PH

 

 

NE

 

 

    
   

 

      
   
 

 

 

 

 

 

 

 

 

 

 

        

  
        
   

SI

 

Michigan State
University

     

This is to certify that the ,_. a»

thesis entitled

PATTERN ANALYSIS ALONG A
PINUS-CBAMAEUAPQfiQ TRANSITION AS
KEY TO VEGETATIONAL STRUCTURE AND ORGANIZATION

 

presented by

MICHAEL DANIEL BYER

has been accepted towards fulfillment
of the requirements for

_Pl'_lD_degree in Botanz & Plant Pathology

“31

 

(7) [/1]? . .rl . "k

_/ Major professor #

II t
5

)ate 3 AugUSt 1965 1

 

 

 

    

*1va m m t m
( tWtif‘l/T‘lgm €33 ,/

 

 

 

 

PATTERN ANALYSIS ALONG A

PINUS-CHAMABDAPHEEfiTRANSITION.AS

 

 

A.KEY TO VEGEIATIONAL STRUCTURE AND ORGANIZATION

BY

Michael Daniel Byer

AN ABSTRACT OF.A THESIS

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

DOCTOR OF PHILOSOPHY

Department of Botany and Plant Pathology

1965

ABSTRACT
PATTERN ANALYSIS ALONG A

PINUS-QHAMAEDAPHNE TRANSITION AS
A KEY TO VEGETATIONAL STRUCTURE AND ORGANIZATION

by Michael Daniel Byer

This study investigates the use of microdistribution
and interspecific association patterns as possible tools for
unravelling the cause-effect heirarchy of biological commu-
nity organization. Changes in these patterns along a gra-
dient of soil drainage and deveIOpment, with its associated

vegetational transition from Pinus banksiana forest on the

 

well-drained soils to Open Chamaedaphne gatherings bog on
the most poorly-drained sites, were also examined for clues
to this heirarchy. The Spatial scale of patterns, investi-
gated via quadrats of different sizes, also proved helpful.
Measures of association are designed to show whether
high values of two Species or other variables are found to-
gether more frequently (positive association) or less fre-
quently (negative association) than one would expect of
entities independently distributed with respect to each
other. Here, a number of correlation and regression meas-
sures proved their worth, as have these and other techniques

1

 

2 Michael Daniel Byer
in previous studies. Used here were the phi coefficient
( fl ), which shows the degree to which two Species tend to
occur in the same or in different quadrats, the zero-order
product-moment correlation coefficient (3;), which measures
the tendency for high values (Lg. per cent cover or thick-
ness in this study) of species and other variables to occur
together, the first-order partial correlation coefficient
Qij'h)’ which determines the influence of correlations with
other single variables on the correlation between particular
variable pairs, and the fitting of linear, quadratic and
cubic regression curves.

The uniqueness of the present: study lies largely in
its examination of enviromnentally-related changes in asso-
ciation and in its use of several association measures in
conjunction. The great bulk of pattern studies use only one
method, thereby examining but one property of patterns which
have a multiplicity of properties. The use of several tech-
niques and the study of changes lessen the chance that we,
like the blind men examining the elephant, may emerge from
our data with an impression completely out of proportion
with respect to reality.

Attempts to place the species into natural mutually
positively associated groupings by means of simple mathe-

matical formulas based upon these association measurements

3 Michael Daniel Byer

seem not to yield meaningful answers, at least in the vege-
tation studied. Perhaps the grouping tendencies within it
are too complex to produce such distinct units. Moreover,
there are various theoretical objections to such grouping
techniques, for in attempting to oversimplify extremely
complex situations they may tend to obscure rather than
reveal relationships. It is therefore postulated that the
several grouping techniques tested are best used merely as
guides, or reserved for those cases where one has reason
to suspect a distinct vegetational mosaic to begin with.
An effort is made here, instead, to "subjectively" recon-
struct organizational hierarchies on the basis of all the
available information about the species involved.

An abundance of non-random microdistribution of
species with respect both to one another and to variables
of the physical environment, within relatively homogeneous
gradient segnents, is clearly demonstrated in the study
area. This comes as no suprise, in the light of the find-
ings of similar studies and of the theoretical improbabil-
ity of randomness in nature. Part of the demonstrated
pattern may be attributable to large-scale heterogeneity
within gradient segments, and as might be expected positive
and negative associations seem stronger on the physically

more heterogeneous segments, but a major portion can only

4 Michael Daniel Byer
be explained by small-scale, or microdistributional re-
sponses of the species.

A tendency of xerOphytic Species to exhibit
stronger positive and negative correlations on the margins
of their ranges across the gradient than on those segments
where they reach their greatest abundance and vigor is per-
haps a consequence of greater sensitivity to small changes
where conditions approach the tolerance limits of the Spe-
cies, and of those portions of the marginal segments least
similar to a species’ Optimal habitat being unsuited to it.
An opposite trend, i._e_. a general reduction in positive in-
tercorrelation among them near their range margins, may be
related to a reduction in competitive vigor as conditions
close to the tolerance limits of these species are approached.

The larger, more abundant, and more clonal species
seem, in general, to exhibit stronger correlations, prob-
ably because they themselves constitute a strong patterning
force with their large sphere of interactive influence and
ability to exclude Spatially the entry of other species
into a microsite. But on all but the most poorly-drained
gradient segment , there is an overwhelming preponderance
of positive over negative interspecific correlation, ap-

parently reflecting a relatively large-scale patterning of

5 Michael Daniel Byer
areas more and less favorable to plant growth generally.
Thus the preponderance of relatively weakly positive,
circumneutral or slight negative correlations of the rarer
and smaller species suggests inability of these plants to
survive or grow well with the larger and more abundant
species. Probably in this community the former play the
role of Opportunists temporarily utilizing unoccupied
Space (and hence do not reach clonal size or large dimen-
sions), and it is postulated that such Opportunists may be
an important component of natural communities in general.

In contrast to the aforementioned large-scale en-
vironmental variations which produce predominantly positive
correlations, there seems, on the basis of changes in cor-
relation with quadrat size and distributions of positive
and negative correlation values, to be a small-scale inter-
specific competition and spatial exclusion effect which
tends to produce negative association. This combination
of large-scale envirornnental, small-scale interactive pat-
terning is in agreement with the findings of other‘studies,
and is corroborated by the much higher proportion of high
positive i coefficients (based upon presence or absence)
than of gs (based upon per cent cover). The larger scale

patterning, according to this hypothesis, must tend to

/

6 Michael Daniel Byer
bring most species together in the favorable patches, where
interaction tends to preclude luxuriant growth of any one
of them near heavy cover of another.

The "environmental" factors considered in this
study, namely cover of vegetation synusia, thickness and
cover of litter, and thickness of soil horizons, are rela-
tively weakly correlated with species, especially the
second. Perhaps these factors are not the ones most di-
rectly influencing species performance here, or perhaps
interspecific interaction is really a more important pat-
tern-inducing force here than the physical part of the
plants' environment.

Field experiments were designed to test separately
the responses of selected species to the environmental
gradient and to certain other species. The species were
selected on the basis of changes in correlation or unusual
patterns of correlation and association discovered in a
preliminary study. Seed, seedling, and mature plant per-
formance were tested.

These experiments demonstrated that representative
Species of the poorly-drained and imperfectly-drained seg-
ments of the gradient are not excluded from the better-

drained portions by any inability to survive there, as

7 Michael Daniel Byer
adults, in the absence of natural interspecific inter-
action. There is an indication, from plots soon excessively
flooded for further plant survival, that certain Species
naturally restricted to well-drained and intermediately-
drained soils might be able to survive on poorly-drained
organic soils under artificial culture conditions. In
these excessively flooded plots, there seemed to be no
semblance of preportionality between ability to survive
inundation and a species' natural position on the drainage
gradient. With one and possibly two bog species, inability
of seeds to germinate on inorganic soil clearly seems to
be one factor which could limit their spread to such soils,
but with the tested species native to inorganic soils there
is no demonstrable difference in ability to germinate
across the whole range of soil-drainage conditions. Most
probably, then, species distributions across the gradient

are limited primarily by ability of adults to spread vege

tatively under natural conditions of interspecific inter-
action, but in some cases by germination and seedling
survival requirements.

Growth rates and vigor, as estimated from various
morphological measurements, as well as performance in a

manipulated environment (in this case one having exces-

 

8 Michael Daniel Byer
sively droughty conditions) show up much more difference
between Species, and within the same species on different
gradient segments, than do straight survival measurements.
The former two thus show more promise as keys to the physio-
ecological nature of species than does the latter. The
more saphisticated measurements sometimes yield results
unexpected from Species' natural distributions, although
in the case of the manipulated environment some of the
differences seem attributable to differences in tranSplant
hardiness.

Controlled experiments designed to test interaction
between species deal more directly with the maj or questions
of a study such as this one than do simple tranSplant ex-
periments. Unfortunately, meaningful results were obtained
from only two such experiments.

As to specific patterns found in this study area,
moss and lichen cover as a whole, as well as individual
Species within these categories, show a bimodal distribu-
tion across the gradient. Although the primary moss
species are relatively mesOphytic and the lichen species
relatively xerOphytic, members of both categories seem to
require sites sparsely covered by litter or vascular plants,

and the low points of moss and lichen cover seem to be re-

 

9 Michael Daniel Byer

lated reapectively to a scarcity of bare patches and a
maximum of vascular plant ground cover. This ground cover
is itself perhaps reduced by thick litter cover, possibly
because this litter retards soil warming necessary to
vigorous bud expansion in the spring. L

Also noteworthy are two mutually positively cor-
related groupings of species among those which seem to
reproduce primarily from seed in the study area; they are
likely brought about by similarities in germination or
seedling survival requirements among the grouping members.
There is a distinct tendency on the better-drained inor-
ganic soils for the more xer0phytic species to have an
affinity for _C__a_rg_:_r ens lvanica, the more mesOphytic ones
to have an affinity for Vaccinium species. Thus, these
two dominant entities may be visualized as the nuclei of
two "phases" in the vegetation, but these "phases" are by
no means clear-cut spatially, and the two dominants' them-
selves are found together in all sorts of combinations.
The "phases" are therefore best thought of merely as trends.

A probable inhibition of understory species by
m serotina litter and an apparent strong interaction
between 9331; groenlandica and 9232113 canadensis are other

highlights suggested by the correlation analyses of the

 

10 Michael Daniel Byer
mineral soil segments.

In the bog, mature glamavegaphne calyculaga and
other shrubs seem to inhibit herbs and Sphaggum moss, and
possibly a shrub and an herb-Sphaggum stage are associated
in a cycle of hunnnock buildup and breakdown. Rarity of
interspecific interdigitation between shrub clones suggests
a strong amensal reaction between bog shrub Species or
their associated mycorrhizae.

Incidentally, the study provides some evidence
which might be helpful in solving a taxonomic problem.

Two phyletically close Vaccinium taxa, which have essen-
tially identical overall distributions and cover modes
across the gradient, show differences in their association
patterns of about the order of magnitude exhibited by
distinct species. They must represent two partially or
completely isolated gene pools, each with its own unique
potential for environmental response and niche occupancy.
Such isolation could result from one entity being diploid
X. boreale, the other tetraploid E. anggstifolium var.

ni mm, which morphological evidence suggests to be the

case .

PATTERN ANALYSIS ALONG A
PINUS-CHAMAEDAPHNE TRANSITION AS

A.KEY TO VEGETATIONAL STRUCTURE AND ORGANIZATION

By

Michael Daniel Byer

A THESIS

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

DOCTOR OF PHILOSOPHY

Department of Botany and Plant Pathology

1965

CONTENTS

CHAPTER
l L1. t o f r‘b 1e . O O O O O O O
’ List of Figures . . . . . .
List of Appendices . . . . .
; Acknowledgements . . . . . .
I INTRODUCTION . . . . . . . .

II GENERAL PROCEDURE . . . . . .
‘A. Location and Character of
‘3. Field Methods . . . . . .

1958 Sampling . . . .
Experimental Plots . .

C. Analysis Techniques . . .

the

III ZERO-ORDER CORRELATION COEFFICIENTS

A. MethOdS o o o o s s e s 0

VB. Demonstration of the Existence

of Pattern . . . . . .

C. Influence of Soil, Drainage . .

D. Relation Between Frequency or Cover

and Strength of Correlation . . .

. Influence of Quadrat Size . . . . .

E
F. Prevalence of Positive Correlation

G. Comparison of d and 3.; coefficients

H. "Environmental" Parameters .

I. Usefulness of Non-significant

correlations . . . . .

11

PAGE

vii
viii-a

10
10
21

21
25’

25
39
39

44
47'

58
62
69
78
83

89

iii

IV THE QUESTION OF GROUPING . . . . . . .
v PARTIAL CORRELATIONS . . . . . . . .
v1 HIGHER POWER REGRESSIONS . . . . . . .

A. The Hypothetical Data . . . . . .
B. Data from Dawson-Greenwood Peat .
VII FIELD EXPERIMENTS . . . . . . . . . .
A. Methods . . . . . . . . . . . . .
1960 Plots . . . . . .

1961 Plots . . . . . .
Analysis of Data . . .

B. 1960 Plots: Survival, Flowering
and Fruiting . . . . . . . . .

C. 1960 Plots: Seed Germination and
Seedling Survival . . . . . . .

D. 1960 Plots: MOrphological
Measurements . . . . . . . . .

E. Interaction Plots: Survival and
Growth Measurements . . . . . .

F. Interaction Plots: Interaction .
G. Comments on Individual Species .
VIII ENVIRONMENTAL VARIABLES . . . . . . .
A. Synusia . . . . . . . . . . . .
B. Litter Parameters . . . . . . . .

C. Soil Horizons . . . . . . . . . .

120
138
189
195
199
212
213
213

218
221

224

235

239

241
258
263
276
276
288
297

 

iv
CHARACTER OF THE VEGETATION . . . . . . . 302
A. Small Solitaries (SS) . . . . . . . . 303
B. Clonal XerOphytes (CX). . . . . . . . 313
C. Margin Dry (MD) . . . . . . . . . . . 321
D. Vaccinium, an Ecotaxonomic Problem . 329
E. Margin we: (MW) . . . . . . . . . . . 334
F. Bog Species (B) . . . . . . . . . . . 342

SUMMARY AND CONCLUSIONS . . . . . . . . . 349
Bibliography . . . . . . . . . . . . . . 367

Appendix A, Separation of Soil Types . . . 373
Appendix B, Abbreviation Symbols . . . . 375

Appendix C, Species Included in Groupings 377

II

III

VI

VII

VIII

XI

XII

TABLES

PAGES

Frequency, Cover, Means and Standard
Deviations of Species and Other
Variables Across the Gradient . . . 27-38

Per cents of Correlation Coefficients
Significant by variable Grouping,
Soil Type and Quadrat Size . . . . . 55

Relationship between Cover and Strength
of Correlation . . . . . . . . . . . . 60

Correlation Shifts between Quadrats of
Different Sizes . . . . . . . . . . . 65

High Positive and Negative Values, MOdes
and Means of all Correlations by
8011 Type a o a a a a a a a a a a a s o 70

Fraportions of Positive and Negative
Correlations on Segment DG . . . . . . 79

List of all Zero-order Correlation
Coefficients . . . . . . . . . . . . 91-119

Influence of Several Variables upon
Correlations with Tree Cover . . . . . 151

List of the more Important Partial
Correlation Coefficients . . . . . 154-188

1960 Plots: Survival, Flowering
and Fruiting . . . . . . . . . . . 229-233

1960 Plots: Significance Differences
between Gradient Segments . . . . . . 234

1960 Plots: Gemmination and Seedling
SUI-Viva]. a o o o a a a a s a a a a s 2 3 8

V

XIII

XIV

XVI

XVII

vi

1960 Plots: Growth Measurements,
Differences between Segments . . .

1961 Interaction Plots: Survival . . .

1961 Interaction Plots: Vaccinium
Taxa, Differences between Treatments

1960 and 1961 Plots: Comparisons
of Growth Measurements .. . . . . .

1961 Plots: Results of Interaction
Experiments...........

242-243

253

257

259

264

10
11
12
13
14
15
16
17

18

FIGURES

Location of the Study Area . . . . . . .

Photographs of the Vegetation of the
StUdy Area 0 a o O a a a a a o a a a

Shifts in Correlation with Quadrat Size

Correlation values for Each Segment . .

Correlation values for Each Segment, cont.

Vicarious Association of Physio-ecologically

Dissimilar Species . . . . . . . . . .

PAGE

12

19
64
76

77

127

Hopkins Basic Units on Each Gradient Segment 137

Hypothetical Linear, Quadratic, and Cubic
Regressions . . . . . . . . . . . . .

Hypothetical Regressions, cont. . . .
Hypothetical Regressions, cont. . . . .
Hypothetical Regressions, cont. . . .

Hypothetical Regressions, cont. . . .

Regression Curves, Segment DG . . . . . . .

Regression Curves, Segment DG, cont. . .
Regression Curves, Segment DG, cont. . .
Regression Curves, Segment DG, cont. .

Survival and Growth, 1960 Plots . . . . .

Survival and Growth, 1960 Plots, cont. .

vii

198a
l98b
l98c
198d
198s
208
209
210
211
244
245

 

viii

19 Survival and Growth, 1960 Plots, cont. . . 246
20 Survival and Growth, 1960 Plots, cont. . . 247

21 Survival and Growth, 1960 Plots, cont. . . 248

c: I: a» '3‘

APPENDICES
Separation of Soil Types . . . . . . . . .

Abbreviation Symbols . . . . . . . . . . .

Species Included in Groupings . . . . . .

viii-a

gag:
373
375
377

ACKNOWLEDGEMENTS
The central figure, among a fair number who helped

me in diverse ways to formulate, construct, and perfect
this thesis was certainly Dr. John E. Cantlon of the
Department of Botany and Plant Pathology, Michigan State
University. As my major professor, he first inspired me
to begin the problem, then, like an experienced traveler,
described the territory to be explored in detail. He did
not, however, pressure me to follow any particular trail.
In fact, his academic respect for me, even to the point of
allowing me to make my own mistakes if need be, greatly
aided me in my intellectual maturation. Without this
personal tolerance, moreover, the work would surely have
been more difficult.

A very special debt of gratitude belongs to the
late Dr. Phillip Clark, of the Michigan State University
Department of Zoology. His aid with the biometrics of the
problem went far beyond mere methodology and technique,
indeed to the very nature of life itself, and his penetrat-
ing insight could not be contained within pedagogic bounds.
The guidance and the mental stimulation provided by this

unique person were irreplaceable.

ix

 

 

 

x
The remaining members of my graduate committee at

Michigan State University, Drs. William B. Drew and John

H. Beaman of the Department of Botany and Plant Pathology
and Dr. Eugene P. Whiteside of the Department of Soil
Science, offered many helpful suggestions concerning the
manuscript. Dr. Beaman, in addition, criticized Chapter

9, Section D on Vaccinium intensely and constructively,

and stimulated me to more critical thinking with regard to
several other parts, particularly Chapter 4. Dr. Whiteside
was good enough to examine the soils in my study area and
check my nomenclatural determinations.

Dr. Dennis W. Strawbridge Of the Department Of
Natural Science Of this imiversity kindly read the manu-
script. He discovered a number of inaccuracies and Omis-
sions, particularly in the discussions of statistical
methods, and Opened my eyes to a number of new and exciting
directions which the problem might take in the future.

Several other persons were also helpful in the
statistical considerations. Dr. TomAs Rodriguez-Bachiller,
of the Department of Mathematics, University of Puerto
Rico College of Agriculture (Mayaguez) discussed with me
my plans for Chapter 5 on the partial correlations. Dr.

Ishver Bangdiwala of the Puerto Rico Superior Council of

 

 

xi
Education, read and criticized the manuscript for the same
chapter, and Dr. Richard Levine Of the Biology Department,
University of Puerto Rico (Rio Piedras) assisted with some
last-minute problems .

Much of the work in a problem of this sort is in
computer programming. In this, I would have been literally
helpless without the aid of several persons connected or
affiliated with the Michigan State University Computer
Center. Particularly outstanding among these are Mrs.
Beth Unger, Mrs. Marjorie Williams and Mr. Francis Sim.

I am especially indebted also to Mr. Jim Clark of the
Michigan State University Department of Psychology. He
participated with me in a partnership of writing and per-
fecting the program for partial correlations, in which I,
alas, played a very minor role. Thanks are also due Mr.
Donald Spyke, manager of the aforementioned computer
center, who cOOperated with Dr. Cantlon and myself beyond
the call of duty.

Mr. Victor Davila and Mrs. Algarin of the Univer-
sity of Puerto Rico Computer Center at Mayaguez programmed
and executed some eleventh-hour calculations.

My wife, Mrs. Francisca Pons de Byer, was very

much in the thick of all phases of analysis and writing.

 

 

xii
She did the bulk of the first drafting of the figures and
the tabulations, much administrative work, and finally
helped with the copying of the manuscript and proofreading.
Her special training and talent in mathematics were useful
on several occasions. Moreover, her moral support, plus
her virtuosity in inciting me to extract the best from
myself, contributed more than I can measure with correla-
tion or regression coefficients.

It is a rare person to whom drudgery is its own
reward, but I was fortunate enough to discover two of them
as the work progressed. Miss Georgina Pons-Otero devoted
many nights to the first typing of parts of the manuscript
and the final typing Of the two large tables. In addition,
she scouted out other valuable secretarial help. Mr.
Samuel Pans kindly and patiently prepared most of the
figures for photographic duplication.

Mr. Norman Smith of the Michigan Department Of
Conservation was most helpful in securing permission for
the experimental manipulation of State Forest land. His
active interest in the project was most encouraging. I
wish to cite also the personnel of the Regional Office Of
the same department at Roscommon and the Crawford County

Office at Grayling for their cooperation.

 

 

CHAPTER I
INTRODUCTION

Statistics is regarded today by virtually all
scientists as one of the basic weapons in their arsenal.

SO universal is the acceptance of biometric techniques for
the objective testing of hypotheses that we are, perhaps,
too seldom moved to question their worth. Yet these meth-
ods'are most frequently used to help answer Specific ques-
tions which one has in mind. Nonetheless, it is in the
nature of science, as indeed of any inquiry into the nature
of the universe, that we sometimes come upon areas fertile
for investigation, but so unexplored that we simply do not
know M questions to ask. In such cases, a statistical
procedure which would suggest the best places to start
investigation would be most helpful.

The organizational basis of a biological community
is, I submit, a little known area which could benefit from
just such treatment. 0n the one hand, we have excellent
autecological "life history" studies of many species, in
which the performance Of organisms is tbo frequently viewed
as a result of physiological reactions to rather than in-

l

 

2
teractions with an integrated biotic-abiotic system. On
the other, we have a fairly recent body of descriptions of
the detailed structure of the communities containing such
Species. Yet this structure must surely result in large
measure from.the nature and activities of the individual
species making up the vegetation. Only by using the two
approaches in conjunction can we avoid the "apparently
barren nature of some statistical work in ecology" which
Graig-Smith (1964) attributes to failure to recognize emr
pirical description of structure as "a starting point for
further investigation of the factors reSponsible and not
an end in itself". The autecological study of a whole nat-
ural assemblage of organisms is vastly more complex than
the investigation of a single Species' interrelationships
with its environment. Statistical procedures revealing
the distribution patterns resulting from the interactions
among the biotic and abiotic components Of the assemblage
help to suggest which, among literally thousands of possible
experiments involving these species, should prove most re-
warding. Thus one may avoid the fruitless testing of count-
less postulates which the existing structural patterns
indicate to be false.

