvae begin to aggregate, apparently as a result of their

277

own conditioning and possibly due to the development of interactions
between different aged larvae. It was demonstrated, for example, that
if one member of a pair of larvae is a 13 day old larva, aggregation
does not occur and larval distribution remains random. Younger larvae
do begin to aggregate and the most significant case was the 7 and 11
day old situation. Analysis further demonstrated that the 7 day old
larva is moving to the food lump on which the 11 day old larva is
residing, probably as a result of the 11 day old larval conditioning
being greater. However, such aggregation behavior would also neces-
sitate that each larva tolerate the close proximity of another larva,
whereas this is not the case with a 13 day old organism. The lack of
aggregation with 13 day old larvae is probably not because 13 day old
larvae are secreting a qualitatively different conditioning than
younger larvae, since isolate 7 or 11 day old larvae have been shown
to be attracted to the same degree of conditioning when the 13 day old
larva is absent (see Chapter IV).

Situations with two 7 or two 11 day old larvae also exhibit
aggregation behavior. This is probably not a case of one larva's
conditioning being greater than the other and thus attracting the
other larva, although there may be individual differences in con-
ditioning behavior. Younger larvae do not initially respond to each
other, but when they each begin conditioning and connecting the food
lumps with silk tunnels this may increase the probability of finding
them in pairs. This behavior may also be related to activity rhythms.
Long (1955) demonstrated that different ages of various butterfly

larvae exhibit different activity rhythms. He found that the closer

278

to pupation these larvae get, the more "restless" they become. This
restlessness does not necessarily mean they move about more, only
that they are more easily agitated and exhibit rather sudden and
jerky movements when touched. Younger larvae tend to be more docile
and move about with more easy-going motions. Therefore, young larvae
may tolerate the presence of young, or not very old larvae, since
their activity rhythms are more synchronous whereas the activity
patterns of older larvae are disruptive to group formation.

Experiment II indicates that larval interactions are associ-
ated with the presence of an already conditioned food lump. The pref-
erence behavior of individual larvae to low-conditioned food is the
same whether isolated or with a conspecific, regardless of the age of
the conspecific. As seen in Chapter IV, however, the response of
isolate larvae to this degree of conditioning, particularly for a 7
day old larva, is already at a relatively low level and it would
almost require repulsive behavior to detect a difference between the
isolate and group situation. When one food lump is highly conditioned,
however, the preference of individual larvae is affected by the pres-
ence of a conspecific, regardless of the age of the conspecific. For
example, the total number of 7 day old larvae on conditioned food when
two 7 day old larvae are present is lower than when only one larva
is present. This distinction is difficult to elucidate since the two
larvae are not individually distinguishable. However, it can be
definitively seen that a 13 day old larva's response to conditioning
is lowered when a 7 day old larva is present and a 7 day old's

response is lowered when a 13 is present, as compared to the isolate

279

situation. This suggests that although the greatest effect is on the
7 day old larva's response, the 7 day old larva also affects the
response of an older larva.

The response of individual larvae to a conditioned food lump
of low or high conditioning is unaffected by the presence of a
resident larva (13 day old larva) in that lump. That is, individuals
respond to such a situation as if the resident were not present. The
conditioning may act as a buffer, allowing organisms to be paired
that do not pair in the absence of conditioning. It is also possible
that the test larva and the resident never encounter one another in
the conditioned lump.

Experiment III elucidated the results of Experiment 11, in
that 7 day old response to highly conditioned food is lowered when
a 15, 13, or 11 day old larva is introduced with the 7. There may be
a relationship between the wandering phase of the older larva and the
rise in response of the younger larvae. The 7 day old response begins
increasing two days earlier when a 15 day old larva is present than
when a 13 day old larva is present, the wandering phase of a 15 day
old larva occurring two days sooner than for a 13 day old larva.
However, the response vector of the 7 day old larva begins rising
earlier than predicted when an 11 day old larva is present. This may
be a function of the initial encounter with an 11 day old larva not
being as traumatic for a 7 day old larva as it is with a 13 or 15 day
old larva. This result also supports the finding in Experiment I
that 7 day old larvae are attracted to where the 11 day old larva is

residing.

280

Further experiments attempted to dissect the behavior of the
younger larva with the hypothesis that its lower response was due to
an interaction between the initial encounter with the older larva,
the extra conditioning the older larva adds to the already conditioned
food lump, and a second encounter when the older larva begins wander-
ing. The results demonstrated that any combination of these variables
is sufficient to significantly lower the younger larva's response from
that of the isolate situation. It was also found that the situation
in which only the initial encounter with the older larva was the
variable that response was significantly lower than when all three
variables were present. No other differences were found, although
there is a decided trend for the situations in which both encounters
are present or only the first encounter and extra conditioning were
present, to be intermediate between the others. There are several
reasons why these trends were not significant and they will be dis-
cussed in the next section.

The general conclusion from all of these studies is that indi-
vidual behavior and distribution are functions of the degree of homo-
typic conditioning, age of individual larvae, presence of a conspecific,
and location of the conspecific. There are some indications from
these studies that differences in larval activity may play a role in
spatial distribution. By larval activity I am referring to the time
in development when larvae tend to move from one food lump to another,
but not movement within any one food lump. For example, it was demon-
strated in Chapter IV that 7 day old larvae are less active during the

first 2—3 days of the experiment, as measured by the number of times

281

they move from one food lump to another, whereas 13 day old larvae are
more active at that time. Mobility differences, on the other hand,
refer to the ability of larvae to move from one food lump to another.
It was previously pointed out (Chapter IV) that older larvae may be
more mobile than younger larvae and can more easily cross the open
space between food lumps.

In the next section I will discuss the possibility that dif-
ferences in larval activity may play a role in the spatial distribution
of Galleria larvae, and propose several hypotheses for future testing.
The discussion, however, will be mostly theoretical.

Possible Mechanisms for the Interaction
of Biological Conditionigg and Larval-

Larval Interaction in the Spatial
Distribution of Galleria Larvae

 

 

 

 

In all of the experiments presented in this chapter, the only
situations in which larval response to conditioning was significantly
lower than in the isolate situation were when larvae initially
encountered each other in the center of the experimental dishes and
when a highly conditioned food lump was present at the start of the
experiment. Assuming high conditioning to be a preferred stimulus,
some sort of competition may be involved in the larval interactions
for it. However, we saw in Chapter IV, that older isolate larvae
exhibit a higher initial preference for highly conditioned food than
do younger larvae, and it was postulated that this is related to
activity differences between the larvae. Older larvae find the con-
ditioned food lump sooner than young larvae and cease their movements

in the preferred stimulus, although differential sensory capabilities

282

may also be involved. The young larvae, on the other hand, tend to
remain in the first food lump encountered for a period of time and
move to the conditioned food lump later in development. If now, a
younger larva encounters an older larva in the center of the dish and
if the older larva is preferentially heading for the conditioned food,
the younger larva may merely head in the opposite direction as an
"escape" reaction, which would enhance the normally lower initial
response of young larvae. I have also observed that these larvae
release some sort of liquid substance if disturbed. Upon encountering
each other, in the center of the dish this substance may be released,
causing avoidance between the two larvae and differentially having a
greater effect on the younger larva. It is not clear what consti-
tutes the interaction, but it is clear from these experiments that
the initial interaction is a prerequisite for lowered responses to
conditioned food. It is also indicated that this interaction is

less traumatic when between two larvae of not very different ages.