The present thesis, representing some earlier ef-

 

 

 

3

forts in a projected continuing study, is necessarily
largely descriptive and concerned with methods. But it is
done in the spirit of thus using statistics primarily to
suggest questions rather than to test the validity of hy-
potheses already formulated. By pointing the way to some
of the best avenues of investigation and contraindicating
many which are doomed to failure, it will hopefully make
possible, in the long run, a far greater return per unit
of expended energy than would otherwise be obtainable.

Thus the seeming prodigality of the vast initial calcula-
tion operation may be compensated many times over. Besides,
the organizational analysis of whole communities, which is
thus greatly facilitated, should in the end also prove more
productive than isolated life history studies done with
only rudimentary knowledge of, or regard for, the communi-
ties which the organisms occupy. It is doubtful that such
investigations of coumunity organization will replace in-
tensive investigations of single Species or make them obso-
lete. They can, however, pave the way for more realistic
treatment of these species as working components of the
biogeochemical wholes with which they evolved, not as
isolated active entities pulling and tugging at a sort of

inert physio-chemical conglomeration which enve10ps them.

 

 

The research herein discussed has the following
objectives: (1) characterization of the large-scale (£13933-
distributional) and various small-scale (microdistribution-
_a_l) patterning present in a particular plant community,
and changes in these patterns along a drainage-related
gradient of vegetation and soil development, (2) an explo-
ration of several types of correlation-regression analyses
as techniques for elucidating these patterns, (3) the
construction of plausible postulates concerning the phenom-
ena responsible for the patterns, which will thereby sug-
gest the nature of some aspects of community organization,
and (4) the testing of a few of these postulates by means
of field eXperiments.

"Macrodistribution", as the term is used in this
thesis, refers not to geographic distribution, but rather
to changes in frequency and per cent cover related to the
area sampled on the soil-vegetation gradient, and hence to
change in the distribution of papulationg. "Microdistribu-
tion", on the other hand, refers to variation in the loca-
tion and luxuriance of individuta_l_.§_ fl _s__m__al._l_. aggregations
(or other convenient units such as individual shoots of a
clone) within areas of a few square meters or less.

By "patterning" is meant non-random distribution.

 

 

Randomness, $.53. the kind of distribution which would re-
sult if every point in the study area had an equal chance
of containing an individual of a Species, has been shown
to be rare and possibly nonexistent in nature. Species
usually tend to be clumped or overdispersed (points nearer
an individual having a greater chance of containing an-
other), occasionally regularly distributed or underdis-
persed (points nearer an individual having a smaller chance
of containing another) in reSponse to their own reproduc-
tive patterns and the patterns existing in other elements
of the community with which they interact (Odum and Odum
1959, Hutchinson 1953, Cole 1949, Graig-Smith 1964).
Correlation-regression analysis (2 above) is here
used to show the degree of non-random distribution of
species with respect to these other elements, the latter
being either other species or environmental variables.
The regression coefficient or the plotted regression curve
shows the changes in a dependent variable, y, as an inde-
pendent variable, x, changes. The product moment correla-
tion coefficient, on the other hand, indicates the relative
goodness of fit of a series of points (values in each

sample) to the best-fitting regression curve, no matter

the slape of this curve. Thus, in a sense, the regression

 

 

6

answers the question of "how much?" or "how strong?" the
mathematical effect of one variable upon the distribution
of another may be (not necessarily implying a biological
cause-effect relationship). The correlation coefficient
shows the consistency or reliability of the relationship.
Since the present study is more concerned with, the exist-
ence of relationships than with their magnitudes, greater
attention is given here to correlation, although regres-
sions may sometimes answer important questions which corre-
lations cannot (Chapter VI).

If non-randomness consists of a tendency for
higher values (e.g. cover, density, frequency) of a species
to occur with the higher values of a second species-or
other variable it is called positive ggsociation, whereas
if each variable tends to be relatively high in value where
the other is relatively low it is called negative associa-
Ligg (cf. Greig-Smith 1964). The same definitions and
techniques can be extended to combinations of three or
more variables, but this is not done in the present study.

The correlation and regression techniques are not
designed to reveal the spatial details of pattern (e.g.
size of clumps, spacing between them, etc.), as some other

procedures are, but since they deal with combinations of

 

 

 

7

variables the correlation and regression coefficients are
more efficient in unravelling the complexities of an or-
ganizational heirarchy. Moreover, they put the investi-
gator in a relatively favorable position for seeking the
causes of this organization. The knowledge that two
Species are positively or negatively associated is obvi-
ously more valuable in attempting to ascertain whether
they interact favorably or unfavorably with each other, or
respond similarly or differently to various factors, than
are figures on mean number of individuals per quadrat or
breadth of their clumps. (The latter may also be of some
help if the scale of variation of certain possible cause-
tive factors is already known). The choice of correlation
and regression analysis over some other possible techniques
was also based partly upon the relative ease of the calcu-
lations and the availability of computer programs.

The "causes" sought for existing patterns in the
present study are mostly on a level which might concern a
practicing agriculturist, e.g., Species A inhibits or
stimulates the growth of Species B. We should not lose
sight, however, of a more ambitious level of understanding
which hapefully will be possible in the future, a level at

which we can describe biological communities as functional

 

 

8

networks of chemical reactions, energy flux, and infome-
tion transfer. The theories of the trophic-dynamic ecolo-
gists (e.g. Lindemann 1942 ) do visualize biotic-abiotic
assemblages in this way but are inadequate to describe the
detailed organization within a single traphic level, with
which I am dealing here. H. T. Odum (1960) has been ex-
ploring the question via his analog circuit studies.

Concerning objective number 1 again, the planners
of the study felt that an investigation of changes in small
scale community organization as a function of community
change along a gradient of some kind would add another
meaningful dimension. Shifts in abundance and pattern
relationships among Species in response to varying condi-
tions along this gradient may provide clues to the roles
of these species in community structure unobtainable from
an examination of association on any portion of the gradient
separately. There has been considerable attention to vege-
tational changes along large-scale gradients (e.g. Curtis
and McIntosh 1951, Curtis 1959), and emphasis upon changes
in association patterns on a similar scale by a few workers
(3.3. Whittaker 1956). But relatively few ecologists
(notably Bray 1956, Kershaw 1963) seem to have explored

gradient-related microdistribution and association changes

in some detail.

The present study is unique, to my knowledge, in
this investigation of association changes along a gradient
combined with the use of several different measures of
association. Each of these measures answers somewhat dif-
ferent questions about the patterns present, and thus per-
mits the viewing of several facets of these departures from

randomnes s .

CHAPTER II
GENERAL PROCEDURE

A. LOCATION AND CHARACQTCER OF THE w. The area
used for all phases of this study was located in extreme
south central Crawford County, in the northern part of
Michigan's lower peninsula (Figure 1 A, page 12). It is
on State Forest land in the southeast 1/4 of Section 30,
Township 25 North, Range 3 West. As shown in Figures 1 B
and 1 C, it is approximately one mile west of Route l8-76
between the towns of Grayling and Roscommon, one mile north
of the Roscomnon County line, about two miles east of
Higgins Lake and County Road 270 (Old 0.8. 27) and two and
one-half miles east of the Lansing-Mackinaw Bridge Through-
way (Present U.S. 27).

The soil parent materials in this part of Michigan
include extensive areas of sandy and gravelly tills and
outwashes of late Wisconsin glacial origin (Veatch g; 9;.
1924). The tills (formerly mapped as Roselawn sand and
related soils by Veatch £3; .a_]_._. 1924), which comprise the
low rolling tapography of the ground moraines, support a
tree cover of the oaks Qgrcus m, 9. a_1_13_a,, and members

10

11
of the Q. ellipsoidalis complex.1 The extensive outwash
plains, on the other hand, characteristically support for-
ests or savannas of 11:93; banksiana, the mid-western Jack
pine (Figure 2 A, page 19), often with interspersed gi_e_rcus
ellipsoidalis and 11-929; resinosa. More rarely, g. resinosa
may form the dominant tree cover and occasionally the
@2th £22. characteristic of tills may dominate. Stands
of Bigtoothed aspen, ngulus ggrandidentagg, may occur on
both tills and outwashes as may Red maple, Ag; rub_______rmn, as
a subordinate component. Understory vegetation in both
pine and oak stands is strikingly similar, dominated by
two Vgccinium species (blueberries, huckelberries) , 23:3};-
igigm aguilinum (bracken) , §_a_!_:_e_z_c_ ngnsylvanigja. (a sedge),
mill-é W (Sweet fern), and (more prevalent in
treeless sites and savannas) Danthonig spicata (Poverty
oat grass). The outwash soils, formerly the only ones des-
1Smited Grayling sand and its relatives, on the whole are
said to be poorer in fine materials and soluble materials
than the hilly sand areas mapped earlier as Roselawn com-

plex. But an individual sample of outwash may assay higher

1Scientific names of vascular plants cited in this
thesis follow Fernald (1950), except when an author is given
the first time the species is mentioned.

FIGURE 1.

 

A.

Location‘gg the Study Argg

Map of Michigan, with Crawford County
and the northern part of Roscommon
County shaded. White dot shows the
study area. (Modified from Veatch,
Schoenmann and Moon 1924).

Portion of a highway map, showing the
shaded area of A. Dot indicates the
study area.

Detail from a soil map, showing four
sections in Crawford County, Township
25 North, Range 3 West. The study area
(square) is located just upward and to
the left of center. (Modified from

Michigan Department of Conservation
1927).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mom), FIGURE 1 c

 

 

 

Dawson-Gmemlood peat

[:l Grayling sand

 

7//// Kinross loamy sand
maelawn sand, gravellyx
maelawn sandy loam, gravelly?I

   

Saugatuck sand
_ Hislmys

‘—-—— Secondary roads
----- Trails

Section lines

C] Study area

 

 

xljlow mapped as Grayling

13
in both than one of till of similar geological origin
(of. data of Veatch 1953). It is quite probable, then,
that the vegetational differences arise from topography
related drainage and evapo-transpiration differences and/
or the greater intensity and rapidity with which fires can
burn over a flat surface, rather than from parent material
differences. The existing oak, sapen, and Jack pine stands
are all apparently successional, dating from the early
1900's when the fire frequency began to decline following
the regional clearcutting and subsequent slash fires of
the late 1800's (Veatch g; _a_1._. 1924, Kilburn 1950).

The current soil classification combines soils in
sandy outwash and till deposits into one soil series, if
there are no other observed profile differences; hence
Roselawn is now included in Grayling (Whiteside _e__ 31.
1955).

The Jack pine-supporting outwashes are pitted with
glacial potholes. Various types of bog and marsh vegeta-
tion have developed in them, and it is along the border of
one such pothole that the study area is located. It is a
poorly-drained, acid (non-calcareous) bag of the type
known as leatherleaf for the dominant shrub , Chamaeda hne

calyculata (the type is also known locally as "huckelberry

14
swamp" although the plants referred to, Vaccinimp boreale
Hall and Aalders and 1. wstifolium var. m, rarely
comprise more than 10-20% of the biomass). The area was
chosen so that it would encompass roughly equal prOportions
of Grayling sand, the bog soil type Dawson-Greenwood peat; "
and soils intermediate between them. Because of scale
limitations only Saugatuck sand, among these intermediate
types, is shown in the soil map from which Figure l C was
taken. The Dawson-Greenwood peat is a poorly descomposed,
fibrous type.

Grayling sand is characterized by the near absence
of a leached, or A2 horizon and a very yellowish zone of
accmnulation, or B horizon; that is, the B has minimal/
sesquaoxide and organic deposition and so is not very dif-
ferent in color from the parent material. "Wet Grayling"
(an arbitrary unit separated from Grayling sand when the
latter proved to occupy ; disprOportionately large area)
reflects slightly poorer drainage in an increased leaching
and increased depositing, which become more pronounced on
the still less effectively drained Croswell sand. On the
latter, also, the red-black/coloration of the B horizon be-
comes quite intense. Continuing in the direction of poorer

drainage, Au Cres sand is distinguished from Croswell sand

15
primarily on the basis of the former's pronounced/B horizon
aggregation; firm pressure of the fingers is required to
break the chunks. This tendency reaches its maximum on
Saugatuck sand, where the B forms a @pan which can be
broken only by chipping. This cementing begins to decrease
on Kinross sand, and concomitantly organic matter (A0 hori-
zons) increases to a depth and consistency resembling that
of the peat. The Dawson-Greenwood peat is distinguished
from Kinross by having a thickness of incompletely decom-
posed organic matter of 12 inches or greater. The areal
divisions between soil types, as listed in Appendix A
(page 373) are measurements intermediate between those
chosen by McKenzie and Whiteside (1956) as "typical" for
each type (except in the case of "Wet Grayling").

From Grayling through Au Gres the overstory vegeta-
tion remains my; banksiana, its cover varying little
throughout this portion. The associated understory vegeta-
tion, however, changes radically. 0n Grayling, it is char-
acterized by extensive sparse patches and a dominance of
9333;: ens lvanica, Vaccinitm: gygustifolig var. m
and X. boreale (Figure 2 A, page 19). The primary change
on "Wet Grayling" is a tendency for the tall bracken' fern

gteridium aguilinum, growing in fairly scattered patches

16
on Grayling, to become extremely lush and sometimes to form
almost solid cover. 0n Croswell, as one follows the gra-
client in the direction of progressively less efficient
drainage, certain species relatively minor on Grayling and
"Wet Grayling" such as Qultheria procumbens, Epigaea
megs and Maianthemum canadense increase to near their
maximum cover. Concomitantly, several important Species
of the middle gradient ("bog margin") first appear in any
but sporadic quantities: 992g; ggognLandigg, w
canadensi , £21333 hispidus var. obovalis and Yaccinium
mntilloides. In addition, total ground cover increases
(Figure 2 B). But the vegetation of Croswell is still
predominantly "dry end", in contrast to that on Au Gres
where Gaultheria etc. , M etc. , and the two gradient-
spanning Zacc iniun taxa (which reach their maxima here)
form an unusually continuous ground cover. This lushness
could be due to several factors, such as drainage favorable
to plant growth generally, a B horizon texture which is
neither impenetrable to water nor conducive to excessively
“Pi-d percolation, and an "edge effect" which provides con-
siderably more light (from the nearly treeless bog) to this
band than to the better-drained soils. The portion of

Saugatuck adjacent to Au Gres is vegetationally similar to

17
the latter, but with considerably Sparser ground cover.
Closer to the bog, however, the tree cover becomes much
more continuous, and along either side of the Saugatuck-
Kinross line runs a band of tall, closely Spaced trees not
only of gm banksiana, but of 13335 mariana and Lag-£5
laricina as well (Figure 2 C, left). Approaching this
band from the "dry end" side, the Au Gres ground vegetation
begins to thin out, and a few shrubs such as Lady; gagg-
landicum, 551.313 olifolia, flgmopanthus mucronggg and
Yiburnum cassinoides assume some importance (they do not
penetrate to soils which are much better drained than
this). Under the heavy marginal tree canOpy itself, ground
cover becomes sparse; there are usually depauperate indi-
viduals of the shrubs just mentioned, as well as of Species
characteristic of both bog and "bog margin" and even the
"dry end", but apparently light intensity is simply too
low here for good plant growth. The part of Kinross sand
lying on the "wet end" side of this tree border supports a
vegetation typical of the bog; dominant Chamaeggphne as
well as scattered clones of 29.1%» Vaccinium anggstifolium
var. p_i_g_rig and y, boreale rooted in a mat of living
§phaggmn moss, with occasional scattered small trees of

gig, Picea, Pinus banksiana, g. strobus and g. resinosa.

 

 

18
On Dawson-Greenwood peat nggm.cover drape but otherwise
the vegetation remains similar (Figure 2 C, right).

In summary, then, there is a gradual shift in the
pine forest floor vegetation between the "dry end" and mid-
gradient, then a sudden transition across a narrow zone of
bog margin trees to the open bog vegetation. There are
but two shrub-herb Species which extend to both sides of
the marginal band of trees to any great extent, so that
this transition satisfies the definition of an ecotone
(Whittaker 1956, Curtis 1959, Becking 1957).

This particular drainage-related soil and vegeta-
tion gradient was chosen for this study because the vege-
tation is floristically poorer than that on any other
gradient of comparable range in drainage in the region.
This is thought to result from the relative coarseness and
low nutrient content of the inorganic parent material, and
the low pH, low oxigination, and possibly low availability
of nutrients on the organic soil. Floristic paucity is
desirable in a pioneering study of pattern, particularly
one which emphasizes association between Species as this
one does, because the number of variables and consequently
of possible combinations is thus minimized.

Indications are that the area remained essentially

FIGURE 2. Photographs 2; the Vegetation.g§ the Study

Area.

A.

B.

C.

i
’ t“

Pinus banksiana stand on Grayling
sand, showing characteristic Sparse
understory vegetation.

génus banksiagg stand on Au Gres or
Saugatuck sand looking towards the
better drained portion of the grad-
ient. Note the very lush under-

story vegetation.

Left, zone of dense trees (Pinus,
gicea and Larix) on Saugatuck and
Kinross sands along the bog margin
or hingeline. Right, characteristic
Open bog vegetation on Dawson peat:
dominated by the shrub Chamaedaphne
calyculata.

19

 

20
undisturbed during the 55 years or so preceding the begin-
ning of the study. Rings were counted on several 3.1.29.9:
banksiana on the inorganic soil, and on several stumps of
each conifer Species present in the bog in the western
portion of the study area, which was clearcut for experi-
ments. These were supplemented by core borings on three
2.19.9.9. banksiana, two 21.532 and two _I:_a_r_i__:_c_ distributed
throughout the remainder of the area studied. They re-
vealed that most trees began growth 55 j; 3 years before
the 1958 sampling, probably following a fire. Virtually
all the may; on inorganic soil probably originated at
this time, since closing of the vegetation cover would
have tended to prevent sufficicient light from reaching
younger seedlings of this shade-intolerant tree. Seedlings
have continued to establish themselves in the more Open bog
up to the present time, however, although some of the oldest
trees here are but 3 inches in DBH. The bog vegetation did
not give evidence of recent rising or lowering of the
water table. The vegetation thus seems to have reached
the relative equalibrium characteristic of older Jack pine.
If the findings of Kittridge (1934) hold for the region in
question here, this vegetation might remain somewhat as is

for another 40-50 years.

21

B. FIELD METHODS. 1358 Sampling. A square area

 

of 100 X 100 meters was staked out in such a way that it
embraced the entire soil-vegetation transition discussed
above. Prior inSpection suggested that an area of this
size would be sufficient to encompass the normal range of
macrodistributional variation found on each of the soil
types in this and similar stands in northern lower
Michigan. Thus, the results should be fairly indicative
of what one might expect generally in the vegetation types
studied. The square Shape was chosen because it best fit
the particular gradient, a rather abrupt one. An oblong
sample universe whose long axis paralleled that of the
gradient would have included too little of the soils de-
velOped under intermediate drainage conditions. Orienta-
tion was with the sides running approximately East-West
and North-South, the latter closely following the orienta-
tion of the gradient.

Four hundred l X 1 meter quadrats were randomly
located within the area, as many as time permitted, so
that standard errors of the coefficients to be calculated
later would be as small as possible. The 1 X 1 meter
quadrat was several times larger than the individual shoots

or tussocks of any species dealt with, so this seemed about

22
the right size to show up the largest scale association
patterns present. Sampling was carried out between mid-
June and mid-September of 1958. With one or two exceptions
(pointed out where apprOpriate) the aerial portions of the
Species seemed neither to grow nor to die back to any ex-
tent during this period, so it is unlikely that seasonal
changes affected the measurements appreciably deSpite the
relatively long sampling period. Nonetheless, reading of
the quadrats was done in such a fashion as to insure that
all parts of the study area would be prOportionately repre-
sented in the sampling done during each of several rela-
tively short time intervals. In this way, apparent
differences between segments of the gradient due to tempo-
ral factors was minimized.

The reason for a complete random sample, rather
than a stratified one which would have insured a reasonably
large number of quadrats and consequently lower standard
errors on the arealy smaller gradient segments, was a
practical one. Any detailed soil mapping done before the
vegetation sampling would have disturbed the plants unduly.
Moreover, since exploratory soil profiles close to the
Sample universe had revealed that soils varied greatly

within a space of two or three meters, it was obvious that

23

profiles dug in the center of each quadrat after vegetation
sampling would insure more accurate determination for each
quadrat than would a soil map based upon profiles dug every
ten meters or so. In the latter case, some quadrats would
have been five meters from the closest soil reference point.

Each 1 X 1 meter quadrat contained a centrally
located 1/2 X 1/2 meter quadrat, and the latter a central
1/10 X 1/10 meter quadrat. The smaller quadrats are not
randomly located, since their positions are predetermined
by those of the largest sized quadrats. However, they
may be thoughtof as being centered at some of the inter-
sections of a l X 1 meter grid, these intersections having
been chosen at random. Their location is thus similar to
that of randomly located point quadrats, used by many
workers including Creig-Smith (1964) and Kershaw (1964),
but they have non-zero area. Location at grid intersec-
tions makes the assunption that there is no Spatially
regular variation in the vegetation of about the same
scale as that of the grid, such as that which Harper
and Sager (1953 in fact demonstrated in a field containing
old plow furrows. But that does not seem a likely eventu-
ality, and so it is probably safe to assume that the

Smaller quadrats give an unbiassed sample, though of course

24

randomness cannot be assumed in comparing results from
units of one size with those from units of another size.

For each of the three quadrat sizes, presence of
all vascular species was recorded and their cover estimated.
Presence of all bryOphyte and lichen Species was recorded
also, but their cover was estimated only in the smallest
quadrats. Voucher Specimens for all Species encountered
in the quadrats are on file in the Beal-Darlington herbar-
ium at Michigan State University. Total moss, lichen and
tree cover were also estimated for each quadrat, while
total shrub and herb cover were calculated later by summing
cover of the component Species. Litter thickness in each
of the largest quadrats was estimated by measuring it at
five equidistant points within the quadrat and taking an
average. Per cent of bare ground and per cent of litter
covered ground were estimated for all three quadrat sizes.
After the vegetational sampling had been completed, a soil
profile‘was dug in the center of each group of nested quad-
rats and thickness of each soil horizon (to as great a
depth as feasible), depth to gley, and degree of cementing
and color of the B horizon were recorded (the B horizon
(immacteristics were later used only in the separation of

soil types).

25

Experimental Plots. Napieces of the 1958 study

area were used in 1960 and 1961, to begin transplant and

germination experiments. Because the details of these

experiments, unlike those of the sanmling, are pertinent

only to chapter 7, they are included with that chapter in
order to make of it a better integrated whole.