For example, the 7 day old larva's response curve rises relatively
sooner when the interaction was between a 7 and 11 day old larva than
between a 7 and 13 or 15 day old larva.

When the 7 day old response curve begins to exhibit a rise, 1
have postulated that it is a result of a second interaction between
the two larvae, or a response to the extra conditioning the older
larva has done to the already conditioned food, or an interaction
between the two. These variables may be linked to activity differences.
After the initial encounter between the two larvae, the 7 day old

response to conditioning is depressed and remains so for about four

283

days. At the end of those four days the 13 day old larva is entering
the wandering phase and leaves the conditioned lump. At the same
time, the 7 day old larva, which is now 11 days old, may be more active
and begins to explore its environment. This was indicated in Chapter
IV. It thus encounters the highly conditioned food which is more
highly conditioned than the food it just left and it remains there.
Thus its response to conditioning increases. This would explain why
the 7 day old response increases sooner when an 11 day old larva is
present but which is not yet wandering. If this were the only
mechanism, however, we would predict that the response curve for the
group in which only an initial encounter with the 13 day old larva
took place, to begin an upswing about four days into the experiment
and reach the same level as the situation in which both encounters
and extra conditioning were present. Experiment III, however, demon-
strated that those two situations remain significantly different.

Of course, in the situation where only the initial encounter occurred,
the highly conditioned food lump did not receive any extra condition-
ing. However, if the 7 day old larva does enter a more active phase,
it should have found that lump and remained there since it was still
more highly conditioned than its own lump. Some other mechanism(s)
must be operating.

It was noted earlier that the variances for the response of
larvae to conditioning, whether paired or isolates, decrease as
response to conditioning increases. This may indicate that less move-
ment between food lumps occurs as larvae find and stay in the con-

ditioned food. This is particularly obvious for the situation where

284

a 7 day old isolate has a choice between a plain and a conditioned
food lump. As more and more larvae find the conditioned lump, they
remain there and activity, as measured by exchanging food lumps,
ceases. In the situation where a 13 day old larva was introduced with
the 7 day old larva, we find that about 60% of the 7 day old larvae
initially choose the conditioned and 40% the plain lump. However,
perhaps there is movement between the two lumps. That is, there may
be a constant exchange between the two food lumps. This would not
be the case if the 7 had been isolated for the whole experiment and
the hypothesis is that the initial encounter with the 13 day old
larva may have in some way altered the activity of the 7 day old
larva and that a second stimulus, i.e., extra conditioning, a second
encounter with the 13, or both is needed to drive the 7 day old
response up to the level where it would have been if it had been an
isolate to begin with. Long's (1955) data tends to support this postu-
lation. He demonstrated that young larvae are relatively less active
than older larvae of the large white butterfly and that older larvae
will disrupt the activity rhythm of a young feeding group.

Based on the foregoing considerations the following system is
being postulated for future testing:

1. Young larvae are initially less active, as measured by
exchanging food lumps, than older larvae but their activity
increases as they age. This was demonstrated in Chapter IV
and is also evident from my data sheets with young isolate
larvae in a situation with two initially plain food lumps.

The response in such a situation is always random over time,

285

but about 5 days into the experiment, many of these isolates

switch food lumps and remain there.

2. High degrees of conditioning are a preferred stimulus and
when encountered activity decreases. This was also earlier
demonstrated and is supported by examining the variances for
such situations. The variances decrease as highly conditioned
food is encountered.

3. Older larvae, being more active, have two effects on younger
larvae. They depress the younger larva's response to highly
conditioned food and possibly alter its activity in such a
manner that young larvae tend to remain in the initially
chosen food lump.

4. Young larvae, which have been so affected by an older larva,
may require a second stimulus to offset their tendency to
remain in the initially chosen non-conditioned food lump and
to find and settle in the conditioned food when the older
larva is gone. Such a second stimulus may be in the nature
of a second encounter with the older larva or the presence
of extra conditioning.

There may be an inherent activity rhythm in these larvae, in
which, as they age, they become investigatory and begin sampling their
environment and moving between food lumps. If a highly conditioned
food lump is encountered they may cease this activity and settle down.
If they had been in such a conditioned food lump initially they may
sample the plain lump but return to the preferred stimulus. If young

larvae initially encounter an older larva in the open, i.e., not in a

286

conditioned lump, the effect is to alter the young larva's behavior so
that its preference for conditioned food is lowered and its activity
is affected in such a way that normally conditioned food is not a
sufficient stimulus to offset it. Therefore, many exchanges might
occur between plain and conditioned lumps. If, however, the larva
receives a second stimulus in the form of a second encounter with the
older larva or extra conditioning, or both, its activity may be

offset and it settles in the preferred highly conditioned lump.

This hypothesis is partly supported by my data, but much of
it is only circumstantially supported and several tests are needed,
all of which relate to an individual's behavior and activity. The
fact that activity rhythms do differ among different ages of larvae
is tentatively supported by the data in Chapter IV. However, high
conditioning is a preferred stimulus and even active larvae will
settle in it, although they may continue to sample their surroundings.
The experiments in this paper support the idea that older larvae
alter the behavior of younger larvae and lower the young larva's
response to conditioning. However, the hypothesis that a second
stimulus, in the form of extra conditioning or a second interaction
with the older larva, is needed to increase the younger larva's
response to conditioning is not well supported. An increased sample
size may be needed to pull these relationships apart.

One of the major problems is that we cannot be sure that a
second encounter with the older larva actually occurs, and if it
does, whether it has any effect. This is a function of the experi-

mental design. The first encounter occurs in the center of the dish

287

when both larvae are introduced into the apparatus. If the second
encounter occurs, however, it would have to occur in the food lump

in which the younger larva is residing, and cause it to leave that

food lump and move to the highly conditioned one. Yet in Experiment 11
the data indicates that a larva's preference for conditioned food in
which an older larva is residing is unaffected by that older larva.

The situation in which an older larva moves to conditioned food con-
taining a young resident was not tested and we do not know what the
effect on either larva would be. This needs to be done.

We also do not know if any interaction occurs between two
larvae in a conditioned food lump. Such a situation may buffer any
interactions demonstrated to occur in the open (in the middle of the
dish) and it is in fact, possible for two larvae not to encounter
each other, at least directly, in a food lump. Tests are needed in
which the experimental situation has been so altered as to maximize
the chances of two larvae meeting in the food lumps. One possibility
is to decrease the size of the food lumps. A better approach might
be to so construct the conditioned food lump that only enough food is
available for the survival of one larva and ask the question which
larva is ejected.

The other part of the hypothesis that is not well supported
is that extra conditioning may be a required stimulus to offset the
lowered activity of the younger larva after an initial encounter with
the older larva. This may be easily tested by the following compari-
son. In the final experiment of this paper group (l-O-O) received an

initial encounter with the older larva but no extra conditioning or

288

second encounter. It was found that the 7 day old response curve does
not significantly rise. That is, even when the older larva is removed
the younger larva does not move to the conditioned food. This group
should be rerun, but after about 4 days an even more highly conditioned
lump would be added to the situation. If extra conditioning is a
factor in offsetting the increased activity, the younger larvae should
move to this new lump and remain there.