C. ANALYSIS TECHNIQUES. Each quadrat was assigned

to a soil type according to the criteria listed in Appendix

A (Page 373). The seven soil types originally designated

were subsequently combined into five segments, for some

of the less extensive types did not contain enough quadrats

for a worth-while correlation analysis. Per cent frequency

and mean per cent cover of each Species considered were

calculated for each of the seven soil types and each of

the five segments. Standard deviations of the cover of

each Species were calculated also, as were the means and
sample standard deviations of the other variables recorded.
Such data for those 47 Species common enough for further
analysis (present in at least five 1 X 1 meter quadrats)

and for the other 16 measured variables are summarized in

Table I (page 27 ). A complete listing of the Species en-

countered, together with their frequency and cover, is

found in Tables I and II of Byer (1960).

26

Several types of correlation and regression coef-
ficients were calculated for these Species and other vari-
ables. For the sake of greater unity of the individual
sections of the work, discussions of these coefficients
and details of the analyses are presented with the appro-
priate chapters. Similarly, Statistical handling of the
data from experimental plots is described in Chapter VII,
to which it pertains.

The R technique is used throughout the association
studies,‘img., the quadrats are treated as the individual
objects observed, and their various characteristics (pres-
ence or cover of Species, values of "environmental" param-
eters) correlated with one another. The interrelationships
between these characters themselves are thus revealed,
rather than similarities and differences between quadrats
in terms of their Species composition, soil profile devel-

Opment , etc .

TABLE I

Frequency, covery means and standard deviations of
Species and other variables across the gradient.

Symbols for soil type names (listed across the top
of the table's pages) and for Species and other variables
(listed in the left-hand columns) are listed together with
their meanings in Appendix B (page 375).

The variables are listed in the following sequence:
Species in alphabetical order (nos. 1-47), synusia in order
of decreasing height (nos. 48-52) and then total cover of
all synusia combined (no. 53), litter parameters (nos. 54-
58), soil horizons in the order in which they occur Start-
ing at the surface (nos. 57-62, if and horizons

are counted here as well as with lit er), an depth to gley
(no. 63).

 

 

LEGEND
ggadrat size parameter_measured
£11122}, W smbol meaning
1 1 m2 F per cent frequency
4 1/4 m2 FC per cent frequency
H 1/100 1112 for combined grad-
ient segments (see
text, page 25).
C mean per cent cover
CC mean per cent cover

for combined grad-
ient segments

3; mean value

XC mean value for com-
bined gradient seg-
ments

3 standard deviation

of cc or Sic

 

 

 

 

*huummmiagthmetmatthmh

QB! PAR 8011. rue
1m» 3m» ms» 3 33 g 53 s g
1 AI) 1 I “.17 10.87 18.13 21.43 10.00 3.45
PC “.17 .18 18.42 .
c 0.07 0.18 0.05 0.08 0.01 0.0!;
00 0.08 010 00% I
I 0.“ 0067 0018 C
h cc 0.08 0.19 0.13 .
I Do“ 1.20 0052 '-
2 AIS 1 P - - I$.55 10.71 ‘10.” 10.35
m " . 180,52 '-
c ' D 0002 0.05- 0022 0.02
CC III I- 0e09 -
I - C 0021‘ -
k x a C 0010 -
e - - 0.47 -
3 m 1 l' 20.00 15.22 9.09 - - -
PC 20.00 11.61 - -
0 0.09 0.05 0.” - - -
Cc 00” 0.04- . -
I 0027 0.1)} II "
h 00 0e“ 00“ - "
I 0e58 0.21 - C
1 ABC 1 1' 6.67 10.87 'I C I "
m 6.67 - - -
C 0.21 0.10 - - . -
00 0.21 n e.
I 1012 . a n
h 00 0021 'l " .
I 1.17 'l ' "
5 l8! 1 r 8.33 6.52 7.58 3.57 - -
1’0 8.33 7 1h - .
C 0.03 0.02 0.02 0.01 - '-
00 0003 00m - -
I 0015 0.08 " .
" cc 0.03 . .. ..
I 0.22 I' " '

IIIII ll lllll'g

TABLE 1, cont.

mm: 3123* m 3

63181

7m

8011.

90m

* Mumfecmthflretmoftbteble

QM P13
1 1'
1'0
0
CC
I
1 cc
I
1 1'
1'0
3 CC
I
1 1'
1'0
E 00
I
1 1'
1'0
0
CC
I
4 00
I
R 00
I
1 1‘
1‘0
0
CC
I
h 00
I
R 00
I

13-33
13-33
0.03
0.03
0.10

0.04
0.15

9017

112.50
£2.50

1.3h
7-77

10.83
10.83
0.5-5
0.35
2.85

0.39
2.82

0.10
0.65

98-33
93-33
9-75
9-93
13.25

10.3h
13.78

8.96
115.29

28

3011. rm
£9. 2. a a a
10.87 6.06 - - -
8.93 - -
0.03 0.03 - - ..
0.02 - ..
0.07 - -
0.02 o a
001° C c
21.75 18.18 32.14» 30.00 13.79
19.6% . ..
0.21 c e.
1.00 a a
28.26 28.79 39.29 80.00 l$8.28
28.57 50.00 II-8.23
0.25 6.39 5-00
1.92 18.83 17.26
6052 90” 10071 - 3°~5
. 7.89 -
0.05 0.07 0.25 0 0.02
0.06 0.19 .
0027 0083 C
0.05 - ..
0.28 - -
97.83 93.94 82.13 ”.00 3.35
95-53 71-05 '-
8.85 ¥.55 3.1-8 1.02 0.&
60.11 2085 .
10086 “025 "
6092 3056 "'
11.09 5062 "
6067 2.80 '
$.03 7°“ "

.""l8

TABLE I cont.
A

mam

110m

13 00?

N503!

QM

SIZE"

1

1

"803” .8 .8 .803“

' 83

P

0.71
0.71
2.80

-
O

16.67 21.74 48.08
16.67

37.50
0.63 1.25

1.00

2.00

1.02
2.55

0-53
2.38

- 10061
6.25
- 0.07
0.04
0.23

0.05
0.28

0.12
0.75

0.03 0.31
0.19
0-95

0918
0.90

29.17 8.70 9.09
29.17

"I-In-d‘un;..¢. thciac the first page as the

SOILTIE
w s.

.0-

3.57 20.00
0.01 . 4.03

0.39 0.28
0.36
0.62

0.37
0.65

1.05
2.88

029 60.00
39 an

1.63 ° 0.57
1.35
2.35

1.35
2.46

tabla

E 92
58. 62 100.00
58.62 100.00

5-99 23.58
5.85 23.51
8.41 15.97
5036 26012
8.29 17.70
3055 26902
8.48 20.67
3.45 -
0.05 -
1003‘ -
0:03 .
1702“ '-
17028' '-
0018 '
0.18 '
0.75 -
0018 "
0063 -
6.90 1.98

TABLE I cont.
fi—L—__

cm!
@215?
16033 1 '
170m 1
3
18M] 1
1
a
19010 1
20m 1
a.
a
21m 1
1
a
flu-lien

00
0

38-33

43.33
43.33

4.17
4.17
0.04
0.04
0.28

30

1‘2. 2.
23.91 12.12
16

30.43 22.73
25.89

. 0.19
1.10

10.87 4.55
.04

0.03 0.01
0.02
0.12

0.02
0.10

26.09 21.21
23.31

8.70 31.82
22.32
0.09 0.99
0.62
1.69

0.65
1.85

SOIL

g

M

E

B

 

D0

. 20.00 2‘10“ 6093

7.14 30.00 37.93 16.83

Mfumtbflratpepettbtebh

24.14

13.79
13.79
0.16
0.16
0.49

0.11
0.32

37.93
37 .93
0.20
0.20
0.44

0.25
0.78

0.13
0.41

1.05

2511:

ll!

31
SOIL. I!!!
E 2 .3.
96.43 100. 00
75-7 97. 37
1. 3.23 3.99 2.42
2.38 3.60
3.12 3.15
2.60 3.51
3.70 3.75
3-39 3.58
6.99 7.96
- 10.61 39.29 30.00
6.26 36.84
- 0.09 0.11 0.07
0003 0010
0015 0016
0.02 0.18
0.13 0.55
n . 1.52 10.”
- - 0.09

 

m

58.62
58.62
1093

1-93
3.44

1.76
3.48

2.03
4.72

27.59
2769
0.22
0.22
0.60

0.43

0010

0.58

0.76
2.57

1.29
3.79

6.93
6.93
0.31
0030
1087

0.36
2.10

0.54
3.60

I7WBID

281I£I

29llfl.

301lll

"lhlnllcullnupzthc

QB! EAR
SIZI'uIIAS'
1 .F
10
C
00
I
4 CC
I
H CO
I
1 .P
10
C
00
I
1 .1
10
C
00
I
4 CC
I
H’ 00
U,
1 .1
10
0
00
I
4 00
I
H’ 00
I
1 .1
'0
c
00
I

.l I I I I I It:

25.83 34.78 74.24

25.83
0.12
0.12

0.35

0.16
0.55

36.67
36.67
0.09
0.09
0.17

0011
0.34

32

[O

0.04

58093

0.17

0.52
0.88

0.55
1.13

1.00
2.83

30.43 ‘53o03

43.75

0.08

0.25

0.39

0.18
0.46

0.41
1.85

1.24

3. 57 ~20.“
0.03 005°

Luthefirumottheteble

79.31
79-31

0.53
0.53
1.73

31.68
31.68
1.98

1.98
4.02

1.95
4.62

8.49

1.98
0.03

TABLE I cont.
——-_.a.___

33

an! an 8011. m
75m and! man 5 1g 3 5g ,3
30 m 4 ’ oc- - .. -
(Mt.) . D a -

31 on: 1 1" 16.67 10.87 4.55 . ..
re .6 .. .
c 0.06 0.03 0.01 - -
cc I - u
I 0023 I Q

4 ca 0.05 .. -

I 002“ - -
320KB 1 1' 19.17 15.22 21.12 14.29 .
re 19.17 18.75 15.79
c 0.30 0.05 0.08 0.31 -
00 0.30 0.07 0.23
I 1012 0027 1015
4 00 0.24 0.02. 0.30
I 1053 0012 1037
1! cc 0.19 .- .. O
I I.“ «I a
33 P08 1 r - - 1.52 25.00 70.00 44.83
10 .- .. -
11 cc .. . .
I‘ C u -
34 mu 1 1 17.50 23.91 25.76 50.00 40.00
10 17.50 23.21 47.37
0 ma 0.24 0.48 1.08 0.07
00 0.22 0.38 0.39
I 0.81 1053 0.71
4 00 0.16 0.32 0.44
I 0.84 1.88 1.21
3 cc - - 0053
I u u 20%
35 rm 1 1! 61.67 76.09 65.15 53.57 10.00
10 61.67 68.75 42.11
c 6.34 5.97 6.88 6.04 0.90
cc 5.74 6.51 4-68
I 6093 7'09 7.11

'I‘llnlanptun‘thnrn

tmotthtcbh

 

 

81.19
81.19

5-72
14.23

TABLE I cont. 34
-—-—~l_____

QB! PAR -- $11. m
21.12 LWM' 9.. 29. .c. 4.9. 2 .13. 2.9.
352111 4 5 00 5.85 6.41 5.99 .. .
Rout.) . 9.10 9C& 8.” C o
H 00 4.” 6083 9032 U I
I 12088 15092 17:68 D II
3683) 1 P 0.83 4.35 43.94 64.29 50.00 13.79 -
J'c .- 27.53 60-53 13079 "
C T 0.09 0.92 1.59 0.19 0.14 -
CC I 0.57 1.22 0.144 «-
I «- 1.39 2.32 0.45 -
4 cc - 0.63 1.40 - -
I C 1070 3.1.2 o o
I CC - 0.97 2.18 - ..
I II 2089 5.64 - II
37 an 1 l' 9016 6052 60m “.29 . 3*5 -
'0 9.15 6.25 10.53 ' .- I-
0 0003 0.04 0.04 0.12 O 1' I
00 0.03 0.03 0.04. a .-
I 0012 0.14 0.14 - .
4 00 0.05 - .- .. ..
I 0.32 c o o a
38m 1 r .- - -- 1.52 7.14 40.00 89.66 98.02
N - - 15.79 89.66 98.02
3 cc . . 2.97 45.4? 47-85
. - . 15.46 40.45 37.48
39. 8P? 1 r . - . 7.14 30.00 79.31 98.02
to . . . 79.31 98.02
I cc - U I 2'“ 15035
I . I - 6'87 2‘.“
40 001 1 1 10,00 13,04 . .. .. .. ..
c 0003 0.03 n n O I U
cc 0003 0001 " " '
I 0.11 0.06 '3 - -
‘5 00 0.04 . . .. ..
I 0.22 a. u. - -

Eggnus I, (309;;

41 HI

42 VA!

43m

”Conny-1.15.13».

mm
@M
11"
re
0
00
I
400
I
1100
I
11'
10
0
00
I
400
I
300
I
11'
10
0
00
I
400
I
1100
I
11
10
0
cc
I
400
I
100
I

59.17
69.17

1.67
1.64
2.15

1.88
2.90

1.99
4.96

35

4.35 37.88
24.11
0.17 0.55

0.52
1.69

0.43
1.29

0.37
2.53

4.55
0.12

2.96

3.24
3.73

3.73
7.12

801:. m
£8. 2
71.43 60.00
68.42
0.44 0.29
0.41
0.43

0.41
0.74

0.75
3.11

50.00 70.00
55.26
1.61 3.27
2.05
2.96

1.84
3.40

2.29
7.93

17.86 .. 50.00
26.32
0.52 2.25
0.98
2.58

1.05
4.64

1.13
5.09

82.61 93.94 100.00 100.00

2.0889.2§.78 4.36 2.61

100.00

3.90
3.81

4.79
4.90

5059
11- 55

mmummu

.11

10.34

0.27

5.35

3079

1.45
4.23

31.03
31.03

1.24
1.48
2.83

1.24
3.40

75035
75-86

4.31
5.68

4.74
5055

9.21

lllll '8

I‘lO'II ll

13.86
13.86

1.20
1.19
3.88

1.44
5017

1.30
6.23

 

_TABLE I, colt. 36
an! m 0011. run
.._.._mm __8m* 5512- 9. £1 9. 9.9.- 9.
45710 1 _ r ’ 14.17 4.35 3.03 . .
N “017 O a
c 0.03 1- r . .
00 0.03 o . c.
I 00” D u
4 cc 0.02 - .
I 0.09 - Q
#6 '01 1 r I u I c Q
,0 - I C
c I a ' - u I
CC C I I
I O . .
4 cc .. - ..
I C - Q
1 cc . .. .
I, u u U
47m 1 I 60.00 71.74 96.97 96.43 100.00
10 60.00 86.61 97.37
c 2.86 5.02 4.25 5.39 2.99
00 2.45 4.48 4.76
. 3.15 5.09 3.76
4 00 2.58 4.69 5.12
. 3.87 6.24 4.73
n 00 2.39 5.02 6.00
I 6.61 1035“ 901‘
48 m 1 I 96.33 97.83 92.42 100.00 100.00
:0 64 100.00
0‘ 33333 26.58h'20.80 29.45 37.70
4 cc 33.31 29.06 41.61
. 25.31 25.72 30.14
498113 1 c 5.82 8.66 11.18 15.34 18.25
4 00 5.88 10.18 16.33
I 6.80 8091 1003‘
50.! 1 c 18.17 17.20 19.58 21.35 7.71

”Mammmrmnuumh

IIIII [:0

13.79
13079
0.08
0.08

0.29

0.05
0.16

9'33:
4:48
4.31
5.55

5.76
6.79

9.14
17.93

82.76
82.76
21.72

28.75
29.00

24.76

25-36
15.89

3.38

"I

 

.

 

TABLE 1. cont.

OD! 213
143.19. 9.1.91" w 9.
50 m 4 ’ 00 18.54
(60.4.) I 15.84
51108 1 1 61.67

rc 61.67
c 2.95
4 00 2.42
I 7035
52 ml 1 I 56.67
rc 56.67
c 0.69
4 00 0.74
I 1.72
53 M 4 cc ..
I o
54 Ln 4 20 11.43
I 9018
55 MR 4 cc 0.42
I 2.12
56 up 4 00 91.93
I 10.82
57101. 1 i 13.6
4 “£0 13.60
I 5.04
584011 1 i‘ 14.3
4 to 13.88
3 7075
5940! 1 i" 28.7
4 2c 25.17
I 9083

37

‘Mummmmmnummu

8011. m
“.9. 9.. £ 9. .13. P9.
18.75 20.14 2.96 2.78
“03" “019 3.89 3'5~
50.00 46.97 64.29 100.00 100.00 100.00
48.21 73.68 100.00 100.00
1.96 0.99 8.42 24.36 55.91 74.47
1065 1205“ 590w 7‘039
5.43 25.60 35-93 27.41
41.30 30.30 7.14 30.00 44.83 20.79
.82 13.16 44.83 20.79
0.38 0.10 0.01 0.15 0.69 0.17
0.24 0.08 0.32 0.16
0.89 0.49 1.48 0.85
I II #31086 -
- .. 97.56 -
13.46 11.42 6.00 3.18
13.61 9.44 9.40 5.78
0.09 0.26 0.25 0.38
0.51 0.92 0.78 1.46
93.36 80.58 33.31 20.26
8.45 28.88 36.02 25.65
14.1 13.3 14.1 13.2 12.3 12.2
13.70 14.03 12.17 11.47
5.83 7.22 8-45 4.91
13.1 11.6 14.9 29.1 42.8 62.0
12.13 ' 18.79 41.59 60.55
5.49 12.45 33.96 41.08
35.2 37.2 57.1 60.0 139.3 579.2
34-26 57.53 136.24 33 1..
17.2? 33-83 $5.69 9-

 

hu-

f 26.1 34.9 30-5

25.13 31093

14.48 17.46
9.1 27.0 620“.
9.58 46.17
8.36 $027

13806 “7.7 “3 .0

46.20
29.20

1500 50.0 83.3

TABLE I, cont.
QD!‘ m
22.12%” 9.
601.1 1
4 'i'c
I
611.2 1 f
1 To
I
6211:: 1 Y
1. fc
I
6301.1 1 P
4 1'0
I

38

99, .9.

108.23
75.53

comm
AG §.

37.3 35-0
32.26
15.64

86.9 18308
110.11.
95.94

18500 183.3
1”.”
76.11

960‘. WOO

149-32
47.45

.uuuamhcmth'm‘t’w “than.

I”

27.0

28.59

24.41
216.7

213.28
99.54

CHAPTER III

ZERO-ORDER CORRELATION COEFFICIENTS
A. METHODS. For each quadrat size on each
gradient segment, the following statistics were calculated.

1. Zero-order product moment correlation coeffi-

cients, based upon per cent cover, were found between all
possible pairs of those species occurring in five or more
quadrats, between all possible combinations of one of
these species with another measured ("environmental") var-
iable and between all pairs of such "environmental" vari-

ables. Assuming a normal distribution, it is found that,

for all segments, presence in five or more quadrats will
leave more than 952 of the observations within the 2 0"

confidence interval, which Sokal and Sneath (1963) deem

sufficient. With a skewed distribution, probably more

cannon in nature, the per cent falling within this interval

is even higher. Rarer species were ignored both because

of possible decrease in this percentage, and because dis-

tributions depart increasingly from the normal as the num-

ber of occurrences becomes lower. Normal distribution is

assumed in the correlation coefficient. The small letter

39

4O

sed to designate this statistic.2

;_is conventionally u
lculated is presented in

A complete list of the g values ca

Table VII (page 9]).

The assumption of normality is probably violated

in most cases by biological data, which tends to be skewed

either to the right or to the left (Clark 1961), but not

2 usefulness of.g

so far in most cases as to preclude th

sure of association (cf. Snedecor

as an.approximate mea
1956). In addition to normality, the correlation coeffi-

cient assumes linearity, which is shown in Chapter VI to

ses. But even in the latter

be invalid in a minority of ca
instances the correlations reflect the general or average

slope of the regression. For reasonably rapid elucidation

of the associations between a large number of variables,

therefore,.; seemed the most satisfactory among the avail-

able measures.
So-called phi coefficients ( i ) were ca

those species which were

2. lcu-

lated for all possible pairs of

east the number of quadrats indicated below

present in at 1
least the same number.

for each segment, and also absent in at

standard formulas

2This statistic is defined by the
214 of Yule and Kendall (1940),

found for example on page
page 184 of Fisher (1958), page 234 of Walker and Lev (1953),
and page 200 of Dixon and Massey (1957).

41

 

Minimum No.
of Quadrats
Segment Drainage for Presence
M3 Name Characterization $12151 Absence
G (Grayling Sand) Dry , «a H ‘ 20
WC (Wet Grayling-Croswell Dry-Mesic 20
. Sand)
AS (Au Gres-Saugatuck Mesic 15
Sand)
R (Kinross Sand) Wet-Mesic 12
DC (Dawson-Greenwood Wet _, u 20
Peat)

we shall allow the letters a, b, c, and d to re-

present respectively the numbers of samples having both

Species A and B present, species A only, species B only,

both absent. Then f is calculated by the formula:

§-__

\/ (a + b)(c + d)(a + c)(b + d) ,

ad-bc

found on page 272 of Walker and Lev (1953). It is a prod-

uct moment correlation, actually. a special case of 5 in

which only two categories, zero and non-zero, are recog-

nized.
This reduction to two categories makes necessary

3Definitions of these segments are given in
Appendix A (page 373). me legend for the symbols is

given in Appendix B (page 375).

42

the larger number of quadrats containing the Species than
for the calculation of 3;, where continuous variation is
taken into account. A minimum of five samples in each of
the categories a, b, c and d is generally accepted as ade-
quate; a smaller number in any category may lead to too
large a standard error (cf. Cole 1949). If one Species of
a pair were present in fewer than 12 quadrats, there would
be virtual certainty of having one or more of the four
categories fewer than five; it is for this reason that
such species are not considered in the § calculations.
If a species is present or absent in fewer than 20 quadrats,
in fact, the chances of making such an error are quite
great, but the smaller numbers were permitted on AS and R
because there was unused space on the IBM cards used to
record data for these two segments. Several of the pairs
whose §s were calculated did have one category of 4 or
3, but in no case were the corresponding chi squares sig-
nificant. There seemed to be no harm, therefore, in in-
cluding such §s in the Table VII in order to help round
out the general picture.

The i coefficients are presented together with the
3; values in Table VII (page 91), in the pink-shaded rows.

A few "environmental" variables, with values of zero in the

43

requisite number of quadrats, e.g., cover of moss and
lichen synusia, were included in the § calculations also.

The g coefficient was specifically designed for
answering questions in which at least one of the variables
is of a qualitative nature and measurable only on a nominal
scale, e.g., eyes brown or not brown, male or female. Thus
it might seem a superfluous calculation here, merely a
simplification of 3.; and hence cruder and less sensitive.
In reality, each coefficient answers somewhat different
questions, as explained in Section G of this chapter, and
comparisons between them can be valuable.