Finally we need more information about the general activity
of these larvae. For example, the initial encounter described is
between a 7 and a 13 day old larva, but the second encounter, if it
is occurring, is between a 16 or 17 day old larva (the original 13
day old larva) and an 11 or 12 day old larva (the original 7 day old
larva). The two situations may be totally different as to the effects
on each larva. The effect of the initial encounter and the second
encounter may be a function of the size difference between test larvae.
This may be tested by systematically decreasing the age differences
between test larvae.

The variability encountered in these experiments also needs
to be investigated. Variability may be environmental, genetic, or
both. Two things need to be done. First, is to begin manipulating
the larva's environment to see what happens to the variability. For
example, there were no controls in my experiments for the fact that
some larvae were reared in more dense colony groups than were others
or that older larvae in the colony are subjected to higher degrees
of conditioning than were younger larvae. These two factors may have

an effect on the activity levels and preference for varying degrees

289

of conditioning. If all larvae were reared as isolates before testing,
this may result in different levels of variability.

A second approach is to artificially select for a "standard"
population of Galleria so that the genetics of my experimental animal
is under control. As an extension of this I would also like to
select for a population of larvae that do not spin silk and begin
asking questions about whether the behavioral response to conditioning
is genetically related to the physical act of spinning.

Once these variables are under control, the next step would
be to make predictions about the behavior of larger larval populations,

and experimentally test the predictions.

CHAPTER VI

SUMMARY

The studies in this dissertation deal with the preference

behavior of individual Galleria mellonella larvae for homotypically

 

conditioned food, its relationship to spatial distribution, and the
effects of degree of conditioning, developmental age, and presence

of conspecifics on these preferences. A mechanism is postulated for
the interaction of these variables with larval activity in determining
spatial patterns. Also presented is an analytical technique for the
analysis of multiple samples taken from the same individuals over

time.

Isolate Behavior

 

1. Response of isolate larvae to homotypic conditioning is either
random or attraction, depending on the degree of conditioning
used.

2. Within a particular age of test larvae, response to condition-
ing increases as the degree of conditioning increases up to
a certain level; the degree of conditioning being measured
in terms of the number of days conditioning occurs, the age of

the conditioning larva, or both. A threshold effect is

290

291

indicated, in which degrees of conditioning above a certain
level do not elicit a greater response.

There appears to be an age difference in response to con-
ditioning. Older larvae exhibit a greater response than
younger larvae, at least initially, to the same degree of
conditioning. However, these age differences cannot be
separated from early experience differences of the larvae
with colony density and colony conditioning, or from length
of experience with the experimental situation.

Age differences in initial response to low versus high
degrees of conditioning are more pronounced between young and
old larvae. Older larvae exhibit higher discrimination between
degrees of conditioning than do younger larvae.

A difference in activity, as measured by larval movement
between food lumps, between young and older larvae was
indicated. Both exhibit increased movement to conditioned
food as conditioning increases, but older larvae do most such
movement in the earlier periods of the experiment and young
larvae move more 3 or 4 days later.

Activity differences are supported by the variance vectors
for each group. As response to conditioning increases, the
variances decrease, whereas the variances remain large when
larval response to low conditioning is low. This indicates
that more movement between food lumps occurs when one of the
lumps is non-conditioned and the other has a low degree of

conditioning. As conditioning increases, larvae select the

10.

292

conditioned food and tend not to continue moving between food
lumps.

Larval movements and preference for conditioning are affected
by food source. Larvae exhibit an initial preference for
deficient—conditioned food but move to normal non-conditioned
food within 24 hours. The initial response to deficient-
conditioned food is also initially lower than to normal-
conditioned food. This may be a function of the conditioning
itself or the food source.

Female (adult) Galleria avoid highly, larval-conditioned

food for egg laying purposes and this was interpreted in
terms of an ability to interpret high conditioning as high
larval population and a behavior of laying eggs in situations
where larval density is low.

Interactions Between Conspecifics and
Biological Conditioning_

 

 

Pairs of larvae initially distribute themselves randomly

when placed in a situation where no previous conditioning has
occurred. With time, however, significant aggregation occurs
between 7+7, 7+ll, or ll+ll day old larvae. Whenever a 13

day old larva is present, significant pairing does not occur.
Aggregation behavior is not due to some discrepancy in food
lumps but is due to either an interaction between larvae,
biological conditioning, or both. For example, pairing in

the 7+ll day old situation is due to the 7 day old larva moving

to the more highly conditioned food lump where the 11 day old

11.

12.

13.

14.

293

larva is residing. Similar movements do not occur to where

a 13 day old larva is residing, probably a function of larval-
larval intolerance, not conditioning.

When one of the food lumps is highly conditioned, aggregation
behavior occurs in that lump, regardless of the age of the two
test larvae. However, the response of individual larvae to
the conditioning is affected by the presence of a conspecific.
If a conspecific is a resident in the conditioned food, the
behavior of test larvae is the same as if the conspecific
were not present. For example, a 7 day old larva's response
to a conditioned food lump with a resident 13 day old larva
in it is the same as a 7 day old's response to a conditioned
food lump lacking a resident. This behavior is the same whether
the food was low or high in conditioning.

When both test larvae are introduced into the experimental
situation at the same time, their responses to conditioning
are lowered if conditioning is high and unaffected if the
conditioning is low, as measured against the response of
isolates to similar degrees of conditioning. When response
is affected, the response of the younger larva is more
drastically lowered than is that of the older larva, but
eventually increases with time.

The younger larva's response was manipulated by varying the
age of the older conspecific and it was postulated that the
behavior of the younger larva is a result of three variables:

(1) the initial encounter with the older larva, (2) the extra

15.

l6.

17.

294

conditioning being done by the older larva, and (3) a second
encounter with the older larva.

Experiments utilizing various combinations of these variables
demonstrated that the initial encounter was the factor involved
in the lowered initial preference for conditioning, but the
effects of extra conditioning and a second interaction were
inconclusive. An hypothesis was advanced relating the inter-
action between biological conditioning, larval-larval inter-
actions, and larval activity rhythms to the observed spatial
patterns of Galleria larvae and various tests proposed. A
larger sample size may be needed to elucidate these variables
Such a sample size may be calculated by N = 3/G(1-G), as dis-
cussed in Chapter III.

The general conclusion from all of these studies is that indi—
vidual behavior and distribution is a function of degree of
conditioning, age of the individual, presence and age of a
conspecific, location of the conspecific, and possibly the
activity of Galleria larvae.

Hotelling's One- and Two-Sample T2 Statistic for Multivariate
Analyses was utilized for the analysis of the data, with
modifications for handling longitudinal data on individuals.
This proved to be a very effective way of handling data on
individual animals and for comparing the total response of

organisms over time.

BIBLIOGRAPHY

BIBLIOGRAPHY

Adolph, E. F. 1931. The size of the body and the size of the environ-
ment in the growth of tadpoles. Biol. Bull. 61:350-375.