3. Significance tests were carried out for the
above coefficients, _t_:_ in the case of 3 and chi-square in
the case of g , to determine the probability that these
coefficients were different from zero. Significance levels
of P< .05 or higher for all coefficients are indicated
after the numbers in Table II with symbols, as exPlained
in the legend preceding that table.

4. A number of special calculations involving the

i and g coefficients were used in analyzing tendencies.

These are explained at the apprOpriate points later in this
chapter.

44

B. DEMONSTRATION OF THE EXISTENCE OF PATTERN.
Perhaps the most striking and convincing result of this
study is the demonstration of the existence of pattern
within each of the five gradient segments, 3.3. , the vege-
tation within each such segment, though relatively homoge-
neous, on a large scale, is evidently heterogeneous on a
small scale. Probably the Spatial repetition of two or
more elements, each tending to be restricted to areally
small microsites, creates the appearance of large-scale
homogeneity.

Considering all soil types and quadrat sizes com-
bined, we find 127. of the _1;s_ and 38.17. of the és signif-
icant at the P <.05 level or higher. Since by definition
we would expect 57. of the possible pairs of randomly dis-
tributed variables to give significant correlations at this
level purely by chance, we have 2.4 and 7.6 times the ex-
pected numbers of significant correlations, respectively.
Even the former figure could scarcely be due to chance when
one considers that some 3700 coefficients were calculated!
At higher levels of significance, moreover, the‘ratio of
actual to expected proportion of significant coefficients
becomes progressively greater so that 4.47. of the 3's are

significant at P < .001, as are 18.1% of the @s at

45
P< .005. These are reSpectively 44 and 36 times the ex-
pected percentages. Practically every individual Species
or variable also exhibits many times the expected prepar-
tion of significance in its correlation with other varia-
bles, and the general picture is similar if prOportions
are calculated for individual soil types and quadrat sizes
separately (Table II, page 55). These high proportions of
significance cannot be attributed merely to a high number
of "obvious" or "meaningless" correlations. After we have
discounted all of these the remaining percentages of sig-
nificant coefficients remain about as high, on the whole.
This can be seen by comparing the subcolumns headed SS,
CX, etc. on the left-hand side of column 3 of Table II
(Proportions of Species-"environmental" correlations which
are significant, including those of synusia with their
major components) with the series 83', CX' etc. to the right
of them (the same, excluding the synusia-component coeffi-
cients).

Some of these correlations may result from reaponses
of Species to gross environmental patterns within the grad-
ient segments, particularly in the three middle segments
which are obviously heterogeneous. Some, however, seem

attributable to the microdistributional phenomena which,

46
because they are generally not consPicuous in the field,
are of more interest to us.

The mere discovery of patterns on a small scale
within vegetation is by no means an original one, as
Kershaw (1963) clearly indicates. Among dozens of papers
published during the past few years in which the authors
attempted to demonstrate pattern on a scale of several
square meters or less in the most diverse vegetation types
imaginable (e.g. Goodall 1952, 1953, 1954, Cole 1949, Dice
1952, Kershaw 1958, Kershaw and Tallis 1958, Phillips 1953,
1954), I have encountered none in which they failed to do
so. It is in its examination of changes in these patterns
as a function both of environmental differences along a
gradient and differences in quadrat size that this study
attempts to reveal previously unreported characteristics
of pattern and to provide useful suggestions for future
investigations .

The phenomenon of the much higher actual-expected
ratio at higher than at lower significance levels suggests
that a relatively high pr0portion of the microdistribution
patterns present are quite marked and distinct. The higher
levels of significance probably represent strong ecological

interactions between variables or narrow tolerance ranges,

L_

47
with strong coincidence (in the case of positive correlation)
or non-coincidence (in the case of negative correlation) in
Spatial optima. The lower levels may reflect broader tol-
erances, weaker ecological interactions and more over-
lapping optima.
C. wUENCE OF SOIL, DRAINAGE. Table II (page
55) was constructed in order to compare the preportions
of coefficients significant on the various gradient seg-
ments and quadrat sizes, for groupings of species with
similar distributions across the gradient and for groups
of "enviromnental" variables which had certain character-
istics in common. A listing of the Species and variables
found in the different categories is found in Appendix C
(page 377), and the criteria used for separating the
Species into these convenient categories is explained in
detail in Chapter IX.
We must caution ourselves, before interpreting the

prOpositions of significant correlations found on each

soil type, that they are not strictly comparable with one
another. The nunber of samples is not constant, and it is
increasingly difficult to establish significance as sample
size decreases. Thus, all other things being equal, we

would expect a far higher preportion of the coefficients

48
on G, on which 120 quadrats were sampled, to be significant
than on R, which includes only 29, but if we find the re-
verse to be true or even the same proportion on both seg-
ments it is noteworthy. G, WC, and DC are roughly
comparable with 120, 112, and 101 quadrats respectively,
as are AS and R with 38 and 29. We must also be aware
that a few of the prOportions listed in the table are
based on such a small number of coefficients that they
cannot be taken seriously. These are indicated in Table
II by parentheses.

The lowest proportion of significance is found
among those relatively rare and/or small Species occurring
on the inorganic soils, most of which possess Optima on
the drier G or WC segments rather than on the moister AS
(called "small solitaries" and symbolized as (SS) in the
table).4 The more clonal and areally extensive of the
Species which possess cover Optima on one of the two best-
drained segments (here designated "clonal xerOphytes" (CX))
seem generally to be more strongly correlated with other

species on the relatively more mesic WC segnent, perhaps

4In order to avoid confusion with the symbols for
the soil types, symbols for these Species groupings will
be shown in parentheses.

49
because their clonal organization itself imposes a non-
randomness upon their own microdistribution which is re-
flected in these coefficients. The positive relationship
between frequency and pr0portion of significant correlation
(next section) suggests this also, for the more frequent
Species are also the more clonal. But these more clonal
Species are scarcely more often significantly correlated
than the (SS). Those Species whose higher cover is found
on the mesic AS ("Margin Drys" (MD) which are also abundant
enough on G for correlation analysis there and "Margin
Wets" (MW) which are not), on the other hand, exhibit al-
most twice as high a proportion of significant coefficients
as the two relatively xeroPhytic group just mentioned (see
grand totals, lower right in (Table II).

In looking at the proportion of significant coef-
ficients for these groupings of Species on different soil
types, one finds some tendency for Species to diSplay
higher correlations at the margins of their ranges along
the gradient than in habitats seemingly Optimal for them.
Thus the bog margin Species, (MD) and (MW), exhibit a much
higher preportion of significance on G, WC and R relative
to sample size ((141)) only on the latter) than on the AS

segment where they reach their maximum cover. 0n R, the

50
(MD) and the distinctly bog Species (B) exhibit prOportions
of significance about twice as high as do the (MW) , which
are apparently closer to their Optimum habitats there than
are the former two. These results suggest that Species on
their Optimum segnents of the gradient tend to have toler-
ance ranges embracing most of the conditions found within
those segments, that they consequently tend towards more
nearly ubiquitous distribution within these segnents, and
thus to circumneutral correlations with other Species.
But on segments which are more "marginal" for them, the
sme Species may be restricted to relatively small portions
of the segment, or may only grow well in these portions.
Accordingly, they perhaps exhibit high positive correla-
tions with other Species which are similarly Spatially
restricted, high negative correlations with those re-
stricted to different microhabitats.

A dramatic instance of such increase in correlation
at range margins is the prevalence of high positive corre-
lations on WC between the clonal xerophytes, (CX) , and the
(MD) and (MW) groups that reach peak cover on the bog
margin. The (CX) are approaching the most poorly- drained
limits of their range, the (MD) and (m) the best-drained

limits of theirs. Hence all may be relatively poor com-

51
petitors on this cannon "meeting ground" and so tend to be
mutually more successful in those microsites most favorable
to plant growth generally, or where the Species dominating
the gromd cover on WC are Sparse (indicating, perhaps,
relative gnfavorableness to plant growth generally). Now
on AS, the mesic segment where the (MD) and (MW) have
their cover maxims, the majority of these correlations
disappear. The (Clo-(MD) and (CID-(MW) correlations shift-
ing in a negative direction by 0.2 or more between the dry
mesic WC and the mesic AS (including positive to less pos-
itive and circumneutral to negative shifts) are about
twice as numerous as those shifting positively by the same
amount (15 against 7, only 2 shifting by less than 0.2).
Moreover, the seemingly less drought- tolerant (MW) exhibit
only negative shifts above this magnitude (9 of them) in
their coefficients with (OK) , whereas about half of the
correSponding shifts in the (Hm-(CK) correlations are
positive (6 of 13). Hence xerOphytes and meSOphytes tend
to produce high cover in the same Spots relatively less
frequently on the more mesic segment, most eSpecially in
the case of more extreme meSOphytes. Possibly the meso-
phytes have a competitive edge on the more poorly drained
Bement and effectively exclude the (OK) from the moister

52
and/or abiotically more favorable microsites by thriving
there themselves. The experiments (Chapter VII) demon-
strate that the xerOphytes grow on these bog margin soils
about as well as on the better-drained segments when such
cwpetition is excluded.

Maximal or minimal conditions may increase the
sensitivity of Species to interactions or small environ-
mental differences, just as reduced vigor in many pOpula-
tions may decrease disease resistance.

Among themselves, the relatively meSOphytic (MD)
and (MW) exhibit a trend Opposite to that of most other
Species, being more positively correlated with each other
in their Optimum habitat, the mesic AS, than on the dry
mesic WC where all are less abundant. This suggests either
that very different factors limit the distribution of dif-
ferent meSOphyte species on the drier segment, or that they
cannot compete very successfully with one another in the
restricted areas relatively favorable to them at their
range margin.

Time and again, we find that Species tend to appear
indifferent to a parameter on those segments where the
parameter's values are low, but are strongly correlated

with it (positively or negatively) where its values are

53
high. This is eSpecially true in the soil horizon corre-
lations. For example, there are overall much Stronger
correlations with soil horizon thicknesses on deep-pro-
files of the mesic AS than on G or WC, and correlations
with bracken, Pteridium Equilimun (253;, no. 35 in the
tables) are much the strongest on WC where cover of this
large fern is so high as to make it decidedly the dominant
among the high herbs. One might expect that if a Species
were positively correlated, for instance, with humus thick-
ness @3195 on AS, it should be even more strongly posi-
tively correlated with it on G where humus is in such
short supply. In fact, however, the latter correlation
turns out to be circmnneutral much more frequently than
not. This leads us to postulate that there may be 5133;};-
ngg. 333., values below which a parameter can have but
little influence, favorable or adverse, upon the growth of
a plant. Thus 10 um. or 20 nm. of humus may make little
difference to a plant because in any case there is an in-

sufficient volume of this organic material in which roots

5Abbreviations used for Species and other variables
used in the tables are listed in Appendix B (page 375).
They may be distinguished from symbols for gradient segments
and species groups by their being underlined and consisting
of three rather than one or two letters or other symbols.

54

of a species favored by it can grow and proliferate. But
30 mm., for example, might be the minimal thickness in
which roots of such a Species could grow without becoming
dessicated by the sun striking the ground surface. Then
60 um. would provide twice as much root-expansion volume
as 30 m. and accordingly favor more vigorous shoot pro-
liferation.

As a demonstration of the value of examination of
correlation and other information from different parts of
the gradient in Specific cases, let us examine the pair
M pensylvanica (Car) and Epigaea £253; (£219 . From
Table VII (page 97) we see that on AS, the correlation
Q1) between this part is 0.48 and 0.55 on 1 m2 and 1/4 an2
scales reapectively, both positive and highly significant
(P< .01 and .001). But the correSponding correlations on
the better drained segments are circumneutral, even slightly
negative (non-significant) on WC.

We see from the sunmaries of per cent cover across
the gradient (Table 1, page 27) that 22.8.1.2. cover is
greatest on well-drained C while Epigaea produces its
greatest growth on the moister AS. This suggests that the
first is a relative xerOphyte, the latter a meSOphyte, so

the circumneutral correlation between them on the "dry end"

 

 

 

 

 

 

 

 

 

I

\x i

 

 

 

 

 

 

 

56

(which might be negative were not both scarce in Sparsely'
vegetated patches) is not surprising. Possibly, then,
their positive correlation on the bog margin results from
both growing relatively poorly on the more poorly drained
Saugatuck portion of this soil deSpite Epigaea‘s apparent
meSOphytism; their maximum tolerances for moisture or
flooding may coincide though their Optima are decidedly
different. Another possibility is that certain other
Species, at their Optima on this part of the gradient and
so vigorous competively, preempt large parts of the area,
mutual absence of M and Epigaea from these parts
making their correlation positive. But the latter Species
do not both exhibit any high negative correlation with the
same third Species, so the first alternative appears more
plausible. By comparing correlations and cover on several
soils we have avoided some possible pitfalls, that M'
and Epigaea's Optima for moisture, humus, or something
else coincide, for instance, or that they interact posi-
tively. In these last cases there would probably be at
least a suggestion of positive correlation on the "dry
end".

If we compare the gradient segments themselves

with each other in regard to the proportions of significant

57

correlations found upon them, we see from Table II that R
contains a higher prOportion of significant correlations
at all four levels than does AS. Since there are about
one third more quadrats on A8 than on R, so that coeffi-
cients of somewhat lower value are significant on the
former, the higher prOportions on R take on added import.
This implies that R, is indeed more heterogeneous than AS,
as suSpected from field observations that the former con-
tains both inorganic soil and bog elements, and also that
its tipup-mound microrelief is much more extreme than AS'S.
Such heterogeneity would tend to make all but the most
broadly-ranging Species Strongly predominate in distinct
patches of one or the other environmental types represented,
and hence give rise to many high correlations. Similarly,
among the segments whose sample sizes are large WC, with
about 72 fewer quadrats than C, appears to support the
stronger correlations, possibly because it contains both

dry end and bog margin elements. But DC, with fewer quad-
‘ rats than C or WC, is the locus of considerably higher
percentages of significance at all levels than either of
the latter. In the case of DC, however, it is probably a
sharply defined small-scale heterogeneity caused by the

marked hunnock-and-hollow microtOpography found on that

58

segment (and discussed in more detail elsewhere) which is
the cause. The DC segment, on a scale several times that
of the largest quadrats appears more homogeneous than any
other, and it embraces neither broad vegetational extremes
nor conSpicuouS elements of other segnents. The absence
of drainage-influencing SlOpe may be an important factor
in this.

D. RELATION BETWEEN FREQUENCY 0R COVER AND
STRENGTH OF CORRELATION. It appeared frequently during

 

the analyses that the more abundant Species or those with
the highest cover might exhibit the strongest correlations
overall. AS a partial test of this question, a series of
linear regressions of the standard form,

y - a + bx,
whose computational formulas are given for example on pp.
230 ff. of Walker and Lev (1953), were calculated for the
data from gradient segnent G (Grayling Sand). In these
regressions, frequency or cover of Species (dependent var-
iable y) were plotted upon per cent of all correlations of
the Species Significant andependent variable x). Letter a
represents the y - intercept and b, the regression coeffi-
cient or slope, the increase in y per unit increase in x.

This was repeated for each of four different significance

59

levels of the correlations, both for the coefficients from
all three quadrat sizes treated as a single pOpulation and
for those from the l m x l m size only. Relevant results
are presented in Table III, page 60 ).

At the three lower Significance levels (P < .01,
.02 or .05) frequency of a Species and its correlation
strength, as measured by the prOportion of its coefficients
significant, do indeed seem to be positively related as
predicted. This is eSpecially true when coefficients from
all three quadrat sizes are considered together, and the
regression of frequency upon per cent significance at
P< .05 for all quadrat sizes, is itself significant at
P< .01. There are no other significant regressions but
their overwhelming positive trend is fairly convincing.
Per cent cover and per cent significance in all three
quadrat sizes are Similarly positively related (the low
magnitude of the b's being due to the relatively low values

2 scale

Of cover as compared to frequency). But on the 1 m
alone, the regressions of cover on significance prOportion
range from flat to quite negative.

The import of these results would seem to be that

the more commonly encountered species and those with the

highest cover, which are usually one and the same, tend to

60

TABLE III

The regression coefficient (column 4) of frequency and
cover of species on per cent of correlation coefficients
of these species significant on gradient segment G
(Grayling Sand). The significance levels at which these
per cents were calculated are listed in column 2, the
quadrat sizes which were considered in column 3.
Symbols used are l = lmz, 4 =1/4m2, H =l/100m2.

 

1 Z 3 4 5
y (abundance Significance Quadrat Significance
criterion) level, of cor- Sizes b of b (dash =
relations, P\< P >. 05)
Frequency 0.05 l, 4, H + 2.68 P<.01
(p coefficients) 0.025 + 2. 29 —~
0. 01 4— 2. 41 ~
0.005 -- 0. 57 _
0. 05 1 only + 1.09 -
0.025 + 0.63 —
0. 01 -1- 0. 13 —-
0.005 4— 0.00 ~
Per cent cover 0. 05 l, 4, H + 0. 13 -
(_z_'_ coefficients) 0.02 + 0. 05 ——
'0. 01 ~4— 0. 07 —
0.001 -- 0. l9 -—
0. 05 1 only --— 0.. 00 -—
0.02 -- 0. 07 -
0. 01 — 0. 12 —
0.001 - 0. 12 -

61

exhibit the strongest correlations as postulated. Since
these Species are generally the more clonal ones, their
wide sphere of influence upon other Species may bring this
about. But the decrease in positive slope of the regres-
sion upon significance proportion at progressively higher
significance levels, and the negative regression of both
frequency and cover on these proportions at the highest
significance level of the correlations, (P< .001) appear
to reflect a counteracting force. This force probably
consists in the Specialized requirements of the relatively
rare Species which exhibit a somewhat higher prOportion of
m high correlations than the large, clonal, extensive
species. Perhaps the highly significant correlations of
the rare Species reflect their specialized requirements,
environmental Specificity and dependence upon the more
robust and ubiquitous elements in the conlnunity. The
higher prOportion of low-value correlations exhibited by
the more ubiquitous Species may reflect their relatively
broad tolerances and non-Specificity of influence.

The overall negative regressions of cover on Sig-

2

nificance prOportion data for l m quadrats alone seem

unaccountable to anything other than small sample size.

62

E. W. The measurement of
correlation on the basis of data from different quadrat
sizes should reveal something of the Spatial scales of
patterns present (Goodall 1952, Kershaw 1963, 1964, Greig-
Smith 1964 among others).

Graig-Smith (1964) and Kershaw (1964) point out
that we should expect more and more Species to be posi-
tively correlated as we increase the size of the sample
units, and conversely fewer positive and more negative cor-
relations as we decrease unit size. On the basis of quad-
rats 30 or 40 miles square randomly located over the whole
surface of the earth, such milikely things as wheat and
spruce trees might be positively correlated while inter-
tidal organisms would be found to be associated with
benthic forms of the same latitude. Diminishing the
Scale toward the other extreme, we would reach a point
where the organisms were larger than the sample quadrats,
so that the probability even of parts of two individuals'
in the same quadrat would be small. In this case we would
obtain predominantly negative associations. As Greig- anith
(1964) theorizes, and as Kershaw (1961) found in practice,
positive correlations should then generally shift in a

negative direction as we reduce the size of the sample

63
units, because of both this factor and the increased inten-
sity of competition on a small scale.

Figure 3 (page 64) illustrates, for the data of
this study, the influence of quadrat size upon strength of
positive or negative correlation. The bar graphs Show the
frequency (y-axis) with which correlations of variable
pairs, based upon data for quadrats of a given Size, ex-
hibit positive or negative shifts of various magnitudes
(x-axis) from correlatiOns of the same pairs based upon
larger quadrats. These graphs are based upon data for all
gradient segments combined. For example, eleven variable
pairs have correlations based upon data for the l/4 m2
quadrats which are between .23 and .27 (inclusive) alge-
braically lower than their correlations based upon l‘mz
quadrat data. Accordingly the bar between - 0.23 and
- 0.28 Shaded in Figure 3A has a height of 11. The values
of the correlation coefficients themselves have no effect
upon this graph. Three variable pairs with correlations
in l m2 and it» m2 quadrats reapectively of + 0.90 and
+ 0.65, + 0.15 and - 0.10, and -0.15 and - 0.40 will all
have a shift of - 0.25, and accordingly all will be re-
presented in the - 0.23 to - 0.28 bar.

Table IV (page 65) lists the numbers of positive

 

 

 

 

FIGURE 3.

Shifts 1.3 correlation with guadrat size

The horizontal axes represent the mag-
nitudes of changes in the correlation coef-
ficient between particular variable pairs,
the vertical axes the number of variable
pairs exhibiting changes of these magnitudes.

A. Changes between 1 m2 and
1/4 m2 quadrats.

B. Changes between 1/4 m2 and
l/ 100 m2 quadrats.

C. Changes between 1 m2 and
1/100 m2 quadrats.

 

64

N0.0F VARIABLE PAIRS

 

 

 

 

 

 

 

 

 

 

 

 

 

   
    

 

 

 

      

use]
a. T)
I
I
a I
zoo I
I
ITO-I :
I
I
ISO-I I
IT ”4270 I/4 M2
Ill-I :
I
I
Ioo-I ”‘1 I
4 I
I
I
m I
I
I
504 I T
I
I
' A
25"I I
I FT—l—fl'fii
I
I
0 t r ' f r 1 r t T r 1—
-.7s 0.5. -.3s -.Is 0 J .I 7 «.37 $.57 .t .77
MAGNITUDE OF SHIFT
N0.0FVARIABLE PAIRS
75"1
sea
T1 I/4 IIIz TO I/Ioo mg
I
I
2&4 I B
I
I
I m
0H , 7 r I * I I i f V j *1—
-.7s -.ss -.ss -.IS IS {.17 .I.37 «.57 L77
MAGNTUDE OF SHIFT
50 no. or VARIABLE PAIRS
2 I III2 TO Moo .2
I
I
: C
. .--. m .
9 I I , 1 1 I [ * r " I ' V— T— W T I I f
-.7s -.ss -.3s -.Is 0 «IA? 4.37 LS? 4.77
MAGNITUDE OF SHIFT

65

and negative changes above certain arbitrary values, in
order to call attention to a fact not imediately evident
from looking at the graphs; that the really large shifts
with decreasing quadrat size are indeed predominantly neg-
ative. The tendency is, however, not nearly so pronounced
as Greig-Smith's reasonable hypothesizing might lead one
to expect, and seems to occur primarily between the l m2
and 2]? m2 quadrat sizes with little if any change between
the latter and 1%6 m2.

TABLE IV

Correlation shifts between large and Smaller quad-
rat sizes. EXplanation in text.