Alexander, Nancy J. 1970. A regulatory mechanism of ecdysone
release in Galleria mellonella. J. Insect Physiol. 16:
277-289.

 

Alexander, P. and D. H. R. Barton. 1943. The excretion of ethyl-
quinone by the flour beetle. Biochem. Jour. 37(4):463-465.

Allee, W. C. 1926. Studies in animal aggregations: causes and
effects of bunching in land isopods. J. Exp. Zool. 45:
255-277.

Allee, W. C. 1927. Animal aggregations. Quart. Review Biol. 2:
367-398.

Allee, W. C. and J. F. Schuett. 1927. Studies in animal aggregations:
the relation between mass of animals and resistance to
colloidal silver. Biol. Bull. 53:301-317.

Allee, W. C. 1931. Animal Aggregations: A Study in General
Sociology. Chicago: University of Chicago Press.

 

Allee, W. C. and J. R. Fowler. 1932. Studies in animal aggregations:
further studies on oxygen consumption and autotomy of the
brittle star, Ophioderma. J. Exp. Zool. 64:1-19.

 

Allee, W. C. 1934. Recent studies in mass physiology. Biol. Rev.
9:1-48.

Allee, W. C., E. S. Bowen, J. C. Welty, and R. Oesting. 1934. The
effect of homotypic conditioning of water on the growth of
fishes, and chemical studies of the factors involved. J.
Exp. Zool. 68:183-213.

Allee, W. C., A. J. Finkel, and W. H. Hoskins. 1940. The growth of

goldfish in homotypically conditioned water: a population
study in mass physiology. J. Exp. 2001. 84:417—443.

295

296

Allee, W. C., Alfred E. Emerson, Orlando Park, and Thomas Park. 1949.
Principles of Animal Ecology, Philadelphia and London: W. B.
Saunders Company.

 

Andrews, J. E. 1921. Some experiments with the larva of the bee
moth, Galleria mellonella L. Trans. Wis. Acad. Sci., Arts
and Letters 20:255-261.

 

Asher, James H. 1970. Parthenogenesis and genetic variability.
Ph.D. Dissertation. University of Michigan. Ann Arbor,
Michigan.

Balazas, A. 1958. Nutritional and nervous factors in the adaptation
of Galleria mellonella to artificial diet. Acta. Biol.
9:47-69.

 

Beck, Stanley D. 1960. Growth and development of the greater wax
moth, Galleria mellonella (L.). (Lepidoptera: Galleriidae).
Trans. Wis. Acad. Sci., Arts and Letters 49:137-148.

 

Bell, J. 1940. The heat production and oxygen consumption of

Galleria mellonella at different constant temperatures.
Physiol. Zool. 13573-81.

 

Borchert, A. 1933. Zur Biologie de grossen Wachsmotte (Galleria
mellonella). I. Uber morphologie und Entwicklungsdauer der
larvae der grossen Wachsmotte. Zool. Jahr. Abteil. Anat.

u. Ont. 57:105-115.

 

Bossert, William H. and Edward O. Wilson. 1963. The analysis of
olfactory communication among animals. J. Theoret. Biol.
5:443-469.

Bovbjerg, Richard. 1952. Ecological aspects of dispersal of the
snail Campeloma decisum. Ecology 33:169-176.

 

Bowen, E. S. 1921. The role of the sense organs in aggregations of
Ameiurus melas. Ecol. Mono. 1:1-35.

 

Box, G. E. P. 1950. Problems in the analysis of growth and wear
curves. Biometrics 6:362-389.

Brand, R. A. 1955. Abundance and activity of the woodmouse
(Peromyscus leucopus novebogacensis) in relation to the
characteristics of its habitat. Ph.D. Thesis. University of
Michigan.

 

Braun. 1867. The bee moth. Amer. Bee Jour. 66:380-400.

Breukelman, John. 1932. Speed of toxic action of mercuric chloride
on daphnids. Physiol. 2001. 5:207—218.

297

Brindley, H. H. 1910. Further notes on the procession of Cnethocampa

 

pinivora. Proc. Camb. Phil. Soc. 15:576-587.

Brown, L. A. and A. M. Banta. 1935. Control of sex in Cladocera.
Physiol. Zool. 8:138-155.

Brown, M. D. 1946. The growth of brown trout (Salmo trutta Linn.).

I. Factors influencing the growth of trout fry. J. Exp.
Biol. 22:118-144.

 

Brown, Margaret E. 1946. The growth of brown trout (Salmo trutta
Linn.). I. 1. Factors influencing the growth of trout fry.
J. Exp. Biol. 22:118-129.

 

Brown, Margaret E. 1951. The growth of brown trout (Salmo trutta
Linn.). IV. The effect of food and temperature on the
survival and growth of fry. J. Exp. Biol. 28:473-491.

 

Brown, R. Z. 1953. Social behavior, reproduction, and population
changes in the house mouse (Mus musculus L.). Ecol. Monogr.
23:217-240.

 

Burkett, B. N. 1962. Temperature block of spinning and deve10pment
in Galleria mellonella (Lepidoptera). Amer. 2001. 2:396.

 

Butler, C. G., D. J. C. Fletcher, and Doreen Watler. 1969. Nest-
entrance marking with phermones by the honeybee, Apis
mellifera L., and by a wasp, Vespula vulgaris L. Anim.
Behav. 17:142-147.

 

Calhoun, J. B. 1948. Mortality and movement of brown rats (Rattus
norvegicus) in artificially super-saturated populations.
Jour. Wildl. Mgt. 12:167-172.

 

Calhoun, J. B. 1950. The study of wild animals under controlled
conditions. Annals New York Acad. Sci. 51:1113-1122.

Calhoun, J. B. 1952. The social aspects of population dynamics.
Jour. Mamm. 33:138-159.

Calhoun, J. B. and W. L. Webb. 1953. Induced emigrations among small
mammals. Science 117:358-360.

Carthy, J. D. 1951. The return of ants to their nest. Trans.
9th Int. Congr. Entom. 1:365-369.

Chapman, R. N. 1928. The quantitative analysis of environmental
factors. Ecology 9:111-122.

Chase, Ruth W. 1921. The length of life of the larva of the wax
moth, Galleria mellonella L., in its different stadia. Trans.
Wis. Acad. Sci., Arts and Letters 20:263-267.

 

298

Christian, J. J. 1950. The adren-pituitary system and population
cycles in mammals. J. Mamm. 31:247-259.

Christian, J. J. 1970. Social subordination, population density,
and mammalian evolution. Science 168:84-90.

Chiang, H. C. and A. C. Hodson. 1950. An analytical study of popu-

lation growth in Drosophila melanogaster. Ecol. Monogr.
20:173-206.

 

Chitty, D. 1955. Adverse effects of population density upon the
viability of later generations. In the numbers of man and
animals. Edited by J. B. Cragg and N. W. Pirce. Edinburgh:
Oliver and Boyd, Ltd., 57-67.

Chitty, D. 1960. Population processes in the vole and their rele-
vance to general theory. Can. J. Zool. 38:99-113.

Cochran, W. G. 1947. Some consequences when the assumptions for the
analysis of variance are not satisfied. Biometrics 3:22-38.

Cole, H. A. and E. W. Knight-Jones. 1949. The settling behavior of
larvae of the European flat oyster, Ostrea eduli L., and its
influence on methods of cultivation and spat collection.
Fishery Invest. London, ser 2. l7(3):59 pp.

 

Cole, L. C. 1946. A study of the dryptozoa of an Illinois' woodland.
Ecol. Monogr. 16:49-86.

Collias, Nicholas E. 1964. The evolution of nests and nest building
in birds. Amer. 2001. 4:175-190.

Crombe, A. C. 1942. The effect of crowding upon the oviposition of
grain-infesting insects. J. Exper. Biol. 19:311-340.

Dadd, R. H. 1964. A study of carbohydrate and lipid nutrition in
the wax moth, Galleria mellonella (L.), using partially syn-
thetic diets. J. Insect Physiol. 10:161-178.

 

Dadd, R. H. 1966. Beeswax in the nutrition of the wax moth, Galleria
mellonella (L.). J. Insect Physiol. 12:1479-1492.

 

Davis, D. E. 1958. The role of density in aggressive behavior in
house mice. Anim. Behav. 6:207-211.

Davis, D. E. and J. J. Christian. 1956. Changes in Norway rat popu-
lations induced by introduction of rats. J. Wildl. Mgt. 20:
378-383.

Dawson, P. S. 1964. An interesting behavioral phenomenon in Tribolium
confusum. Tribolium Inform. Bull. 7:50-52.

299

Dickman, A. 1933. Studies on the wax moth Galleria mellonella with
particular reference to the digestion of wax by the larvae.
J. Cell and Comp. Physiol. 3:223-246.

 

Doull, K M. 1953. Phase coloration in lepidopterous larvae. London.
Nature 172:813-814.

Dryl, S. 1963. Contributions to mechanism of chemotactic response
in Paramecium caudatum. Anim. Behav. 11:393-396.

 

Drzewina, A. and G. Bohn. 1921. Variations de la susceptibilite

aux agents mocifs avec le nombre des animaux traites. C. R.
Acad. Sci. 172:485-487.

Drzewina, A. and G. Bohn. 1926. Activation par la lumiere des effets
de l'argent sur convoluta. C. R. Acad. Sci. 183:677-679.

Durbin, J. and G. 8. Watson. 1950. Testing for serial correlation
in least squares regression. I. Biometrika 37:409-428.

Durbin, J. and G. S. Watson. 1951. Testing for serial correlation
in least squares regression. II. Biometrika 38:159-178.

Dutky, S. R., J. V. Thompson, and George E. Cantwell. 1962. A

technique for mass rearing the greater wax moth. Proc. Ent.
Soc. Wash. 64:56-58.

Edmondson, W. T. 1945. Ecological studies of sessile rotatoria.

II. Dynamics of populations and social structures. Ecol.
Monogr. 15:143-172.

Edney, E. B. 1954. Woodlice and the land habitat. Biol. Rev. 29:
185-219.

Edwards, J. S. 1966. Neural control of metamorphosis in Galleria
mellonella (Lepidoptera). J. Insect Physiol. 12:1423-1433.

 

Edwards, T. G. 1910. On the procession and pupation of the larva of
Cnethocampa pinivora. Proc. Camb. Phil. Soc. 15:431-436.

 

Ellis, P. E. 1953. Social aggregation and gregarious behaviors in

hoppers of Locusta migratoria migratorivides. (R. and F.).
Behavior 5:225-260.

 

El-Sawaf, S. K. 1950. The life history of the greater wax moth
(Galleria mellonella L.) in Egypt, with special reference to
the morphology of the mature larva. Bull. Soc. Fouad. Ent.
34:247-297.

 

Faure, J. C. 1932. The phases of locusts in South Africa. Bull.
Ent. Res. 23:293-428.

300

Fitch, John H. 1975. Behavioral habitat selection as a factor in
the distribution of woodland deermice, Peromyscus maniculatus
gracilis, in the Kingston Plains of Northern Michigan. Ph.D.
Dissertation. Michigan State University.

 

Fraenkel, G. S. and D. L. Gunn. 1961. The Orientation of Animals.
New York: Dover Publications.

 

Free, J. B. 1961. The stimuli releasing the stinging response of
honeybees. Anim. Behav. 9(3-4):193-l96.

Friedlander, C. P. 1964. Thigmokinesis in woodlice. Anim. Behav.
12:164-174. .

Gee, J. M. 1965. Chemicals stimulation of settlement in larvae of
Spirorbis rupestris (Serpulidae). Anim. Behav. 13(1):181-186.

 

Ghent, Arthur W. 1963. Studies of behavior of the Tribolium flour
beetles. I. Contrasting responses of I: castaneum and I,
confusum to fresh and conditioned flour. Ecology 44(2):
269-283.

Ghent, Arthur W. 1966. Studies of behavior of the Tribolium flour
beetles. II. Distributions in depth of I, castaneum and I,
confusum in fractionable shell vials. Ec010gy 47(3):355-367.

Gill, J. L. and H. D. Hafs. 1971. Analysis of repeated measurements
of animals. J. Anim. Sci. 33:331-336.

Graham, W. M. 1958. Temperature preference determination using
Tribolium. Anim. Behav. 6:231-237.

Gunn, D. L. and D. P. Prelou. 1940. The humidity behavior of the
mealworm beetle, Tenebrio molitor. III. The mechanism of
the reaction. J. Exp. Biol. 17:307-316.

 

Harris, V. T. 1956. An experimental study of habitat selection by
prairie and forest races of the deer mouse. Peromyscus
maniculatus. Contrib. Lab. Vert. Biol. Univ. of Michigan
56:53 pp.

 

 

Harshbarger, John C. and B. Shain Moore. 1966. Some effects of x-
irradiation on larvae of Galleria mellonella. J. Invert. Path.
8(2):277-279.

 

Haydak, Mykola H. 1936. Is wax a necessary constituent of the diet
of wax moth larvae? Ann. Ent. Soc. Amer. 29:581-588.

Haydak, Mykola H. 1940. The length of development of the greater
wax moth. Science 91:525.

301

Heatwole, Harold. 1962. Environmental factors influencing local
distribution and activity of the salamander, Plethodon
cinereus. Ecology 43:460-472.

Hotelling, Harold. 1931. The generalization of student's ratio.
Ann. of Math. Stat. 2:360-378.

Howard, Eliot. 1920. Territory in Bird Life. New York: Atheneum.

 

Imms, A. D. 1948. A General Textbook of Entomology. New York:
Dutton.

 

Issakowitsch, A. 1905. Geschlechtsbestimmende ursachen bei den
daphniden. Biol. Centralbl. 25:529-536.

Ito, Koichi and William Schull. 1964. On the robustness of the T 2
test in multivariate analysis of variance when variance-
covariance matrices are not equal. Biometrika 51:71-82.

Jacobson, Martin, Meyer Schwarz, and Rolland M. Water. 1970. Gypsy
moth sex attractants: A reinvestigation. J. Econ. Ent. 63(3):
943-945.