Magnitude Numbers of correlations exhibiting shifts of

of shift the magnitudes and signs listed at the left
between the quadrat sizes indicated

2 2 2 2 1 2
lm to? in toi-Efi-dmz lm tomm

 

 

- + + - +
0.30 or over 19 11 10 12 18 14
0.20 or over 58 28 17 25 36 23
0.15 or over 79 47 30 30 58 33
0.10 or over 137 92 46 43 70 49
mean mean mean

shift- -0.013 shift.I +0.007 shift"I -0.0l7

66

The modes, or values exhibited by the largest num-
ber of correlation pairs, are very slightly on the positive
side of zero, and small shifts are seen to be predominantly
positive in Figure 3. But this is counterbalanced by both
greater numbers and higher values of negative shifts. This
combination makes for nearly equal area under the curve on
either side of the mean, and hence for mean shifts (bottom
of Table IV) as well as coefficients of skewness, g3 (Clark
1961) not significantly different from zero and, in fact,
very close to it numerically.

A possible cause of the slight predominance of
small positive shifts in correlations with decreasing quad-
rat size, despite the tendency of larger shifts to be neg-
ative, might be a pronounced tendency of Species to be
grouped together on a small scale. The decrease in nega-
tive shifting between the 1/4 m2 and the 1/100 m2 quadrats
tends to corroborate this hypotesis. Such small scale
grouping might be largely independent of larger- scale pat-
terns or groupings, which would tend to break down on a
smaller scale due to spatial limitations and competition
as discussed above. Thus, it is possible to postulate pat-
terning forces of at least two distinct scales in the eco-

systems of the study gradient. The larger may be large-

67
scale heterogeneity within gradient segments such as parent
material or soil moisture differences, whereas the latter
may be related to changes in exchangable soil ion concen-
trations brought about by the species which recently occu-
pied a microsite, or to the approximate area occupied by
interacting species themselves. At any rate, the smaller
scale patterns would appear to be stronger and more specific
than the larger ones, for the Greig-Smithian factors predis-
posing toward negative shifts should otherwise tend to move
the modes as well as the longer and areally larger tails of
the Figure 3 curves to the negative (left) side.

As an example of the uses to which correlations
from quadrats of different sizes may be put, let us exam-
ine the pair @313 hisp'idus var. obovalis (_R_h_o, no. 36 in
the tables) and Vacciniun boreale (m, no. 47). On WC
the correlations are + 0.01 on the basis of data from 1 m2
quadrats, - 0.00 from 1/4 m2 quadrats and + 0.35 from
1/100 m2 quadrats; on AS the respective correlation are
- 0.13, - 0.09, and + 0.47 (Table VI, page 114). The high
positive correlations on a 1/100 m2 scale, significant at
the P< .001 and P< .01 levels reapectively on the two
segments, are in vivid contrast to the circumneutral ones

from larger quadrats. This combination rules out the pos-

68
sibility that the correlation is due to mutual Optima with
respect to any environmental variable acting on a fairly
broad scale, such as tree cover or parent material textural
variations. An examination of the correlations of _13_u_13_u_s_
and Vaccinium with other variables and species reveals none
with which both have a significant positive or negative
correlation on both soil types, so similar responses to
such a species or to one of the environmental variables
measured is also ruled out as a possible cause. Moreover,
flacciniun is a shrub and Lubgg a repent vine, and both form
large clones and spread by rhizomes. Thus the microsite
where the cover of both is measured in a 1/100 m2 quadrat
is usually some distance from the bulk of the underground
parts of each, and a correlation between them on this
scale could scarcely be the result either of interactions
between them underneath the ground, or of mutual responses
to a soil factor varying on a small scale. By a process
of elimination we are alm03t forced to the conclusion that
a positive interaction is occurring between the 2.9.122}.
parts of the two Species, and that where a m vine
passes beneath a lush Vaccinium clump it puts forth more
and/or larger leaves. At present we do not know whether

this might be related, for example, to a simple stimulation

69
of _l_h___1b__u_s_ leaf expansion in dense blueberry shade (a re-
sponse similar to "shade-leaf" formation in many trees),
to a metabolic exudate released by one of these species
and stimulatory to leaf and shoot growth of the other upon
contact, or to some other factors perhaps beyond our wildest
imagination. What is important is that by examination of
correlation behavior in quadrats of different sizes, we
have considerably narrowed down and crystalized the field
of possible causes. Some now absurd ones might easily have
seemed acceptable postulates had we possessed only the in-
formation from the smallest quadrats, and the relationship
would not have come to our attention at all had we used
only one of the larger sizes.

The information from two gradient segments has been
useful in this case also, but here the two coefficients
reenforce one another, instead of indicating a change as
do the examples cited in Section C.

F. EREVAIENCE 0F POSITIVE CORRBIATIQN. Almost as
striking as the unusually high prOportion of significant
coefficients is the undeniable predominance of high posi-
tive over high negative correlation values. If we count
those coefficients from all three quadrat sizes whose ab-

solute values are 0 . 2 or over , (which approximates the

7O
lowest value significant for any segment), we obtain the
figures presented in Table V below. The complete distri-
butions of coefficient values are illustrated in Figures 4
and 5 (pages 76, 77).
TABLE V

The numbers of coefficients for all quadrat sizes
of 0.2 or over in absolute value which are positive and
negative, the mode and the mean of all coefficients on each
soil type. See also Figures 4, and 5, pages 76 and 77.

5?

r
Soil Type G WC AS R DG Total Tota

__

Coefficients +0. 2

or more positive 77 125 164 101 26 493 126
Coefficients ~0.2

or more negative 21 20 91 65 21 218 5].
Mode of all

coefficients -.03 -.03 -.08 -.12 -.05 --- ---
Mean of all

coefficients +. 020 +. 029 +. 005 +. 007 - . 015 --- + . 067

M

Overall, over twice as many of the 3's above 0.2 are on the
positive side as are on the negative. Similarly, the num-
ber of positive fie larger than 0.2 is about two and
one-half times the number of negative ones. By contrast,
the modes, _i_._.§_. the intervals which contain the most coef-

ficients, are slightly negative on all five gradient seg-

71
ments. This means that, among coefficients of low value,
there are many more negative than positive ones. In com-
bination with the prevailing positiveness of high values,
this accounts for the nearly neutral mean correlation values
on each soil type, and for our inability to statistically
demonstrate skewness in the distribution of coefficient
values just as was the case with the distributions of cor-
relation shifts with quadrat size. (Here, however, the
curves are mirror images of those of the shifts, with the
right-hand tail longer, the left-hand tail higher near the
zero point).

The striking tendency towards positive correlation
could possibly be due to a clustering or overdistribution
of the vegetation as a whole, such as is visually evident
in the field. Lush and sparse multispecific patches would
correspond to a mosaic of microsites relatively more and
less favorable for growth of vascular plants generally.
Such a mosaic could theoretically owe its presence to the
distribution of some dominant and inhibitory species, but
such a species should exhibit predominantly strong negative
correlations with others. Only on one segment, DC, is
there in fact a dominant species which does (Chamaedaphne
W . On G and WC dominant bracken (Pteridium

72
aguilinum), which is known to inhibit other Species in an
antibiotic fashion and which may leave behind a concentra-
tion of inhibitory substances in patches where it has
recently died (Watt 1947), shows only a weak negative trend
in its coefficients. Nonetheless, the antibiotic-impreg-
nated patches need not necessarily be strongly correlated
with patches of living Pteridium, and hence might be the
sparse patches in a vegetational mosaic deSpite living
bracken's lack of a striking negative correlation trend
with other species.

But there are no plausible causative species on the
remaining two segments where gteriditg and Qhamaedaph’ng are
sparse or absent, nor in fact do any of the measured phys-
ical factors exhibit the tmusually strong trends toward
high positive or negative correlation which one would ex-
pect of a dominant patterning force (grunus serotina dem-
onstrates possible antibiotic activity on AS, but is
scarcely widespread enough to account for the positive
correlation trend). It thus appears, then, that the pat-
terning force consists of some unmeasured abiotic parameter,
or in the distribution of unmeasured organisms such as soil
microbiota or insect larvae.

The negative modes may reflect the same general

73
tendency towards negative correlation on a small scale due
to interspecific competition and Spatial exclusion men-
tioned by Greig-Smith, and postulated above (Section E) as
a reason for the predominance of high negative over high
positive correlation shifts with decreasing quadrat size.
They may, on the other hand, be a result of large-scale
patterning which, as mentioned in Section E, perhaps pro-
duces relatively weak correlations. If so, then it also
tends to produce negative ones, while the correlations
attributable to the more sharply-defined, hence high cor-
relation-producing forces on a small scale are mostly posi-
tive.

On those gradient segments which contain much
Sparsely vegetated space and where such spaces are rela-
tively large, namely G and WC, the number of large positive
coefficients according to Table V is respectively about
four and six times as large as the number of negatives.
But on segments where vegetation is more uniformly dense
and competition would appear to be at a maximum we obtain
less than twice as many positives (as on A8) or only a few
more (as on DC). The modes on the latter two segments,
moreover, are more negative {-.08 and -.05 respectively)

than those on the two better drained, more sparsely vege-

74
tated segments (-.03 on both), and considering the several
himdred coefficients contributing to each of these figures
it is improbable that the differences are trivial. This
is just what we would expect; the most marked positive
trend should occur where a force like spatial or moisture
competition, predisposing to negative correlation, is rela-
tively important (although under extreme drought conditions
moisture competition might be more vigorous on the "dry
and" than elsewhere). A relatively high proportion of
slight to moderate negative correlation on R and the lowest
mode of all here (-.12) may be due, not to competition, but
rather to the great heterogeneity of the physical environ-
ment here.

We can now utilize the predominance of positive
correlation in conjunction with information previously
discussed in order to demostrate additional aspects of
pattern in the study area.

Overall, proportions of significant coefficients
decrease on the smaller quadrat sizes (Table II, page 55).
Since most significant coefficients are positive, including
nearly all on A5 and R, the negative shift with decreasing
quadrat size discussed in the preceeding section would

tend to make most coefficients less positive and hence less

75
significant as this size becomes smaller.

If we now examine the behavior of different physio-
ecological groupings of Species separately,'we find the
bog species (B) on DC exhibiting by far the greatest de-
crease in correlation significance proportions between 1 m2
and l/lOO m2. The mesOphytic, mostly clonal (MD) and (MW)
as well as the larger clonal xerophytes (CK), each exhibit
a somewhat weaker tendency of the same sort on at least
one soil type, but for the small, mostly non-clonal (SS)
Species of the inorganic soils there seems to be no such
trend whatever.

The lack of any apparent general trend as to the
part of a species' range where the greatest quadrat size-
related significance decrease is found makes interpretation
of these results difficult. The (CX), (MW), and (B) ex-
hibit their greatest decreases on their Optimal segments,
WC, AS and DG respectively. The (MD), on the other band,
decrease most on G, the xeric end of their range. The
better the drainage where e.grouping has its optimum, the
less the tendency of its positive correlations to decrease
in value as quadrat size does, but this may be a coinci-

dence .

The (B), the only species exhibiting a really

' *‘ .‘ -.—

FIGURE 4.

 

Correlation Values for each Segment

The number of correlations within.eadh
value range are indicated by the height of
the bars. Diagonal hatching - species-species
correlations, clear - species-environment
correlations, black - environment-environment
correlations.

A. Correlations on Segment G, E.

B. Correlations on Segment WC, ;_

 

 

76

no. or cosrrncusurs

 

 

 

 

 

 

 

 

   
 

 

 

A m

u l...
lx/////////////

I lllll wVV/Vflflxx/x xxVxxxxx/ flV/xflflfllx . o
x//////////////.V//////////////////////
a . a m a W . m M.

VALUES

NO. OF COEFFICIENTS

 

SEGMENT

WC

 

 

 

 

 

x/

l.|7

 

 

FIB-T

73-1

504

{.57 11.77

.37

'3' '0"

n3.

VALUES

fl...

 

 

 

FIGURE 5.

Correlation Values for each Segment, cont.

Explanation and legend as for Figure‘4,

page 76.
A.
B.
C.

D.

Correlations on Segment AS,.£

Correlations on Segment R,‘5
Correlations on Segment DG,,£

Correlations on all segments
combined,jf

 

 

77

NO. OF COEFFICIENTS
l7!-

 

I25-

SEGMENT
AS

 

BO-

 

 

25—

\\\\
x>

 

 

 

 

I ,
-.78 -.68 -.38 -.ID 0 4J7 {.37 {.57 {.77

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

VA L U E S
no. or coenucnsurs
n-
so. 3 sousur
R
20- : 8
.J
-.n -.sa -.3s -.l8 0 J.” {.37 £57 {.77
Value. OFCOEFPCIENTS V A L U E S
I
004
sesueur
25‘ I . no
/' ‘ c
0.] / ' k
I I
-.n -.5s -.3e -.u'a 6 4.17 237 x." x."
V A L U E S
no.0: coerncnsurs
I

 
 

COEFFICIENTS

  

 

78
strong decrease in significance in smaller quadrats, are
also the only species which exhibit a high proportion of
significant negative correlations. As we can see from
Thble VI (page 79), the prOportion of negative coefficients
does not change appreciably or even decrease slightly with
decreasing quadrat size, although on the basis of prior
results we might reasonably expect it to increase. But
the prOportion of positive coefficients for the smaller
quadrats decreases tremendously, a probable consequence of
interspecific inhibition on a relatively small scale by
clonal shrubs such as Chamaedgphne. The large number of
high negative correlations may owe itself to this inhibi-
tion also. The relatively large scale of the hummock-and-
hollow microtOpography, probably the primary "environmental"
Pattern-inducing factor here, perhaps accounts for the
hiSher‘prOportions of positive correlations on a larger
scale.

G. COMPARISON OF 2 AND 1' COEFFICIENTS. The much

higher prOportion of significance among the g; coefficients

 

 

than among the as is striking. On G, WC and AS there are
considerably more significant 133 at each significance
level and only on DC are the proportions about the same

(Thble 11, page 55 )(the prOportions on R cannot be taken

79
TABLE VI

A comparison of the percentages of positive and
negative coefficients which are significant on the DG gra-
dient segment.

A ”Ma-H- -.-.. .. ~—4..—9- _ __.¢ —. ~1.- .._..

 

 

914 What size
Significance 1 m2 1/4 m2 1/100 m2
level, p< + - total + - total + - total
0.05 12.7 5.5 18.2 5.2 8.8 14.0 2.6 5.3 7.9
0.025 5.4 5.5 10.9 5.2 5.3 10.5 1.4 3.9 5.3
0.01 5.4 5.5 10.9 3.5 3.5 7.0 1.3 1.3 2.6
0.001 0.0 3.6 3.6 3.5 3.5 7.0 1.3 0.0 1.3

 

 

 

 

 

seriously because of the small number of coefficients cal-
culated there). Part of this may be due to the predomi-
nance of the more abundant Species in the { coefficient
calculations, and to these species' tendency to have

higher 53 than the rarer species also (see Section D of
this chapter). Yet even among just those species for which
is were calculated on each soil type and quadrat size,
there are nearly twice as many significant is as £8 at
corresponding significance levels (bottom of Colman 9,
Table 11, page 55).

Probably this again reflects a mosaic of patches of

80
lush and patches of sparse vegetation generally. If we
postulate such a mosaic then, in the Sparsely vegetated
microsites, only the most common species may have much
probability of occurring; accordingly most of the species
may be clustered together in the lushly vegetated spots,
mutually absent from Sparse locations, and so are posi-
tively correlated on the basis of presence. But within
each lushly vegetated microsite, interSpecific competition
and Spatial exclusion may tend to make the cover of Species
inversely prOportional to one another, as already explained
in other connections. Now, most of the significant corre-
lations on the inorganic soils are positive (the import of
which was discussed in Section F above) so that the prOpor-
tion of significant correlations is a good measure of the
prOportion of high positive correlations. Thus, correla-
tions based upon cover, although overwhelmingly positive
on the whole, are numerically lower than the f3 and
hence less liable to be significant. The similarity in
the prOportions of significant _r_s and 9’s on DC may be
due in part to the absence of extensive sparse patches,
and to the absence of slape or level differences with re-
Spect to the water table making for nearly ubiquitous pres-

ence of the cannon species on this segment. Both conditions

81
would tend to lcmer the proportion Of positive and signif—k
icant is. At the same time, the rather distinct and
sharp1y-defined nature of the microtOpography and its re-
lated clear-cut vegetation mosaic, plus the apparent in-
hibition effects of lush bog shrubs, may tend to increase
the numbers of both high positive and high negative rs.

It is evident, then, that the g and _r_ calculations
do not yield the same information and do not duplicate one
another, although if we allow the two types of coefficients
for each variable pair themselves to be variables the cor-
relation between them comes out very highly positive
(£513 - 40.70, P (.001). The f coefficients, by merely
considering presence or absence, without regard to the
amount of cover, are more inclined to reflect preferences
of species with regard to broadly-varying factors, such as
soil moisture or texture or minerals perhaps. The as , on
the other hand, may reflect competition or interaction be-
tween species, as well as response tO factors varying
sharply on an extremely small scale, to a much greater
degree. This is interrelated also with quadrat size
(Section E of this chapter).

As another illustration, if two species possessed

quite different microhabitat Optima but a great deal of

82
niche overlap, then the two would occur together much Of
the time, perhaps giving a positive é. , but the per cent
cover of each would be high in different microsites giving
a circumneutral _r_. Or a combination of circumneutral fl ,
negative _r_ would be possible under similar circumstances.
Thus we see that similar f - _r_* combinations could be due
to quite different sets of circumstances.

There is in addition a substantial minority of
instances in which either 3;. is greater than if positively,
or where _i; is circumneutral or even positive while i is
substantially negative. A good case in point is the pair
Gaultherig procunbeng (952, no. 22 in the tables) - 3115.1"
gnthemum canadense (_M_a_i_, no. 28) , which on both 1 m2
and 1/4 m2 scales, on both C and WC, yields quite positive
and highly significant _i; values with P (.01 or .001 (see
Table VII, page 106). The corresponding fl coefficients,
while also positive, are consistently much lower in value
and significance (one of them not even significant at
P < .05). The situation here may indeed be somewhat the
Opposite of the one just mentioned; the two species may
have a relatively small area of overlap within the range
Of habitat conditions under which they have a high proba-

bility of growing, so that Maianthemum is frequently found

83
without Gaultheria and gaultheria without Maianthemg.
The area of overlap, however, could include the cover Op-
tima Of both (observations suggest the remains of Old logs,
stmnps, and other large bodies of decomposing organic mat-
ter in this case), giving rise to a correlation based upon
amount of cover which would be more highly positive than
one based upon simple presence.

H. "ENVIRONMENTAL" PARAMETERS. It is incorrect
to arbitrarily designate particular variables "environ- ‘
mental" when, in fact, every species is in some degree
part Of the environment of every other growing within its
sphere of influence. In fact one group Of variables which
are here called "environmental", the layers of vegetation
or synusia,6 are themselves made up of species. They are
included in the analysis because frequently they may have
an influence on members of other synusia which are not
evident from individual species correlations. Also in-
cluded are characteristics of the litter, non-living but

h

6A8 designated here, tree and shrub synusia are
separated on the basis of height (below 6 feet for shrubs).
Those woody species with a repent habit are included with
all the species having non-woody stems in the herb cate-
gory, whereas all other woody plants less than 6 feet in
height are designated shrubs.

84
organic, and the soil, most of whose mass (but not neces-
sarily influence) is inorganic. The epithet "environ-
mental" is patently one of convenience, and should not be
thought of as implying any fundamental differences in the
ways in which "environmental" factors as Opposed to other
species impinge upon organisms.

The "environmental" parameters which at first
glance seem to exert the Strongest influence upon species
distributions are the synusia ((SY) Column 2 in Table II,
page 55). However, a considerable proportion of the high
synusium-Species correlation is due to the fact that the
correlated species themselves constitute these synusia.
For example, the positive correlation between shrub cover
and Chamaedaphne in the bog is nearly unity (+0.83), which
looks most exiting until we realiza that this species,
making up over 80% of the shrub cover here, is practically
synonymous with the shrub synusimn. If we now consider
only those species-synusium correlations for which the
Species is not a member Of the synusium with which it is
correlated or does not comprise over 207. of that synusium
in more than two quadrats, we Obtain the proportions listed
just to the right of the synusia. (SY) in Column 2 of Table

II, namely (S-). (S-), it should be stressed, is no perfect

85
gauge Of the relationship between variation in the synusia
and in the cover of Species, either. For one thing, there
are almost certainly relationships between cover of synusia
and that of the Species which comprise them aside from the
participation of the latter in the former. For another,
the Species for which correlations have been eliminated
are naturally the more comon ones, leaving a disproportion-
ately high number of correlations with relatively rare
Species. Now, as discussed in Section D above, the rarer
Species tend to exhibit the lower proportions of correla-
tions. The combined effect may be to make the adjusted
figures lower than they should be.

In so far as these figures can be trusted, it is
evident that the synusia have considerably less influence
upon the microdistribution of Species than might have been
assumed. On G, WC, and DC the corrected prOportions of
significance in these synusium-species correlations ((S-),
Column 2 of Table II) are now somewhat lower on each seg-
ment than similar means for all correlations ("tot" in
Column 4), and on AS they are only about one-third as high.
Only on R do synusia seem to be strongly correlated, prob-
ably because Of the heterogeneity in that segment, and the

great variation in tree and shrub cover associated with

86

the coming together of upland and bog elements. But based
upon these adjusted prOportions Of significance, the
synusia do seem to exert (or merely predict?) as much pat-
terning influence as the horizons overall ((HO) in Column
2) and somewhat more than the litter parameters (LI).
Synusial correlations appear particularly strong on G and
WC.

Soil horizon thicknesses ((HO), Column 2) seem to
be more closely related to species distributions on WC
than elsewhere. This most probably stems from the rela-
tively higher cover of vegetation as a whole on relatively
deep-horizoned Croswell, that part of WC which borders on
A3. Also, there is a surprisingly large prOportion of
hi8th significant coefficients with the horizons on DC,
the bog soil, which means here entirely with the thickness
Of the peat itself. One would think that here, where the
Organic matter is several times thicker than most roots
can penetrate, its thickness would be less related to
Species distributions than elsewhere. Also, the zone of
Possible root penetration or shading by upland Species is
relatively small. But concentrations of mineral nutrients
may be differentially stimulatory to different Species,

and may be inversely proportional to distance to mineral

87
substrate below from which they can seep upward, or to
upland soils from which they can wash and seep laterally.
Or possibly the patterns reflected here are relict ones;
the peats farther from the margin, though thicker, are
younger and in the more recently filled-in part of the
lake basin. Conversely, the marginal peats are more likely
to have been dry at the time of fires, and so to have been
more recently replaced. Local differences in the magni-
tude of surface level fluctuations, such as those described
by Buell and Buell (1941), may also have their effect.

But overall, the prOportions of Species-environ-
ment coefficients significant are generally smaller than
those of Species-Species correlations, particularly at the
high significance levels (P< .01 and .001). This is true
for each gradient segment as well as the five segments com-
bined, and suggests that these parameters have rather less
pattern-inducing influence than species upon other species.
If we now examine the significance of correlations of
"environmental" parameters with other such parameters
(Column 3 Of Table II), however, just the Opposite seems
to be true, both generally and for Specific categories of
parameters and segments.