Jespersen, P. 1924. The frequency of birds over the high Atlantic
Ocean. London. Nature 114:281-283.

Johnson, R. P. 1973. Scent marking in mammals. Anim. Behav. 21:
521-535.

Johnston, James W., David G. Moulton, and Amos Tuck. 1970. Communi-
cation by Chemical Signals, Vol. I. New York: Appleton-
Century-Crofts.

 

Karlson, P. and A. Butenandt. 1959. Phermones (ectohormones) in
insects. Ann. Rev. Entom. 4:39-58.

Kaston, B. J. 1964. The evolution of spider webs. Amer. 2001.
4:191-207.

Knight-Jones, E. W. and J. P. Stevenson. 1951. Gregariousness during

settlement in the barnacle, Elminuis modestus Darwin. J.
Mar. Biol. Ass. U. K. 29:281-297.

 

Krebs, C. J. 1966. Demographic changes in fluctuating populations
of Microtus californicus. Ecol. Monogr. 36:239-273.

 

Krebs, Charles J., Barry L. Keller, and Robert H. Tamarin. 1969.
Microtus population biology: Demographic changes in fluctuating

populations of M, ochrogaster and M, pennsylvanicus in Southern
Indiana. Ecology 50(40):587-607.

 

 

302

Lack, David. 1954. The Natural Regulation of Animal Numbers.
Oxford: Clarendon Press.

 

Legay, J. M. and J. L. Chasse. 1964. Etude experimental de la
sortie d'une enciente chez Tribolium confusum. Acad. Sci.
(Paris), Compt. rend. 258:717-719.

 

Lerner, I. M. and N. Inouze. 1968. Behavior genetics with special
reference to maze—running of Tribolium. In Dronamraju K. R.
(ed.). Haldane and Modern Biology. Johns Hopkins Press.

Loconti, Joseph D. and Louis M. Roth. 1953. Composition of the
odorous secretion of Tribolium castaneum. Ent. Soc. of
Amer., Annals 46:281-289.

 

Loomis, W. F. and H. M. Lenhoff. 1956. Growth and sexual differ-
entiation of hydra in mass culture. J. Exp. 2001. 132:555-568.

Long, D. B. 1953. Effects of population density on larvae of
Lepidoptera. Trans. R. Ent. Soc. Lond. 104:544-585.

Long, D. B. 1955. Observations on sub-social behavior in two species
of Lepidoptera larvae, Pieris brassicae L. and Plusia gamma
L. Trans. R. Ent. Soc. Lond. 106:421-437.

 

 

Loschiavo, S. R. 1952. A study of some food preferences of
Tribolium confusum Duv. Cereal Chemistry 29:91-107.

 

Lwengood, W. F. 1937. An experimental analysis of certain factors
affecting growth of goldfishes in homotypically conditioned
water. Copeia. 81-88.

Magnuson, John J. 1962. An analysis of aggressive behavior, growth,
and competition for food and space in Medaka (Oryzias latipes)
(Pisce: Cryprinodontidae). Canadian J. of 2001. 40:313-363.

 

Maschwitz, Ulrich W. 1964. Alarm substances and alarm behavior in
social hymenoptera. Nature 204:324-327.

McDonald, D. J. and L. Fitting. 1965. Activity differences in black

and wild type strains of Tribolium confusum. Tribolium Inform.
Bull. 8:124-125.

 

McDonald, Daniel J. 1968. Mobility in Tribolium confusum. Ecology
49:770-771.

 

Milum, V. G. 1952a. Characters and habits of moth larvae infesting
honeybee combs. Amer. Bee Jour. 92:200-220.

Milum, V. G. 1952b. Descriptions and habits of the adults of some
moths whose larvae infest combs of honeybees. Amer. Bee
Jour. 92:155-156.

303

Milum, B. G. and H. W. Geuther. 1935. Observations on the biology
of the greater wax moth, Galleria mellonella L. J. Econ.
Ent. 28:576-578.

 

Morrison, Donald F. 1967. Multivariate Statistical Methods. New
York: McGraw-Hill Book Company.

 

Myers, Judith and Charles J. Krebs. 1974. Population cycles in
rodents. Sci. Amer. 230:38-46.

Naylor, Alfred F. 1959. An experimental analysis of dispersal in
the flour beetle, Tribolium confusum. Ecology 40:453-465.

 

Naylor, Alfred F. 1961. Dispersal in the red flour beetle Tribolium
castaneum (Tenebrionidae). Ecology 42(2):231-237.

Naylor, Alfred F. 1965. Dispersal responses of female flour beetles,
Tribolium confusum, to presence of larvae. Ecology 46:
341-343.

 

Niemierko, W. and S. Cepelewicz. 1950. Studies in the biochemistry
of the wax moth (Galleria mellonella). 1. Growth of the
larvae and their chemical composition. Acta Biol. Exp.
15(4):57-68.

 

Niemierko, W. and P. Wlodawer. 1950. ...2. Utilization of wax
constituents by the larvae. Acta Biol. Exp. lS(5):69-78.

Niemierko, W. and P. Wlodawer. 1952. ...7. The digestion of wax
and utilization of unsaponifiable substances by larvae.
Acta Biol. Exp. 16(12):157-l70.

Niemierko, W. and L. Wojtczak. 1950. 3. Oxygen consumption of the
larvae during starvation. Acta Biol. Exp. 15(6):79-90.

Oertel, E. 1962. Behavior studies of the greater wax moth. XI
Intern. Kongress fur. Entomologie, Wien. 532-536.

Oertel, E. 1963. Greater wax moth develops on bumble bee cells.
J. Econ. Ent. 56(4):543-544.

Ogden, John C. 1969. Effect of components of conditioned medium on
behavior in Tribolium confusum. Physiol. Zool. 42:266-274.

 

Ogden, John C. 1970. Aspects of dispersal in Tribolium flour beetles.
Physiol. Zool. 43:124-131.

Ogden, John C. 1970b. Artificial selection for dispersal in flour
beetles (Tenebrionidae: Tribolium). Ecology 51:130-133.

Orgain, H. and M. W. Schein. 1953. A preliminary analysis of the
physical environment of the Norway rat. Ecology 34:467-473.

304

Paddock, F. B. 1913. The life history and control of the bee-moth
or wax-worm. Texas Agri. Exper. Station Bull. 158:15-30.

Paddock, F. B. 1930. The bee moths. J. Econ. Ent. 23:422-428.
Park, Thomas. 1932. Studies in population physiology. The relation

of numbers to initial population growth in the flour beetle
Tribolium confusum. Ecology 13:172-181.

 

Park, Thomas. 1933. ...11. Factors regulating initial growth of
Tribolium confusum populations. J. Exp. Zool. 65:17-42.

 

Park, Thomas. 1934. ...III. The effect of conditioned flour upon
the productivity and population decline of Tribolium confusum.
J. Exp. Zool. 68(2):l67-182.

 

Park, Thomas. 1935. ...IV. Some physiological effects of conditioned
flour upon Tribolium confusum Duval and its populations.
Physiol. Zool. 8:91-115.

 

Park, Thomas. 1936a. ...V. The oxygen consumption of the flour
beetle, Tribolium confusum Duval. J. of Cellular and Comp.
Physiol. 7(3):3l3-323.

 

Park, Thomas. 1936b. ...VI. The effect of differentially conditioned
flour upon the fecundity and fertility of Tribolium confusum
Duval. J. Exp. 2001. 73(3):393-404.

 

Park, Thomas and Woollcott. 1937. ...VII. The relation of environ-
mental conditioning to the decline of Tribolium confusum
populations. Physiol. Zool. 10:197-211.

 

Park, Thomas. 1938. Studies in population physiology. VIII. The
effects of larval population density on the post-embryonic
development of the flour beetle T. confusum. J. Exp. Zool.
79:51-70. '-

Park, Thomas, E. V. Miller, and C. 2. Lutherman. 1939. ...IX. The
effect of imago population density on the duration of the

larval and pupal stages of Tribolium confusum Duval. Ecology
20:365-373.

 

Park, Thomas and W. Burrows. 1942. The reproduction of Tribolium
confusum Duval in a semisynthetic wooddust medium. Physiol.
Zool. 15:476-484.

Park, Thomas. 1948. Experimental studies of interspecies competition.
1. Competition between populations of the flour beetles,
Tribolium confusum Duval and Tribolium castaneum Herbst.

Ecol. Monogr. 18:265-307.

 

 

305

Parker, J. S. 1930. Some effects of temperature and moisture upon
Melanoplus mexicanus saussure and Camnula pellucida Scudder
(Orthoptera). Univ. Montana Agri. Expt. Sta. Bull. No. 223.

 

 

Pearl, Raymond. 1927. The growth of populations. Quart. Rev.
Biol. 2:530-548.

Peebles, Florence. 1929. Growth regulating substances in echinoderm
larvae. Biol. Bull. 57:176-187.

Pfeiffer, W. 1963. Alarm substances. Experientia. 19:113-168.

Pipa, R. L. 1963. Studies on the hexapod nervous system. VI. Ventral
nerve cord shortening; a metamorphic process in Galleria
mellonella (L.) (Lepidoptera, Pyrallidae). Biol. Bull.

Woods Hole. 124:293-302.

 

Plath, O. E. 1934. Bumblebees and Their Ways. New York: Macmillan.

 

Prus, T. 1963. Search for methods to investigate mobility in
Tribolium. Ecology 44:801-803.

Przelecka, A. 1956. Studies on the biochemistry of wax moth
(Galleria mellonella L.). 14. Cytochemical study of phos-
pholipids in the intestinal tract of wax moth larvae. Acta
Biol. Exp. l7(l):23l-234.

 

Rathcke, Beverly J. and Robert W. Poole. 1975. Coevolutionary race
continues: butterfly larval adaptation to plant trichomes.
Science 187:175-176.

Regnies, Fred E. and John H. Law. 1968. Insect phermones. J. of
Lipid Res. 9:541-551.

Reynierse, James H. 1967. Aggregation formation in planaria,
Phagocata gracilis and Cura foremani: species differentiation.
Anim. Behav. 15:270-272.

 

 

Reynierse, James H., Kathryn K. Gleason, and Robert Ottermann. 1969.

Mechanisms producing aggregations in planaria. Anim. Behav.
l7(l):47-63.

Rich, Earl R. 1956. Egg cannibalism and fecundity in Tribolium.
Ecology 37:109-120.

Richards, C. M. 1958. The inhibition of growth in crowded Rana
pipiens tadpoles. Physiol. 2001. 31:138-151.

 

Robertson, T. B. 1921. Experimental studies on cellular manipulation.
II. The influence of mutual antiquity upon reproductive
rate and the part played therein by the "x-substance" in
bacterized infusions which stimulates the multiplication of
infusoria. Biochem. Jour. 15:595-611, 612-619.

306

Rohlf, James F. and Robert R. Sokal. 1969. Statistical Tables.
San Francisco: W. H. Freeman and Company.

 

Rose, S. Meryl. 1959. Failure of survival of slowly growing members
of a population. Science 129:1026.

Rose, S. M. 1960. A feedback mechanism of growth control in tadpoles.
Ecology 41:188-199.

Roth, Louis M. and Ruth B. Howland. 1941. Studies on the gaseous
secretion of Tribolium confusum Duval. I. Abnormalities
produced in Tribolium confusum Duval by exposure to a secretion
given off by the adults. Ann. Ent. Soc. Am. 34(1):151-l75.

 

 

Roth, Louis M. 1943. Studies on the gaseous secretion of Tribolium
confusum Duval. II. The odoriferous glands of Tribolium
confusum. Ent. Soc. of Amer., Annals 36:397-424.

Salt, George and F. S. J. Hollick. 1946. Studies of wireworm popu-
lations. 11. Spatial distribution. J. Exp. Biol. 23:1-46.

Sharp, David T. and I. B. Gray. 1962. Studies on factors affecting
the local distribution of two sea urchins, Arbacia punctulata
and Lytechinus variegatus. Ecology 43:309-313.

 

 

Shaw, Gretchen. 1932. The effect of biologically conditioned water
upon rate of growth in fishes and amphibia. Ecology 13:
263-278.

Shelford, V. E. 1913. The reactions of certain animals to gradients
of evaporating power of air. A study in experimental ecology.
Biol. Bull. 25:70-120.

Silverstein, Robert M., Otto J. Rodin, Wendell E. Burkholder, and
John E. Gorman. 1967. Sex attractant of the black carpet
beetle. Science 157:85-86.

Silverstein, Robert M., Robert G. Brownlee, Thomas E. Bellas,
David L. Wood, and Lloyd E. Browne. 1967. Brevicomin:
Principal sex attractant in the frass of the female pine
beetle. Science 159:889-890.

Skaife, S. H. 1954. African Insect Life. London: Longmans Green.

 

 

Sladen, F. W. L. 1912. The Bumble-Bee. London: Macmillan.

Smith, G. 1915. The life-cycle of Cladocera, with remarks on the
physiology of growth and reproduction in crustacea. Proc.
Roy. Soc. London B. 88:418-535.

Smith, Stephen D. 1969. The effects of crowding on larvae of the
meal moth, Ephestia kuhniella. J. Exp. Zool. 170:193-203.

 

307
Skal, Robert R. and James F. Rohlf. 1969. Biometry. San Francisco:
W. H. Freeman and Company.

Sokoloff, A. 1966. A behavioral difference of two mutant strains
in Tribolium castaneum. Tribolium Inform. Bull. 9:109-110.

 

Stanley, J. 1932. A mathematical theory of the growth of populations
of the flour beetle Tribolium confusum Duval. Can. J.
Research 6:632-671.

 

Stanley, John. 1949. A mathematical theory of the growth of popu-
lations of the flour beetle, Tribolium confusum Duv. VII.
A study of the "Re-tunneling Problem." Ecology 30(2):
209-222.

 

Stickle, L. F. 1946. The source of animals moving into a depopu-
lated area. J. Mamm. 27:301-307.

Strecker, R. L. 1954. Regulatory mechanics in house mouse popu-
lations--the effect of limited food supply on a confined
population. Ecology 34:375-385.