As a corollary of the last-mentioned trend, the

88
cover of synusia should be more Strongly influenced by
the distribution of other synusia and other "environmental"
parameters than are the individual Species which constitute
them. Thus shrub cover may be reduced under tree cover,
herb cover reduced under shrub cover, or either increased
in patches of relatively thick humus, etc. Competition
and Spatial exclusion of Species 1.43.1112 these synusia may
tend to lower individual Species -"environment" correla-
tions.

The importance of "environment" in the microdis-
tributions of particular groupings of Species can be in-
ferred from the prOportions listed in Columns 3 and 4 of
Table II, uncorrected and corrected for "ridiculous" cor-
relations reSpectively. The "small solitaries" (SS) seem
to be relatively little influenced throughout. The more
abundant "clonal xerophytes" of well drained soils (OK),
on the other hand, exhibit a prOportion of significant
correlations with "environmental" variables two to four
times as high as do the (SS). This parallels the statis-
tical behavior of the two groups with other Species, but
here the difference is more marked. Similarly, as in their
correlations with other Species, those species with cover

Optima on the more poorly-drained soils, (MD), (MW) and

ill—u—ui'V“.E—'flflf'

 

 

 

 

 

—e

 

....— ..__x___

warm——

 

I- (J
\.I| y.
' a Q
.

C. ".e

eaves I
l .
it..- h -

t
.
F's-1...
“5:...

v
I. _
.“F 1w

.
Yl"r..
A ..t_‘
a...

. Ive
‘4‘."
o

" e
I .
w "~

Vi .5
N.

D
I
A...
n... .
‘-
.
‘A I
Hi.
‘._..f
‘v
.
I\
., -- ‘

89

(B), seem from the uncorrected prOportions (SS, CX etc.)
of Colmnn 3 more strongly correlated with "environmental"
parameters than are the aforementioned xerOphytes, esPe-
cially far from their Optimal habitat as on G or WC. But
this difference disappears when the corrected proportions
are considered (88', CX' etc., Colman 3), suggesting that
it reflects nothing but correlations of synusia with their
components. The relatively high corrected prOportions for
(MW) on R are based on very few coefficients, so probably
reflect chance fluctuations.

A more detailed discussion of the roles played by
individual environmental parameters follows in Chapter VIII.

1. QSEFULNESS OF NON-fSZIthIFICij CORRElflQNS.
Isolated non-significant correlation coefficients naturally
cannot be given much credence, since we have no assurance
that their magnitude is not due to random sampling error,
or to random fluctuation in nature which represents a sort
of "background noise" in the system. However, a whole
series of non-significant positive or negative correlations
with a conmon reference point may indicate a trend. To my
knowledge, no Significance test for the difference of
series of correlation values taken as a whole from random

eXpectation has yet been devised. It is, accordingly, im-

90
possible to state that this or that trend is "Significant”,
and the use of such tendencies in interpreting the data may
receive criticism for that reason. By the same kind of in-
ductive reasoning which first led men to seek objective
significance tests, however, it seems permissable to now
consider seriously those trends which, according to our
experience and judgement, stand a 3531 small chance of
being due to random fluctuation. An Objective measure of
this "very small" chance would, of course, increase one's
conviction. But in the meantime it seems pointless, indeed
irreSponsible, to ignore perfectly useful information.

As an example, bracken fern @eridium _acquilinum,
abbreviated .1353, no. 35 in the tables) on the Wet Grayling-
Croswell‘am) soil type, in the l m2 quadrats, exhibits
with other Species two positive correlations and one nega-
tive which are significant at the P< .05 level, and none
more highly significant. But if we count the non-signifi-
cant positive and negative correlations whose absolute
values are above 0.05, we Obtain only 2 positive and 11
negative coefficients, apparently corroborating the rather
Obvious field observation that other Species tend to be

absent or to grow poorly under or near lush bracken cover.

TABLE VII

Complete listings of correlation coefficients,

_r_ and i.

This is a right-hand matrix, 1. e. the lower left
half of the matrix (shaded portion in the— example below)
has been eliminated because each coefficient in the ith
row, jth column,r j’ is identical with the reciprocal one
in the jth row, ith column, r31. This is illustrated by

andr zin the example. The diagonals r11 are also
eIgminat:2 , since the correlations of variables with them-
selves are unity.

Variables

. ' “T’W‘ 9
. I ‘ l I 8y
- : :@l£26: I :
I E , 4 , . , I 1
I I

, ‘ + e I
1.....- - I - cs 1
. . I L . if ___i
'@I

I

p—I
N
w
b
U1
0\
\l

I
l
I

 

 

“7'1!
II;

 

 

i

-————-¢—-— — 0 , r 4 ‘ '

 

'I'
I

Hm m Ho‘s: H-H m<:
OQOONO‘MJ-‘le-i

I

The table is divided into pages as indicated in
the example by circled numbers and heavy lines. As many
of the lowest numbered rows as will fit are placed on the
first page, and the columns are placed on the first and as
many subsequent pages as necessary with this series of
rows. Then the next lowest numbered sequence of rows is
begun, and the colmnns within it treated in the same fash-
ion, etc., until the end is reached.

Within each block, gradient segments are listed
vertically from the tOp down in order of increasingly poor
drainage, and quadrat sizes are arranged horizontally with
the largest size at the left.

Nunber 0f quadrats upon which these coefficients
are based are for G, 120; for WC; 112, for AS, 38; for R,
29; and for DG,101.

Rows of d coefficients are Shaded in red.

 

 

 

The variables are ordered as explained on the
page preceding Table I, page 27.

Decimal points are omitted and negative coeffi-

cients are underlined.

Significance is indicated by

symbols after the numbers as follows:

symbol

P<
a ii

0.05 0.05
0.02 0.025
0.01 0.01
0.001 0.005

The legend for all abbreviations for Species and
variable names, soil types and quadrat sizes is found in
Appendix B, pages 375-376.

FOr example, on page 91, last figure in the tap

row of coefficients
on Segment C, 1/4 m2

P<.01.

‘we see that‘g for the pair Amo-Car
quadrats is +0.24, significant at

TABLE BEGINS ON THE NEXT PAGE

TABLE VII, cont.

FIN
00

SISI

N
O

SISI

M
O

M

“I
el

egfl

422°

\‘I’
H

M
N

91
Male! 3
4,83 S

"o”é’ll SI
,SIeI SI
00
P4,

j‘I
ISISI s
Ifld‘l \O

o O
I s
ssII a
E::.

In
‘JIN (5
11.3: ‘m"

~1‘
o 3
42}:
M

\1‘
O

N'cr
O o

N

8I8l

F4
‘I
'I
°°I
0‘ ,0
o 1
C I
H
M
O
\‘I‘ i,
o I
I,
I;
+:I--l
I40
U
B
Inmst
h
D
03
<mp-1
O

w.

20 21
ERI
1 4 H

EPI
14H

19

DIG
1 4

18
DAN
141i

-
x?“
on

SISIS

SISI

SISI

8|?

can
00)

SI
‘1'
cSISI
:3 8 SI

e33
«mo

[N
O

N
I-l

N
H

\D

Q

HI

12m
N

92

r;

n

0
_"ISI S!
j :

‘8
' SSH S
SIS! !S! S
at
a!
I
SI
.1 d'
: o
I? ‘0
j o
f S!

S!
fiSIS S
SSS! S!
o s o

‘2: 4ND
M ‘1'

8

SIS! SIS!

‘1' Nln'

o o o

8 SI SIS!

n

0

8| ‘8

n

o

3! 8

n nlol

O O O

“I“ SIS!

c ‘e’ e S
d m :s

n o

02”,, .-

O6

93: A

duo-1 mad

w.

22
GAU
14H

('1
O

Q
0

SISISI

S!S!S!

SIS 3

@MH
COO

S!

‘D
O

N
O

N
H

O
N

‘ M'hl
.. 0°
tench,
H'OO

9 SIS!
S S a!

01 04 ”
03 9;

89". f:

MWC

93

0604

\O
0

SI

0

n
O

07 06

37#51#
03

"‘I

00

07 O4

13 91
15 04

10
13 93

8|

8|

NI"'I
CO
H"
00'
o2

don-l
\D

06 ‘_"

RA- ”f '.

B WC

TABLE VII , cont.

41
nu
14H

38 39 40
SP1. SP? 0!.
H H 4

37
RUB
1 4

('1

SISISI

H'N
o n

HO‘
OH

°‘EHEH

OHO

SISIS

HMI-I
000

%I ISI
WI ISI

033

<20 <20)

H

ls

N

jg

HI

V}
H

M
O

|\
O

O
N

In
0

[N
O

S!

N

92m

.OS 22"_,.,
331E

:I8

[x
O

N
H

61 02

L09 ".
03 00

39:22“‘
03 03

17.21.".

N
53
53
u
U 3

94

31#07

M
O

in
o

06 05

O

.2

12 31#
00

05 13
g;

25 '.'30#

9;

pt

SRSI

3
do“:

HI
.-1

ts
OO

06 93
12

0

\od'
Ho
H'v-l
OO
NIH
OH

00

0‘3 '

W.

54

51
IRE SHR HRB N28 ISN TOT LIT

49 50 52 53
4 4 4 4 4

48

A

47
VYP
4

46
VOX
4 H

l

45
V10
4

43
V18
4

1

42
VAN
1 4 H

SIS SI

03%

\oofla
c>o¢u

HINV‘
CON

\0
O

O
O

SISI

SIS]

«3&3

<lx<3 4

H

r.
H

r-l
N

.‘L’I

SI

12 08

08 04

A8

IIm
ea

15

'20' ‘
01 pg

h~
O

.' Q
we 0

U
U 3

.4 2:: '¢¢£€) '4 m m d and

M

95

oo
_7.

O
SI S!
3! SI
01 O
O O
M e
H O
«1 ‘0'
H O
H m'
0 0
ml MIA
0 O O
we
MI H
o a
h N
o H
q H
o H
H \D In
C O O
H
O
0 Gig

.q m

03
9..

g
9.9.

03
21'

2
g;

10
.2

[\H
OH

33m:

30#17
01 03

3|

8|

0
0 3

'0

are. 01. A

" ‘04

96

TABLE VII cont.

 

 

 

 

 

 

on x S .oM ma. Po. 8 mm m. mlm
m S 3 oo oo «o no oo oo no
d 0
a | I I
a oo 3 8 no 8 so so No
on 4 on
u I
o 8 «o oo so no 2 no Mm
I m «4 o
I mm B on a o
o no oo o o
m
l a I I l I I
2 «o No :8 .3 2 8 3 oo oo
I no a
no 4 o I I I I .l I I
I on c on B m Mm no 3 co 3 .G
8 no no oo mm so 3 B No mm
| o oo no oo no oo o no no
a c a a o a a c a e e o e c e c c a
EU a EU HAG Hun N11 .71 ”3 § A3 NH 3
.3 OH a no No .3 on an on km. on an

2

42:: 4:60 duel: den-1 and
o

o3

<38

0
~¢

o 3

2

<IO 41:1!)
v-c

0:3

0V d G. d an 9% d z 9\ d Q d 1 Q d
63.x Unfi- Z<A~ 5:0 ”VIC 23° 2.70
:5 91 as- h u C - at: «\H

 

 

TABLE VII , cont.

21
ERI
14H

20
”I
1 4H

19

D10
1 4

02

01

    

 

 

 

I
8 8
'3' 3.: 2|
gm 3 a g
N N
O O
SISI 8
gig! 2
s 3 3
8
SISII
N no,
0 O
('1
O
3::
3 8 3 5 a a
SISL a com
3183' 2: 2

04.4 on”: 04:: cm:
a S :‘I

W.

31
ON
1 4

 

N
O

6')
O

SISI

‘Olnln

SIBI

ZHEBSH

O2

01 01

   
  
 

   

29

   

07 08
93 04

20 191:

 

 

   

 

19 14 04 15:06

"I
F!

_1;

lg

14 02 07

20 16

06

H

 

ITSVK

 

 

TABLE VII, cont.

04

03

g 3
04.:

09 40’

g

M

NH.
00

N

SIB ‘o°|

813!
In
SISI

SISIS

an
0°

SIS SI

of:
.4
In
.-I
I;
o
ChH
o
[x
O
.400
00
.-I
.-a
n
.4
o
.-4
«In
H0
.40
..-co
can
OP‘
NIO
CO
olh
HO
~1‘
o
\fo
OH
on
OO

05 05 08

1_o

24 15 31 19 13 01

_—

62#28 08

38"49?

10
00

13

25"

09

n
9
o
N
00
.-I
ma
Ho
In
0
o
00.4
o
I».
h
<f
00
r‘
-0
co
N
Incl
0.4
palm
H0
000
HO
<tlv-I
00
mp:
OH
.—c
O
.400
CO
on
OO
O
3
00::
N
.4

22 09 07

09

H

<7
0

In
v-l

\D
H

('1
H

\D
O

O
H

N
O

\D

II.

TABLE VII cont.

46
V01
1 4 H

45
V10
1 4

 

°°I 3|
It
:‘
7:: :il a
i
‘ °°| “I
O O
;

09

32..
0“04
a:

EIS 8|
co o E
O H m
sls 3|
«n a“ -n
cc: 0
3| 3|
«1pc w
9.4 o
" 31%| 2

il

‘1'
3| '5
°o°I8I 8
SI'SIBI 5'
8|
2|
8| 3|
::
3 ‘3
('1

 

42'70#
08 11 03

SISI

”Pl
0 v-l

Slfll

gal.

06 36 10

10

38# 16

11 07 36# 07

09 05 03

11 29912

10

_1_§

18 09 03

V0
0

In
C

12%29

11 06 12

101

TABLE VII cont.

e ox ¢ ¢ Q Q e
HAG a Nld H01 as 23 as m
no No .3 8 an on kn on

Q
HA

 

.o Vi.

 

 

 

 

 

 

 

 

IEI. . III..-

.I. I. . I .III.~.. a- n.
w-
I . I I, ful. I IN}. .u. I... l‘l..1-.'w1.ilul I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.oIHNH fino ammo. 3%.
MI ouooo H H

(I I.. ihf‘Al/N.‘

I . Ll .I. r? :1. II I LI“ h. __ I . . .
In I. L :I. III . .II ..I.. 1 I I...... I . 4-.. .. I . I I . ., I a. a. _ _
I II IIIIIIiiulllaJ-I‘jlqflfisg‘vlfs I . u..HbfiLlufrhllllll-$§R$IIPH ..I_I.. .

Iq uhflflunrr..IU‘-‘l.oilill. .uIlnnll.|I.I “Nun-Pg
. r . . mh
-

oo oo «Haou

\.u-ILII
L.» IIHHI‘I

.IF-I'III II—‘. . .4 I can}!
.. I. II ... ...-..lflhrI EIIII I...
I II I....thf «ful.... . . .
It. :2 ..Iu .. I I.. . , I I H I .I... .0 I
II 'l-IJIfi-II..I IanI 1.. I. I. . I I . I I...» II . .‘I .. IIIIIII .
III LIIIIII‘IIIIIVIii-‘thlfm .._IEm.m n .U- H Furn- .. It. . . . a I I I . I
H II .. I. I o. u.. h, M AI .u‘l ‘ . \l c I I
I... .. .2 . .. . I.
. I. J, .\.I|u.
'l

 

.— m e .n m ¢ .n m o» .— 1» .— 2
Q N m e .n n .v H
OH on H m
hNA ”MA 3 5 “3H 36 Ham Ham 0H9 awn
N n N #N nN NN H N ON

 

TABLE VII, cont.
N
O
('3
O
In
0
SI
.1
O

103

TABLE VII cont.

 

  

I

«~11

“snooze .

o».—
and
on

rkr 1C. .nJuI.

M¢H
E
nn

3 No.3 nanlom. «a 3 2.

.ivlwwg ”my

me."
D“
«M

 

 

 

 

.1

 

a. . . . . . .I-.. I I?! I. -I. -III. I.IIIIIIIIII.IIIIIII.IIIIII:I II.I..II:II . .I.II.quI..-I.II,.II,.II.IIuId I . . I I“ 03 H “H
WI . . . . ‘MMIMIH. . . - , I... ....1..I..IIII III.I..I.IIII§I . . . ...~.J,......r.1. HI ....-......I “EucaI . . . . III. I & a a

IBM Iwwoo olmlo. Ioddm oo mm oaz
.o .N m..H._ :I--II I.-.HIo .HHII In IIH Ho Ho .3 No no 42

III-i. Ir II II _.\. ‘l

.L . | I I I I B I I III Eli 5i. ‘3'. II a- fig
I. V.. 3,. IF .0 .u.. I-I.I_I . .. I ...~. .. .v I.. ... . . u a
n r.thII&m.III.II III .I. .a . I i. . . .II .flIIIIJflVIt . . 31.1.. . . I 1 . . . . w
i

104

TABLE VII, cont.

 

 

 

 

 

 

 

 

I .I . .. I , ,x. ,. «I.. ,u....I I . r... I.. I. . H. r. .II. . ...n~,..I..-..I. no IIIIW. .vh. .. ..I.I.I.4H. ... I..... . . . . o D
. . , ... .
I .a. I .bIIthvaluu-n; 5.. I, ‘ “ I‘VE-i! Elli-I..:i! I‘lll a

 

. IIII I I . IIIII . III . I I I III. III I I . . .
. I. . .. oI . . I . I . . . . I . . I . .. n .. . . A I
”J .. a
I I. II .I I . Ilia-.fi .. I I ..I I . . . I. I. r Ill-Inciflhlilfif ..I IIII‘I...“ ii 'ai .1411 I‘ ”“11 .‘IH: I.--4.93.11! III...\.I\..1\II!‘.III‘I‘I1§II i!.fl'dlfiai“iarg I.....mfi‘ “ I l.|.hI. \‘ I .

nH nn Nm .3 Noin mm .3 NH no

Imo oo III- -I . . No .olo No «P. 3 3 H NH
fiIIIIuIIIIIIII .IIIIIIII, . . . .- _ .

oo oo 2 oo no «H .392

oo oo «o IMHIIHNHI .oo oo 2.. no. 3...? n.N Ho 3 ..

 

 

It. I'iu

 

    
 

.I» II II III I II I III . .llln. II... :leI ll Ir, Ii! 4 l’ 1..., ml I
In. n-II‘F-HI .13.! “Ll-l. . I.. 1.0 . I I , .. . r |.I_ II- I.- .I I .
I i IEII I
{LII . I I . I I. I... I I.I . 5.. ... 1 In I... II iii-“1:4fi

 

.Ho .8 no «a g No 5.3 mo co Xenon

m Q .n m d .n .w .— m e .n m e A m c .— m .0 H d .— m m e H
E “R; OH> UH> flH> 24> HE. do» an"; Aha mam
he we né 3 no» No» .7» 3 an mm mm

 

105

TABLE VII cont.

 

 

flIIIII. I I III. IIIIIIIIII.II..L:IOII I r v I ruIF '- .’n.tLI!~IDIPI;JrE$.J¢.I’Iradrlnuflgl PRICEIEKITII IIQIrIrIPIt-VQIIDSYI {I.IIB... .I...I...IH I ...1D III . J . .. .M. I. . I..... II

I. I..IH.ELI1Iani—n
m I. II I I .. I. I. .IuI-IIIII I... HI I: U4 r. .. III..|- .nCIIIIl‘lIIIII'aIR‘. .. .. n I.. ...

 

”PI
0 O

 

 

IoIHINHmImMmmIHNIoHHINIoII InomIIooonIonld...

 

III... SI 3.. NH nHIIIIoHNIIINIon N2 IIIIHIIIIIIIIINon «IoIIINoIIImI
m I....MIII ”I..”..- III..- I I .IIII I.. I. I I
«HI.-- “I... I... I III! IIIIIIIIIIIIIIIIIIHINIIIHIHINI I..“...HIIIIIIIIIMII:IIII IIHIIII. Iv

 

 

 

. 1‘: I ‘thsgaj

 

 

“I! I . . . _ I . I .. . . 7.» .I. r..4 HII . IIIIIIIII I.. ”W IIH.I.I.I....I...I..sI.. . i . \IIAIIIQLIFIImfisIQnI “VIII-r... Kai... I‘IIakitulIIIIEE IE... nebgrirIBI‘.3rrIn¥l ..

 

 

 

I a I ...I.I|.. I I.. I I I! I. II. I .I III. I I I I .III. I..»; ‘ItlII’IiIHIIIIID III . III -iLBInIFuI!’ Alllxcg- a . VII!
I... I I I . .I .
I I 1 A v. 0.4!... I . . .IF... 4. .I.lu‘.‘ . I-

I: .I.‘ . Isa-i JIIHOIPAI O‘I‘IIQIb

no oofl

.Iv-ix'!‘
HUN

.2
.IH
III;

QIHINNoNMoI.mIoInIoNNflS

_HVI «N. MNIHINMIIHQMIEWMH I OH .00. MH.. ININ I . no no .mm: mo .Ho

a

..mo .mo. NN .HN HH g¢H No «N NH no we :éN Ho no

.8 IoIIIH NnInI. oIINI .oo oo Na .8 No NH oIH. no mo Na. a

a I |13I Fij- i :1 Pl]... .011 4):. ITEM. I.. I..“I'IImlj. I dank. .v\.IID|.vr1.III.IIII..PI IPII £11.: I..-I. . :.ll=.::ll1= .1

If. IT! .nlIl-I. Lil-1L: I:Il..lrIIIITi}ul.-lls.v.uiltiulll.l1iilll L Irl .u IFI'I .. . t-f I..: _I.‘ II

IIHo NH HH :3 .o_II.H. no no o .3 mm Hod :3 IN mm

m
e Q ¢ ¢ ¢ Q e ¢ ¢ ¢ ¢ w w e Q #

HAG ”an «.14 Ht< m3 204 A8 NHA a HHA Hon. 6H an! a 5.5 a

no No Ho 00 on an mm on an Qn an N“ Hm. on Ge 3

cont.

TABLE VII

m I 3.:
HH HH
low .H.Hu
H N

m1 , H -H -

HP. 2-

Hem-.938 HH..H.H,*HH 2n HH 3.338 H

m N
mm no 3 HH 8 8 no 3 .H HH
8 Ho .HHHHH o: H

mm H 138 Ho zoiaw

lain-I..: 1- H: «$.th “.L.-u...»....-.s.r.,...M.H.-.-H-i. xii--- u H

HH HH HHm HH HH HH 3 H HH 2 2 a
.,.HH...HH%H ; .. H- H .. . mH- HH 4 HH

In
‘- 0
I53."

0
§n:
. H
3
E

i

t

I

I

i

it

E

1

i

I

3

x
{‘1‘
E
U

-.o~..HHN.-.HH $533+é-.M.-HM-H-H -$1¢- ,--..-w.-HH HH.-H. Wm-,.wam HH .w o

H , . H mm mm 8. :nlo- .mm Halo 8 wlo no 3:13.:- H

- .H. E. B ..,.,.....--. - 2,-8.5 .. .L.: ..,;:.H.H. . .IH 8 H HH
:88 33 SH HHH HH

I NW «H S o 8 .3 HH HH H
«H HH.--- 2 HH. :H S 3 8 HH HH HH.