Stuart, C. A. and A. M. Banta. 1931. Available bacteria and the
sex ratio in Moina. Physiol. Zool. 4:72-86.

Surtees, Gordon. 1963a. Laboratory studies on dispersion behavior
of adult beetles in grain. I. The grain weevil,
Sitophilus granarius (L.) (Coleoptera, Curculionidae).
Bull. Ent. Res. 54:147-159.

 

Surtees, Gordon. 1963b. ...II. The saw-toothed grain beetle,

Oryzaephilus surinamensis (L.) (Coleoptera, Silvanidae).
Bull. Ent. Res. 54:285-296.

 

Surtees, Gordon. 1963c. ...III. Tribolium castaneum (Hbst.)
(Coleoptera, Tenebrionidae) and ngptolestes ferrugineus

4

(Steph.) (Coleoptera, Cucujidae). Bull. Ent. Res. 54:297-306.

 

Surtees, Gordon. l964a. ...VI. Three-dimensional analysis of dis-
persion of five species in a uniform bulk. Bull. Ent. Res.
55:161-171.

Surtees, Gordon. 1964b. ...VII. The effect of isolated pockets of
damp and mouldy grain on Tribolium castaneum (Hbst.)
(Coleoptera, Tenebrionidae). Bull. Ent. Res. 55:327-337.

 

Surtees, Gordon. 1964c. ...VIII. Spontaneous activity in three
species and a new approach to analysis of kinesis mechanisms.
Anim. Behav. 12:374-377.

Surtees, Gordon. 1964d. ...IX. Some differences in behavior of
mated and unmated grain weevils. Entom. Monthly Mag. 100:
105-108.

308

Surtees, Gordon. 1964e. ...XI. Some effects of temperature. Anim.
Behav. 12:378-381.

Surtees, Gordon. l964f. ...XII. The effect of isolated pockets of
damp and mouldy wheat on Cryptolestes ferrugineus (Steph.)
(Coleoptera, Cucujidae). Bull. Ent. Res. 55:673-680.

 

Talbot, Mayr. 1934. Distribution of ant species in the Chicago
region with reference to ecological factors and physiological
toleration. Ecology 15:416-439.

Taylor, Sneyd J. 1931. A case of "assembling" in the phasmidae.
Proc. Ent. Soc. London. 6:45.

Terman, C. Richard. 1961. Some dynamics of spatial distribution
within semi-natural populations of prairie deermice. Ecology
42:288-302.

Thorpe, W. H., A. C. Crombie, R. Hill, and J. H. Darrah. 1946. The
behavior of wireworms in response to chemical stimulation.
J. Exp. Biol. 23:234-267.

Uvarov, B. P. 1928. Grasshoppers and Locusts. London: Imp. Bur.
Entom. 352 pp.

 

Van der Kloot, William G. and Carroll M. Williams. 1953. Cocoon
construction by the Cecropia silkworm. I. The role of the
external environment. Behaviour 2+3zl4l-156.

Van der Kloot, William G. and Carroll M. Williams. 1953b. ...II.
The role of the internal environment. Behaviour 5:157-174.

von Frisch, Karl. 1974. ’Animal Architecture. New York and London:
Harcourt Brace Jovanovich.

 

Walrecht, B. J., Jr. 1953. Similarity of the building instinct of
social bees and wasps. Levende Natur. 61(3):70-72.

Walter, H. E. 1907. The reactions of planaria to light. J. Exp.
Zool. 5:35-162.

Warburg, M. R. 1964. The response of isopods towards temperature,
humidity, and light. Anim. Behav. 12:175-186.

Waterhouse. 1959. Axenic culture of wax moths for digestion studies.
Ann. N. Y. Acad. Sci. 77:283-289.

Watson, H. S. 1953. Serial correlation in regression analysis.
I. Biometrika. 42:327-341.

309

Way, M. J. and M. Cammell. 1970. Aggregation behavior in relation to
food utilization by aphids. In Animal Populations in Relation

 

to Their Food Resources (ed. Adam Watson). 1970. Bkackwell
Scientific Publ. Oxford and Edinburgh.

 

Wecker, Stanley C. 1963. The role of early experience in habitat
selection by the prairie deer mouse, Peromyscus maniculatus
bairdi. Ecol. Monogr. 33:307-325.

 

Wellington, W. G. 1948. The light reactions of the spruce budworm,
Choristoneura fumiferana Clemens (Lepidoptera: Tortruidae).
Canad. Ent. 80:56-82.

 

Wellington, W. G. 1957. Individual differences as a factor in
population dynamics: the development of a problem. Canad. J.
Zool. 35:293-323.

Wellington, W. G. 1960. Qualitative changes in natural populations
during changes in abundance. Canad. J. Zool. 38:289-314.

Welty, J. C. 1934. Experimental explorations into group behavior
of fishes: a study of the influence of the group on individual
behavior. Physiol. Zool. 7.

West, L. B. 1961. The nature of growth inhibitory material from
crowded Rana pipiens tadpoles. Physiol. Zool. 33:232-239.-

 

Whitcomb, Warren. 1936. The wax moth and its control. U.S. Dept.
of Agri. Circular No. 386:1-10.

Whitcomb, Warren. 1942. The wax moth and its control. U.S. Dept.
of Agri. Circular No. 386:1-11.

Williams, 0. L. 1934. Chapter 4, Termites and Termite Control.
University of California Press. Berkeley, California.

 

Witt, Peter N. and Charles F. Reed. 1965. Spider-web building.
Measurement of web geometry identifies components in a com-
plex invertebrate behavior pattern. Science 149:1190-1197.

Wojtczak, L. 1952a. Studies on the biochemistry of the wax moth
(Galleria mellonella). 10. Respiratory enzymes of the larva.
Acta Biol. Exp. 16(15):199-222.

 

Wojtczak, L. 1952b. ...ll. Respiratory enzymes in development.
Acta Biol. Exp. l6(l6):223-238.

Wojtezak, L. 1952c. ...12. Respiratory enzymes of the larva during
starvation. Acta Biol. Exp. l6(l7):239-246.

310

Woolever, Patricia and Rudolph Pipa. 1970. Spatial and feeding
requirements for pupation of last instar larval Galleria
mellonella Lepidoptera. J. Insect Physiol. 16:251-262.

 

Wynne-Edwards, V. C. 1962. Animal Dispersion in Relation to Social
Behavior. Edinburgh and London: Oliver and Boyd.

 

Young, Roger G. 1961. The effects of dietary beeswax and wax com—
ponents on the larvae of the greater wax moth, Galleria
mellonella (L.). Ann. Entom. Soc. Amer. 54:657-659.

 

Young, Roger G. 1964. Digestion of wax by the greater wax moth,
Galleria mellonella. Ann. Entom. Soc. Amer. 57:325-327.

 

Zielinska, A. M. 1952. Studies on the biochemistry of the wax moth
(Galleria mellonella). 8. Nitrogen metabolism of the larvae.
Acta Biol. Exper. 16(13):171-186.

 

Zromska-Rudyka, H. 1966. An appreciation of the laboratory set
for investigating the emigratory activity of Tribolium.
Ekologic Polska (A) 14:125-132.