5 8 - 1.199%:- -\- . Lyl\lvaJ u .5 I. (v {-3 3.. H- H, 1 .(._.!xI-'.N

14(\.l ) 1,34...) y .1‘- )\ v(\1»ulyu¢- w» ML A - iJIrj‘u-‘w'lil-HJCKP TL - )-|.l..vu-\Jv--3 ‘47 fill-x \IO t ,

H
E . g8: , .u .._..HH.HH....,-H.-H....-...H...-.....<-- HH...M...H-_--3,3HH--.HH-H..H- .,.:H.-.-H-.H....,HH.-., EH. .H..-H..M..-.........H.-.-..,.._H u
3 3n HH 0

 

 

 

8 3 mm
.Hlo 8
I. - 1
HH 2 I NH 2 HH B B S no mm HH
328 3 3 3 3 HH HH 3 3 3 HH
.3 mm no

 

 

 

OH Ho 2 I. 8 Hlo HH 8 we H. 8 3 no 8 OH .Hb HH
1
8 HH mo 8 _._HH.HH._HH

.8 ..HH le

no on 3
. mic
m. 8 HH 2 2 HH 2
o i
c 32% 2.823 HH. 3. 8 8 no
I I
w 8 «lo 2 Ho Ho Hlo
EHHH HHHHHHHHH HHH HeHmeH HHH .HH
M Ha nan mun Aha man can E Dun won as 80
T HH 8 an 8 B on n..- 8 an Ha HH

8

HAN “4.: “<2
H: In
N N

8
d
d
2 MN
03
u
“(D
dun
030
u
H
SMAN
mm
m
3H
how
can
u

no no no HH mm

     

a
mm no Mm mw =mw HH
mm mm mm Ho Ho
oH o o «o .HH

ug‘am

N m nllo S 3 .HIH oo-Hnonnn floHon IIHJ flgoo
no no nH oH Hnn nnn Ho oH no nH no=HH.HH nnHHHH nH
oi
1L nH . «H nH HH on no oH.HH.oH m: m: nH Ho HH

  

.H Q HH .0 ow HH
HHA Hg 5: an a an
en nn Nn an on me

TABLE VII cont.

... «N

mm «Ho on.nn

"I
0
3|
:3

Mm.mm mm
: a...

ma
no mm
Ho H
Ho co
'H mm um

«o o o o o o H «o HH HH

HH oH HH HH nH nn nH nH mm

mm nH.nn HH.:d Mm mm
.nn.nn no n

E
3

no mm
nH HH
no no
nH=HH

wagon Hn

man
s." NN

 

:1

HAN “4.4 M43
2 :3

82H «8
man-a 049
.4 N
N N

:33 3 as :43
"NH

O

N

 

 

 

TABLE VIII cont.

48'.‘

   

 

 

”:3
00" 8
=0
33* a 8:!
53* § has
l-l
8%” slam
age "
3 SIQRI 3|
3“! as N N
3 cl 33° 0'
Nn-‘l
"‘8" 2! 38:: 3|
WI:
"‘5" 88:! SI
:25" c on"
0' OISSI 3'
mg 3 H 33H H 3
”NH Hu-‘IN M49 “42
8 £3 3?. if}

 

0
1
1

TABLE VII cont.

8.: 88 IH.mH| Hono Mano 3:
o
.9.
nlo. no 8 S nnnnonnnn HI.Hn. Hn H83 3

«H 3 oo no 2 no oo no HH HH oH no no

no no HniHn no Ho no 8 Ho no Ho oo
nonnooH .I a mlw nHano WWW Hon@.H.H. .HlooH
:83.» HH no anHHizn 2 H 2 B H no no

no No no mo #mniuu no .3 no no mo co

 

no .3 .nlo .nlo .Hd .Hlo Mm Ho oIo no
NH. 3 t: w. fl Hlo oo .3 HM Ho mm
mm mm m an HIn HI nnn IH oo

oH._HH ...nH.nH .oHan
|
1 Q W N No. nIH .nn noHHoH Sam-m
u Ho no no no 3 S 2 2 Ho no 8.2

no NH NH .HN co he OH ma OH HO

oo

8 no no

t. I I I I I I I I
m no no 2 HH Ho no oHnHion o HH H Ho no S
c i

nH no 2 .HH no in 2 8 no 8 H H no.8 2 H H H S H
I i
W nH HH Hon 3 3 HH. N IH. um oI lm .lo a HH
m o n n n n n n H n H H n H n H H n H n H H n H
M HHHH ROM. 23 an a gm 3 E NO> OH> UH> E Z‘D
T «n nn Hn Hn on nn 8 S n8 nn 3 8 H8

Hnmm

HooIHm

no No No

oo.o~

H o

I

ooflno

finm%mnwuo

TABLE VII cont. 112

 

 

 

:II o 8 3
Guam
at!
'5‘ SI ‘SI
SI 8 HI 3 8|
8 '3 HI '61 SI
2 I S 8
d 8“ 3I SI
H ‘3 531 Ea‘I ‘81
‘3 H ‘8 8
8 ‘8 31 8‘ 8
o E 3 o 3 3 H o
:HH HH: oHo

 

113

TABLE VII cont.

 

II

I351
sBI
88I
o 33 «8

8B. 38

. HH Ho mm I
Saw 3.33 ”fin no”

H H H H H H n H H n H
QH> a E 84> Ha A8 a and BM. 9
n. .3 n. an nn

 

TABLE VII cont.

 

 

 

 
  
  
  
  
  
  
  

 

8| 8!
' 3| 8‘
d SI
SI :1!
a1 :1!
3| 8I
a a
81 a ,
l 8!
8| 8|
SI 8!
°'~ 8
l 5
5|
SI 3
2|: fiSI‘é‘I
8‘!
3m2°~ :33
:9- I..-

TABLE VIII cont. 115

XII SH
:1 fill
a H SISI 8|
3 an , 5:131 38!
a: m a sale” as:
‘8I ‘8I an
3 OH >Hfl
a SI8I 3*"
SI 3 3 H
8 \8 >3”
3

   

TABLE VII 3 cont: .

 

 

 

 

$3: t 8
3E: w
53* am
33* ma
mg, 1 SB
R§* ; 8h!
R , f' 2%
*5: g 83
RE: ¥J 5151
*5= F %s
“‘91: 2.1
“5* sh
Ҥ* 32!
3E= mm
95* ea
33* am
55: 8mh a? :3
.. ,g as
Ezfi 2%“ :3“ inf g,3,“ 5:.

‘8 3 3 S g 9

 

TABLE VIIa cont.

 

w.
99* 2| 5'4
35* 81:13! 8&1:

 

    

3“ 5:29:31 8338 9.2 3':
3: 23mm! exam :3 a
9‘54" 881585: 859833 3 am :1
RE: 335an 8318132st 0 832%! s
*3» suama 535mm: 0 flmaa m
“a: smash 31335313: ma m
”‘2‘ We! 22318153 3 29's 51
3E: 3WRI§ MEM! 5 aisles! :2!
mg, 3r 8 , :1 m
“55* 313mg 8mm ‘8 9328B 3 a $.48 may
“"9" 888833 swam a g 3&8 “.3“
RE» 33mg 533::8 gr?
322mg "Q"
09:.“
s a _
88! 83:13!
A a slam sis-:12:
mg a 3 figs aaams
box PHm 9mm

13 S 3

 

i

TABLE VII, cont. 119
83* 2? ’8' a s a '
a]; 15215! _ 8183 $51 a8: “52*”
5“: '3 3> man
4 ' 33mg «[338 @339: \3
H "“
34* 331833 ogag: . s
«on

aBSmS 3

‘Ovlfl ,
35* 3 3 3| 1:
“5*. ”Hats #851 33:: 38159!
3" 3% aalsa mm sum
33* ”9851 dBflf-‘l 318% 3mm
9“? §§mls masts Salem £22325
3;... 553mm Selassie 28m: E3533:
“‘5 ma; 3:52:98: halal“??? 023mg

. «on
“5* 83%? Estates $33053 5:
In t-‘lHflo
“‘3' 3881381 gamma 8
fl‘fi
533mg 3
HHH

3
Still?
ssfss‘

SEER!

Eéa's ml

agamg

<0.

R

CHAPTER IV
THE QUESTION OF GROUPING

In studying anything as complex as a biotic
cmmmmity, any opportunity to group variables such as Spe-
cies into conveniently handled subcategories is most wel-
come. Such groupings must, of course, reflect real
prOperties of the variables and the community if they are
to have any value. A sizable body of literature concern-
ing grouping techniques has deve10ped, but frequently the
authors seem to have made one or more kinds of errors which
have partially or completely negated the value of their
efforts.

Such errors may be categorized as follows.

(1) There is a failure to realize that, more often than
not, species are not classifiable either spatially or
physioecologically into discreet units, but have various
and complex interrelationships with one another (e.g.
napkins, 1957. and Goodall, 1952, 1953). Only in rela-
tively rare situations such as the polygonal patterns
found in some tundra vegetations (cf. Wiggins 1951) are
the methods thus deve10ped liable to provide a close ap-

120

121
proach to reality, and in such cases the groupings are so
obvious that the pains required to elaborate them statisti-
cally man be wasted.

(2) Originators of a method or other users may try to
apply it to biota in which it has little use. Application
of Hapkins' or Goodall's techniques to understory vegeta-
tion in a mesic forest would be an example. Their pro-
cedures at least partially reflect the natures of the
communities composed of two or more heterogeneous elements
in which they were deve10ped, but could not owe with as-
semblages of species each having broad and non-coincident
contours of probability of occurrence over the whole
sample universe.

(3) One or two factors are assumed to be the dominant
pattern-producing forces. Species are arranged in accord-
ance with these factors without considering that other
factors may be of prime importance for particular species,
and that environment is multidimensional. The approaches
which view vegetation not as discreet grows but as con-
tinuous 1y varying between two or more extremes , though
frequently more in conformity with reality than growing,
tend to err in this respect. We have for example De Vries

(1953), De Vries 21; a1. (1954) and even the classic contin-

122
uum work of Curtis and McIntosh (1951) on a larger scale,
Bray and Curtis (1957) on a smaller one. Whittaker (1956)
demonstrates that, on a continuum axis based on only two
variables, a given species may have demonstrable associates
which nonetheless fall at widely separated intervals along
the axis.

(4) A technique may be used simply because it is empiri-
cally found to produce groupings , without taking adequate
pains to ascertain whether or not these growings re-
ally represent the natural situation. Most of the authors
already cited are guilty of this in some degree.

(5) The fact that a number of different patterning tend-
encies may occur simultaneously and may require several
different approaches for their elucidation, is not con-
sidered. McIntosh (1962) demonstrates empirically that
different methods such as those of napkins, of Goodall and of
Agnew (1957) yield quite different results, although there
was more similarity between the grow ings obtained than
could reasonably be due to chance. These techniques pro-
vide essentially two-dimensional views of the existing
multidimensional patterns, and in fact I earlier erred in
the same direction (Byer 1960).

(6) The imposition of relatively simple mathematical

123
models won complex vegetational structure may tend to
link species which are neither distributionally nor physio-
ecologically related (Hopkins 1957) . Conversely, they may
fail to reveal close associations between species (Goodall
1953, 1954). (This is discussed below).

Perhaps the temptation to allow calculations which
provide a neat-looking classification of the commmity to
substitute for sound judgement is too strong. We may un-
conciously attenpt to squeeze everything into the frame-
work of a method even when it should be obvious that it
does not fit, so that so-called "objective" techniques
deteriorate into a sort of ecological nunerology. Mean-
while, our original objective of understanding the comm-
nity goes begging for attainment. Statistics require
competent interpreters just as machines, to accomplish
useful work, require competent operators.

With reference to number 6 in the listing above,
growing procedures may fail because they link chains of
vicariously correlated species. Thus A is placed with 1?.
because A is positively correlated with B, B with C, C
with D and D with E when, in fact, A is circumeutrally or
even negatively correlated with E. Vicarious correlation

is, in fact, often lacking, and dozens of examples may be

124
found in the results of the present study (Table VII,
pages 91 ff.). 531:3; "h__a_i_r‘y _g_v_a_t_:g" @313, no. 5 in the
table) and W lineare @131, no. 29), for example,
are high positive correlates on G (3; = 40.37 and +0.51
on 1 m2 and 1/4 an2 scales, respectively, both with
P < .001). £153; is significantly positively correlated
with m serotina (1332, no. 34) and Solidago "lanceolate
.1933 petiole" @21, no. 40), whereas W coeffi-
cients with these two are circumneutral (very slightly
positive). Mel , on the other hand, has very high
positive values of P< .001 with m canadensis @,
no. 37) and Vaccinium mtifoliun var. giggly (21.3, no.
44), with both of which Lats; is only slightly positively
correlated. Only with 11.31; ad_____tm_c_a mg, no. 45) are both
Agter's and W coefficients significantly posi-
tive. An examination of napkins' data, moreover, reveals
that the species within his "basic units" are by no means
mutually intercorrelated, either.

To illustrate a situation where two species are
negatively correlated, yet vicariously linked by positive
5; or i coefficients, let us take a simplified hypothetical
(albeit unrealistic) example of a universe containing only

five Species, whose distributions are absolutely controlled

125
by the presence or absence of only two factors. As illus-
trated in Figure 6 (page 127), a combination of shade
(diagonal hatching) and a mycorrhizal soil fungus (stip-
ples) is necessary for the survival of Species "A". New,
species "B" requires the same mycorrhizal fungus but is
indifferent to the amount of shade, so that it co-occurs
with.A and, since it is absent from a large part of the
area where A.is also absent (clear, cross-hatched only),
is positively correlated with A. If species C is an obli-
gate heliOphyte but indifferent to the presence of the
fungus then C and B could be positively correlated by
virtue of their co-occurrence (stippled, non-hatched),
mutual absence from shaded fungusless microsites (hatched,
non-stippled). But each occurs in some microsites without
the other. Species D, which is intolerant of both shade
and the fungus is positively correlated with heliophyte C
(co-occurrence in clear patches) just as A.is correlated
with B. Obviously A and D would be highly negatively cor-
related, yet by virtue of a fortuitous unbroken series of
habitat overlaps these species of exactly apposite environ-
mental and spatial requirements would be included in the
same basic unit! To complete the absurdity, a species E,

which requires both shade and absence of the fungus

126
(hatched, non-stippled) will be excluded from the unit be-
cause it does not co-occur with A, B, C or D, although it
has requirements in comon with both A and D. In physio-
ecological terms the two groupings, one containing A, B,
C and D and the other E alone, are meaningless.

One other factor tends to produce imprObable
matchings in napkins' basic unit technique. Since roughly
one correlation in twenty significant at the P <:.05 level
can.be expected to be a random fluctuation, a few of the
links in a long chain of vicarious correlations may be
spurious. Especially where groups tend to be large to
begin with, this error might tend to increase their size
still further. But with small groups, with few species
which could exhibit "random" correlations with others out-
side the group, this might be less serious. Since the
chance of a random correlation decreases at the higher
significance levels (only 1 in 1000 coefficients at
P (.001), so also does the chance of this error.

Goodall (1953, 1954) perhaps errs in the Opposite
direction, in that his method divides the area into
Spatially mapable types, in each of which no significant
positive correlations are found and which are therefore,

presumably, homogeneous. The fact that this ignores inter-

FIGURE 6. Vicarious association of physioecologicallz
dissimilar_§pecies

Shaded areas are cross-hatched, areas
occupied by fungal mycelium stippled.
Letters represent plant individuals.
Explanation in text.

 

 

 

 

 

 

 

128

gradations even where they exist is not in its favor.
Neither is the somewhat arbitrary way in which the area
is divided and subdivided. Goodall always begins with the
most frequent species showing one or more significant pos-
itive correlations and divides the quadrats into one group
which contains it and another which does not. We could as
well begin with the least frequent positively correlated
species or with one showing negative correlations and sep-
arate out a different series of grows. Besides, as we
break the quadrats into smaller and smaller grows, higher
and higher correlation values are required for significance,
so that one of our final grows might really be nearly as
heterogeneous as the whole universe. In any event, since
this method involves several calculations and recalcula-
tions of correlations it is most time consuming.

I wished to compare the results of one growing
or ordering technique as applied to the present study area
with results obtainable without using such a procedure.
Accordingly, I placed the species on each of the five gra-
dient segments into Hopkinsian basic units. This technique
was chosen both for convenience Q._e_. the possibility of
using the technique without calculating additional statis-

tics) and because it does not seem quite so arbitrary as

129

Goodall's analysis. The results are presented in Figure 7
(page 137). The method used was essentially that of
Hopkins (1957) as summarized by'MbIntosh (1962), and in-
volves placing together those species which are positively
correlated, or which are linked by a chain of positive cor-
relations*with.other Species. There are three modifica-
tions: (1) except in the cases of three species for which
only is had been calculated, correlations used were based
won cover (3;) rather than upon presence (¢ or chi square),
(2) information obtained from different size quadrats was
combined, so that Species were considered significantly
positively or negatively correlated if they were so on any
of the three scales (in no case was a species significantly
positively correlated on one scale, negatively on another),
and (3) basic units were calculated at each of four levels
of significance (P < .001, P < .01, P< .025, P < .05) in-
stead of only at P < .001 as suggested by napkins.

Number l‘was largely a measure of convenience, for
53 'were available for a far greater number of Species
pairs than were @s, but I felt also that it might yield
a result closer to reality. .As explained in Chapter II,
two species may frequently occur in close proximity to one

another, and hence yield high positive correlations based

130
solely upon presence, even though each may produce its max-
imum.cover in microsites Spatially remote from the other's
Optimal spots. Thus, with the is, it might be even
easier to combine two groupings which had little in common
with one another than with theirs.

Melntosh (1962), in addition to considerable dif-
ference, did discover considerable similarity between the
results of various grouping and ordering techniques. It
is not surprising, then, that the picture of the community
which.emerges from a non-grouping approach (Chapter III),
or from an effort to group on the basis of macrodistribu-
tion rather than correlation (Chapter IX) is reflected in
the basic units obtained here.

On G, WC and AS, for instance, there is a single
large "core" unit at P <:.OOl including most of the spe-
cies (Figure 7 Arc, page 137), which.probably mirrors the
mosaic of lush and Sparse patches present on all three
segments, and also the broad bands of lusher and sparser
vegetation associated with drainage on these segments.

The Species remaining outside of these units, moreover,
are among those which one might logically expect not to be
included. Qanthonia Spicata @351) and Cladonia "cup" L62)

on G (Figure 6 A) remain outside of the group even on the

131

P< .05 level and form a grow of their own, joined by
OryzoEsis pungens (922') at the P < .01 level. All of these
species seem to occwy areas of hard-packed surface soil
where other vegetation is Sparse or absent. £3195 £93311-
35131.95 (3a!) patches are extensive and seem to leave no
room for other Species inside of them, except for m-
_c_h_i_¢_a£ oblongifolia (fl) seedlings, which seem to germinate
within them and join 933335 at the P< .01 level. Comptonia
perigrina @033), a species observed to have few discernable
microsite preferences, is not included in the large tmit on
G unless correlations significant at only the P< .05 level
are considered.

Several of the small, non-clonal Species (SS) in-
cluded with the large grow on G are "squeezed out" of it
on WC (Figure 6 B), perhaps because they have low tolerance
for excess moisture and/or for lush cover of other Species,
and so are scarce where the dominants produce maximum cover.
In addition Pteridim aguilinum (Egg) , the large fern which
seems to inhibit most other Species here on WC where it
reaches its highest cover, is included with the main "core"
grow only at the P < .05 level. On AS (Figure 6 C) both
relative xerOphytes and relative meSOphytes are included

in the main grow, which is in harmony with field observa-

132
tions and other evaluations of the correlations. All mem-
bers of the large "inner" unit calculated at P < .001 on R
(Figure 6 D) are species having their greatest cover on AS.
0n DG (Figure 6 E) there are four rather distinct groupings
which correSpond fairly closely with what seems to be true
on the basis of other criteria; (1) Chamaedaphne £91133—

lata (Chm) by itself and seemingly inhibitory to most other

 

Species, (2) golytrichg strictum @3) occurring in micro-
sites less heavily covered by anmaedalhge which also sw-
port lacoinium oxycoccus Q53) and £11333 gngustifolia
(K_an_), (3) Erigphorium spissum (Eli), growing with the
lushest Sphaggum plumosum ($21) in the interstices between
Chamaedgphne hummocks which also swport high cover of
Vaccinium aggustifolium var. 93.3323 (Vig) and, (4) _V_a_§-
ELM boreale Qlyp) growing with _L_e_<_i__w; groenlandicum LL33)
in patches where the cover of all other species is low.
Moreover, these four growings show an unusually high num-
ber (five) of significant negative correlations (arrows)
between them.

Negative correlations, as pointed out in Chapter
III, are relatively uncommon. Of the nine from data for
this gradient which are significant at the P < .05 level

or higher, seven are between units and only two within

133

units. The latter, moreover, are found only within the
"looser" units based upon P< .05 significance. Thus, al-
though vicarious correlation is unreliable, at least the
negatively correlated species tend not to be vicariously
positively correlated here.

Noteworthy is the tendency exhibited by Vaccinium
boreale (5112) to be associated with the large grow while

flacciniun wtifalimu var. nigrum (Vig) exhibits no

strong associations on either AS or R. 0n DG they are

 

members of different growings. Although these two blue-
berries may be phyletically quite close and primarily by

a ploidy difference (see Chapter IX, Section D), they would
seem to be quite different in the roles which they play in
the comunity.

Certain of the results of these growings, however,
are not so easily explicable on the basis of other avail-
able information. It is not startling that four of the
six Species on A8 (Figure 6 C) which are not correlated
with members of the "core" unit even at the P < .05 level
are those whose Optima are on the more poorly-drained
Saugatuck Sand within this segnent, most "core" members
reaching their maxima for this segnent on Au Gres sand.

The two remaining Saugatuck "peakers", moreover, are ex-

134
cluded from the "core" unit of Species based won P< .001
significance. But why, then, is not one of these Species
characteristic of the more poorly drained soil signifi-
cantly positively correlated with any of the others? Why,
on R, are three bog Species included with several wland
Species in the unit at P< .05, while the other two bog
Species are not included with the first three at this
level?

Such results may, it is true, reveal things about
the community which one might not have suspected otherwise,
or might have by-passed in the complex matrix of coeffi-
cients. For example, the Species which peak on Saugatuck
may possess in common an intolerance of lush vegetation
cover. It may be for this reason, rather than because of
relatively great tolerance to poor drainage, that they pro-
duce their greatest cover on sparsely-vegetated Saugatuck,
and this same intolerance of dense vegetation may even pre-
clude positive correlation between themselves. Similarly,
the apparent inconsistency of the growings on R may re-
flect previously unsuSpected qualities of species or en-
vironment, although at present it is difficult to say just
What. An examination of all the correlations calculated

with variables which could conceivably be influencing the

135
microdistributions on R reveals no consistent reSponse on
the part of the growed Species.

Other growing and ordering techniques might reveal
still other facets of coumunity structure, precisely in
their non-coincidence with the basic units and with each
other discovered by McIntosh (1962). But in the paucity
of our knowledge, we cannot know always which of these
revelations may be the more valid nor, indeed, what each
one really means.

I am convinced, however, that if used to the ex-
clusion of painstaking examination and re-examination of
the data, accompanied by intelligent interpretation and
the tireless formulation and destruction of postulates,
growing devices based won a single mathematical measure
of association tend to obscure more relationships than
they reveal. They oversimplify extremely complex situa-
tions, for it seems unlikely that growing tendencies
within a comunity conform to any single simple mathemat-
ical model. We might, thus wrongly conclude from the units
on G that the small solitaries (88) are physioecologically
similar to, or occupy the same Space as the more clonal
(GK) and meSOphytic (MD) and (MW). Representitives of all

types are included in the same units and (SS) and (CK) both

136

have representatives outside of the unit. On AS we might
conclude that xerOphytic g_a_1_'_e_x_ (Car) and bog margin W
@95) are very similar in their moisture requirements or
flooding tolerance, and so on.

I suggest the use of some fairly simple growing
or ordering technique when it is necessary to examine a
great many stands within a short period, so that Speed is
at a higher premium than accuracy. Otherwise, they may be
used as "scratch-work" to help visualize some order in a
community or as a starting point in analysis in conjmction
with other approaches. But, if only a Single overly-sim-
plified technique is relied won to completely and accu-
rately describe the existing patterns, it becomes merely

an excuse to neglect one's responsibility to interpret.

FIGURE 7. ngkins Basic Units 22 each Gradient Segment

Groups enclosed by the lines are the
basic units. Species correSponding to the
code symbols are listed in Appendix B, page
375. Code for the lines is as follows:

 

- united at the P<1001 level
----- - united at the P<L01 1¢V91
--.—-—-- united at the P<.025 level
“""“"""- united at the P<105 level.

A. Basic units on Segment G.

B. Basic units on Segment WC.

C. Basic units on Segment AS.

D. Basic units on Segment R.

E. Basic units on Segment DG.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

CHAPTER V
PARTIAL CORREIATIONS

In the preceeding two chapters we have been dealing
with the tendencies for pairs of species to occur in the
same or in different microsites ({ coefficients) and for
high cover of both members of such pairs or for generally
high values of any two variables, to be present in the same
or in different Spots L1; coefficients). In both cases we
measure such correlation just as it is found in nature,
thus obtaining the Egg-figs; correlation between the pairs.

Now, in his Search for the functional basis of the
patterns, a worker may often wonder what the correlation
between a pair x1 and x2 might be if a third variable, x3,
or a whole series of variables x3, X4, 2:5,... , xk were not
influencing their co-occurrence or coincidence of high
cover, great thickness, etc. in any way. Wright (1934),
through his path coefficients, devised an excellent and
precise way of measuring this, but unfortunately his tech-
niques are difficult to use if we are dealing with the in-
fluences of large numbers of variables. Partial correlation
analysis, on the other hand, is quite amenable to computer

138

139
programming, and provides a theoretical estimate of the
influence of a large number of variables won an associa-
tion. In the words of Crow 93; _a_l_. (1960): "The sample
partial correlation coefficient Qyi.345...k)°°- estimates
the correlation between y and x1 after the influences of
(x3, x4, x5,..., xk), as estimated from the sample have
been removed... The (partial correlation coefficient)
systematically eliminates the variation caused by (x3, x4,
x5,..., xk)".

This statement seems rather incautiously worded,
however, for in practice we cannot know to what extent we
are eliminating the influences of x3, x4, x5.... , xk them-
selves rather than that of other, perhaps unmeasured vari-
ables which concurrently influence the distribution of
X3, x4, x5,..., xk along with those of XI and x2. In re-
ality the partial correlation coefficient merely measures
the degree of co-occurrence of high and low values of x1
and x2 when the influence of x3, x4, x5,.... xk are removed
£193 1:113 correlation coefficient, but go; necessarily from
the correlated variables x1 and x2. Like other statistical
techniques, partial correlation analysis is not a push but-
ton device for cranking out answers; as always we must proc-

ess the results with other available information in order

140

to arive at "educated guesses" and eventually test these
postulates experimentally. Kendall's (1952) explanation
is therefore more accurate: "If a variate l is correlated
with a variate 2, may this not be due to the fact that both
are correlated with variate 3? The question may be decided
by considering the correlation of l and 2 in the suprpula-
tions for which variate 3 is constant".

Mathematical treatment of the partial correlations
is presented in Kendall (1952). The majority of Statistics
texts seem to ignore them, or give them only passing men-
tion.

In the present study, all possible combinations of
the variables x1 and x , with all other variables xk par-
tialled out one at a time, were calculated to give all pos-
sible coefficients 3” .k for each soil type and quadrat
size.7 Such coefficients are called 331.33; 235133; partial
correlations. Unless one can reasonably assume before
hand that a grow of variables all influence two other
variables in approximately the same way, which is not true

—._

7Although many such combinations 51 -k are meaning-
less or ridiculous, in terms of what we kn about the par-
ticular variables, a computer routine which calculated
matrices of all possibilities was easily devised.

141
fbr most cases in the present Study, such first order par-
tials yield the maximum amount of information in that they
pinpoint the exact variables influencing a correlation.
In partialling out more than one variable a time, moreover,
the combined influences of several variables on.a correla-
tion may average out so that they cancel one another. The
more variables we partial out at once, in fact, the smaller
the difference between zero-order and partial coefficients
is liable to be.

Table IX (page 154) presents, for each Species
treated all of its zero-order correlations which are over
0.2 in absolute value, and below these (indented) all the
first-order partials which differ from them by more than
0.1. Zero-order coefficients of 0.15 or more are also in-
cluded if and only if one or more of the partial correla-
tions of the pair are over 0.25. A detailed analysis and
discussion of every one of these coefficients would in it-
self fill a volume larger than this one. Here I am.dis-
cussing only a few'examples to illustrate the characteristics
and use of these coefficients. Others are brought in.where
apprOpriate in other chapters.

Let us mention, at the outset, some types of situ-

ations in which partial correlations are ridiculous or

142
meaningless. A most common one in this study, because some
variables are combinations of several others is that in
which a variable partialled out from a pair is a major
component of one of the pair members. For example, from
page 11 of Table VII,the fall litter thickness @531.) -
shrub cover (gig) correlation on DC, 1/4 1112 is +0.21
(P < .05). But when Chamaedaphne calyculata (cg-g, which
constitutes about 807. of the shrub synusium, is partialled
out the correlation is circumneutral (+0.07). If one mem-
ber of the correlation pair is a major constituent of the
one partialled out, we obtain similar results. Thus 29;-
Chamaed__aphne (page157 of Table IX) has _r_ - +0.21 ( < .05)
but with shrub cover (sh!) partialled out it is only +0.07.
A somewhat special case is that in which the variable par-
tialled out is the sum of both correlated variables, giving
a correlation of -l.00 as in the case on DC of laccingg
oycoccus (193;) and Egophortm Spissum (El-3L) with herb
cover (133:3) partialled out. (If, on
the other hand, the sum of the two is correlated with one
and the other is partialled out, we obtain +1.00) . In
cases where the partialled-out variable could not reason-
ably be expected to influence the other two, e.g. moss

cover partialled out of a tree- shrub correlation, we may

143
safely ignore them also.

As an illustration of a question which lends it-
self to analysis by partial correlation, we find that on
the hingeline between wland and bog, the R segment, sev-
eral Species and synusia are substantially negatively cor-
related with tree cover (Egg) , although the small sample
size on this segment precludes many significant coeffi-
cients there. Lichen cover (153) , on the other hand, is
positively correlated with tree cover here (Table VIII,
page 151 , tap figure in each subcolumn headed 3,302). We
would like to know whether it is the tree cover p_e_r_ ,s__e_
which brings about these correlations, _i._e_. whether tree
shade itself inhibits survival and growth of vascular Spe-
cies while favoring lichens, or whether some other thing
or things associated with the trees may be the controlling
variables. Now, trees themselves produce litter. In this
segment tree cover should also be highly correlated with
soil horizon thickness since the great bulk of the trees
are fowd on the portion of R bordering on AS, therefore
more similar to the latter in its soil profile deveIOpment
than to DC. We see from colmnn 1 of Table VIII that tree
cover is indeed moderately to highly positively correlated

with all litter parameters, most likely because the trees

144
produce most of the litter themselves. It is positively
correlated also with the A2 horizon (3:2) which is thicker
on the AS side of R, and negatively with humus (93;) which
is much thicker on the less continuously tree-covered sites
closer to the bog proper.

If now the thick litter and A2 horizons associated
with tree cover are responsible for the negative correla-
tion coefficients ofiggg with species and synusia, we
should find that these coefficients are less highly nega-
tive,‘i.g, shifted in a positive direction, when thickness
of these horizons is partialled out. As it turns out, the
mean correlation‘when these horizons are partialled out,
-0.284, is only slightly lower in value than the mean of
the corresponding zero-order coefficients, -0.3l7. (The
figures averaged are, in the first instance, all those in
the‘;t3,2 subcolumns of Column 3, Table IV exclusive of
the last subcolumn W139" and the tap and bottom rows
'23-'53" and "99h"; in the second instance, the tap or
"5332" row exclusive of "_lgg"). But individual cases are
much more striking. Note, for example, the correlation of
tree cover and moss cover @393), (next to last variable to
the right, Column 3) with litter cover (lip) and thickness

(lit) partialled out. Other good examples are tree cover-

145
shrub cover Q1335), tree cover-Vaccinium oxycoccus (193:)
and tree cover-Erigphorum gpissum (ggi), all with litter
cover (1.1.2) partialled out.

In the £Q3 subcolumns of Column 3, Table VIII are
presented the zero-order correlations of these Species and
synusia listed across the tap with the litter and horizon
variables of Column 2, which were partialled out in the
calculations of'gt3,2. Each figure is the coefficient of
the variables heading the row and column which cross at
that point. Again excluding the last column (15:3) and
row aoh , most of them.tend to be negative. Moreover, in
general, the more highly negative the zero-order correla-
tion of a Column 3 with a Column 2 variable, the more the
former's negative correlation with tree cover ("AV sub-
culumns) is reduced, 3.2. positively shifted, when the
Column 2 litter or horizon.parameter is partialled out.
New, we see from Column 1 that tree cover is positively
correlated*with all Column 2 parameters but AOH' and nega-
tively (tap row across) with all Column.3 variables but
1&2. But as just discussed, the Column 3 Species and
synusia are also negatively correlated with Column 2 var-
iables. Thus, part of the negative Column 3 tree correla-

tions are due to these negative correlations of Column 3

146
with Column 2 litter and horizon thicknesses, for the
latter are positive correlates of tree cover. A.negative
correlation‘with them will "coincidentally"’prediSpose
towards a negative one with tree cover which has its high
values where they do (cf. Crow _e_t _e_l, 1960).

To illustrate, Erigphorum Spissum (Egg) has a
zero-order correlation of -0.22 with tree cover, which
becomes -0.12 when litter cover is partialled out. So
about 0.10 (that is, (-0.22) - (-O.12)) of Eriophorum's
correlation with tree cover is due to this Species' coef-
ficient of -0.25 with litter cover, itself a positive cor-
relate of tree cover at +0.50. This does gg£_necessarily

mean, let us emphasize once more, that Eriophorum is Sparse

 

under heavy tree cover partly because this Species is in-
hibited by litter. Still another factor may coincidentally
preclude Erigphorum's growing with heavy litter. It gggg
suggest such litter inhibition as a possibility to be
tested.

A0“, the one soil horizon with.which tree cover is
significantly negatively correlated, is highly positively
correlated with two of the Column 3 bag species and with
moss cover M). We might guess, then, that part of the

negative correlations of Chamaeggphne, Vaccinium oxycoccus

 

147
and moss cover with tree cover is due to their affinity for
areas of relatively thick A0“, since tree cover here dem-
monstrates an "aversion" for microsites with thick. humus.
Indeed the partial correlation of tree cover with m-
51521133, 1. oxycoccus and moss cover, AOH partialled out is
substantially less negative than the correSponding zero-
orda' coefficients in all three cases. This is simply a
modification of the preceeding case, with the difference
that here Column 3 Species and synusia occur with high,
tree cover with low values of a third variable instead of
vice versa.

Lichen cover (1.33;), which is positively correlated
with tree cover, fits into the develOping picture. It is
also positively correlated with bare ground cover (133;),
litter cover and A2, just as is tree cover. Now when these
three variables are partialled out of the positive lichen-
tree correlation, the value of that correlation is sub-
stantially reduced, or shifted in a negative direction.

Not only, then, may a negative correlation be brought about
by correlations of gpposite Sign of each member of a pair
with a third variable, thus suggesting differing reactions
of each to that variable. A positive correlation of a pair

may result from their correlations of like sign with, or

148
Similar reactions to, such a third variable. In the first
case, partialling out the third variable will shift the
correlation in a positive, in the second case in a negative
direction.
Now, in general, part of the Column 3 tree cover

correlations, which we shall designate , are due to the

3:3
similar or different correlations of tree cover and Column

3 with Colman 2 variables r 2 and 3, reSpectively). If

£2
£t2 is substantially positive (ignoring its magnitude for
the moment), then the more positive is $23, the more the
correlation of the 001mm 3 variable with _t_r_e_, £t3’ is
likely to be due to its positive correlation with the trees'
close Spatial associate from Column 2. Consequently, the
more positive 323, the more in general will the partial
correlation, which we shall designate £t3-2’ differ in a
negative direction from the correSponding zero-order coef-
ficient 5:3.
should £t302 shift in a positive direction from 31:3. By

Conversely, the more negative 523, the more

an extension of the same line of reasoning we can also
Show that the opposite relations hold for the case where
£t2 is substantially negative. In other words:

(313.2 ' 5:3" 'f “23"

if 3:2 is positive and

149
($93.2 " 22:3)" +f (£23):
if 5:2 is negative, all other things being equal.

A number of things are not equal, however, one
being the magnitude of 5:2, and others being the degree to
which all variables included in the calculations are influ-
enced by still other variables. Nonetheless, the quantity
Q1334 - a3) should at least be positively correlated with
when r is negative. The value of this

+323 --:2
correlation coefficient for the present data, apprOpriately

'523 , or with

changing the sign of £23 where A011 is the Column 2 variable,
in fact comes out to be +0.93, for above the P< .001 level
of significance. This forcefully corroborates the suspected
influence of litter and soil parameters upon species and
synusia correlations with tree cover.

But is it not equally possible that tree cover in-
fluences correlations of Column 3 variables with the litter
and soil variables of Column 2? If so, to what extent? AS
a test of these questions, partial correlations of each
001mm 3 with each Colmn 2 variable, tree cover partialled
out were arranged in Column 3, @234). Substituting in

the above equations, we now have, assuming hypothetical

constancy of all other factors:

(323-: " £23" ”5 02,3).

150

and

G234: " £23" +f uca)’
for those cases where £t2 is positive or negative reSpec-
tively. The corresponding correlation coefficient betwaen
(523.: - £23) and 31:3 comes out to be +0.82, substantially
lower than in the reciprocal case but still with a proba-
bility well below 0.001.

Hence, correlations with litter and soil horizons
influence correlations with tree cover very markedly,
while correlations with tree cover influence those with
the litter and soil horizons substantially, but somewhat
less so. As discussed above, these influences on the cor-
relations suggest, but do not imply, a cause-effect rela-
tionship between the microdistribution of the two variables
correlated and the one partialled out. Since this relation-
ship exists between the partial correlation coefficient of
a variable pair and the correlation of each member of this
pair with the variable partialled out, all the information
contained in the partials is in fact "hidden" in the zero-
order correlations.

But the practical advantage of partial correlation
actually lies in this fact; we can eliminate the several

steps of looking at the correlation of x1 with x2, then

 

 

 

 

.m unevenness 5 mean: cam.

.00, )\Jd\JUlO

1..

 

 

 

 

O H ”.0... c e n. .. . . mm .11.. 3.11.. .
a m ounce . cm menus
oumnwmfiwfida of. dam hot/00 menu so?» Maugham U

 

152
looking for other variables with which one or both are
positivelyor negatively correlated. We went through these
steps aboVe by way of demonstrating the relationships, but
if the computer program is so arranged that all possible
first-order partial coefficients of a pair are arranged in
a colunn below the zero-order we can simply look down this
column and pick out those variables most influencing the
correlation.

In addition to revealing the factors influencing a
correlation, analysis by partials can bring to light pat-
terns and trends not evident from the zero-order coeffi-
cients. For example, on page 176 of Table IX we see that
[geridium gguilinun (Lt_r, no. 35 in the table) is positive-
ly, but not significantly, correlated with 933335 253211-
359313 (99;, no. 10) on AS, 1 m2 (5A8 - +0.24). But with
the shrub m sergtina (Egg, no 34) partialled out, this
correlation is highly significant (EAB°C - +0.43, P < .01).
Pteridiun, the bracken fern and gal-35, a sedge, are both
dominant and conSpicuous species "peaking" on Croswell
and Grayling, reSpectively. On the more poorly-drained
Au GreS portion of AS both tend to drop slightly in mean
cover but on still "wetter" Saugatuck both are nearly ab-

sent, as we might eaqaect of relative xerophytes. The afore-

153
mentioned partial correlation suggests that they do tend to
"stick together" on the better-drained soil and perhaps in
better-drained microsites. But an apparent Strong affinity
of 9933; for the ground under m (for this pair on AS,
1 m2 EBC - +0.62, P <.001, page 156) perhaps connected
with alterations of soil chemistry by this shrub's litter,
contrasts with an indifference or slight aversion to it on
Eteridium's part (_rAC - -0.16, -O.21 and -0.11 in 1 m2,
1/4 m2 and 1/100 m2 quadrats respectively). These differ-
ing reactions to a third Species may thus tend to keep
9353; and Pgridimn apart in nature, masking their similari-
ties in microsite preferences.

A somewhat different example is provided by the
pair _;R_ub_e_u_s_ hispidus‘var. obovalis (Rho, no. 36), a creeping
blackberry, and Oryzopsis asperifolia (Org, no. 32), a
broad-leaved caeSpitose grass, on A5, 1/4 m2 (Table IX,
page 173). Their zero-order correlation here is circum-
neutral GAB - -0.08), but with m partialled out this
becomes _r_

AB°C
for areas occwied by that shrub on the part of both Spe-

- -0.36 (P < .05). In this case an affinity

cies Q for Orzzgpsis-Prtmus and Rubus-Prunus +0.44 and
+0.47, reSpectively. both with P (.01) apparently masks

other differences in their microhabitat preferences.

TABLE IX

List of the more important first-order partial cor-
relation coefficients.

The Species abbreviation symbols are listed in
alphabetical order at the headings of the columns (capi-
talized, underlined, numbered as in Appendix B, e.g.
"l-AMO"). Below it are listed all those variables with
‘which it exhibits "substantial" zero-order or partial cor-
relation, whose abbreviations appear in capitals along the
left side of the columns(e.g. "CAR"). All of the pertinent
zero-order correlations of the variable heading the colmn
with this latter variable are listed by gradient segment,
in order of increasingly poorer drainage, and horizontally
by quadrat size, beginning with the largest size at the
left, and labeled 1, 4, H across the top of the column.
Below the zero-order correlations for each segment labeled
G, WC etc. at the left are listed all of those variables
(indented, lower case, e.g. "Epi", "Car") which, when par-
tialled out, change the correlation "substantially", along
with the cofreSponding first-order partial correlations.

Decimal points are omitted and negative coeffi-
cients are underlined. Significance levels are indicated
by symbols following the coefficients, as explained in the
legend for Table I, preceding page 27.

For example, near the tap of the left-hand column
on the first page (154), the zer -order correlation between
Amg_and‘ggg on Segment AS, 1/4 m , is +0.33, significant
at P<;05. With.§pi,partialled out, this correlation.be-
comes +0.19.

The criteria for "substantial" correlation and
inclusion in this table are given on page 141. The legend
for abbreviations of variable names, gradient segments.
and quadrat sizes is given in Appendix B, pages 375-376,
and within each category variables are listed in the same
order given there.

‘r‘k— .—

 

-) m._.-—' —_..._ ~. ._

 

TABLE IX. cont.

l-AMD
1

CAR G 17

WC ‘9;

AS 22
Epi

COP WC 00
AS 42?
Cor l9
Vam.52
a-2

COR WC 99
AS 38
Cop 09

Rho

Car 00

ILX WC ‘gg
Ase;
Cop‘lg
Cor 24.

LYC AS 19_
Cop‘gg
Corpgg

MEL G
WC
As—

Cor‘gl
Rho‘gl
hrb

H'O'O
0 O\ M

0R3 G
WC
“g

Car.yg

BIBLE”.

154

212-10 , ANS , ANM

1

PTR G ‘01
hrb

WC 13

AS 01
hrb

RHO WC 0].
AS 19
Cop 06
Cor‘Qi
Mel 31

TRI WC 99.
A512
Cop 39.
gli

VAM we 9;
AS .19.
Cop 2.6.

VIBAS 92
Cop‘gl

VIG G ‘91
WC ‘91

shr
AS 25
VYP G .Qg
WC O7

shr
Ase.

Epi

HRB G
Car
WC
AS
Car
Epi
Ptr
Rho

CARAS_2_Z_3_

COM AS 12.
Gau‘gg
a-l

COP AS .16
a-2

GAU AS 23
Com 33
Vyp 12
a-l
a-2

LED AS ‘95
aoh

MAI AS

MEL AS 26

AOH AS
a-2

IO r-I
00 «P

|H

N

14

IO
\1

lo
\l

_r_;

N O
u NNU
U!

12

A-l AS 27
A-2 AS 40'
Cop 54

3-ANM

1

ARC G 11
32%
05
WC 1;
Rho 07

MAI G

34%
37#
Pru l7
Shr
wc 38#

ORS G

21"
39#
Ors 22
WC ‘0;

PRU G

08
Ors 94
we 33#

SOL G

20'

15
Shr
WC 01

VIC G

15
20'
Shr
WC 27?

VYP G

SHR G
VYP
wc

VYP
lip

25?

c>c>
Ioolw

O
ON
LOCI)

O
”H

22H

22"

155

TABLE IX. cont.
Am, ARC, ASH, ASL, CPD, CAR

 

3-ANM(Cont.) 6-ASL l 4
4 1 4 ILX we 09 475%
Epi 37
M05 G 02 ASH G 03 9.1 Tri 37
we 401'; we 27‘.‘ AS 12
lip 13
CPD G 9}; 9_3_ MAI G 05 04
LIP G 02 v10 _2_; we '60" 07
we 5.2%.! we 9_7_ 93 Epi— :5}.
mos 22 11x ‘22
VIC G 30.? 17 m _1__
4-ARC we 91 03 AS _1_g
1 4 v10 G 33;? 9_5_ 0115 G 9f; 9;
we 23" __§_
ANM G 11 25': mos G 03 AS 05
we 21' ""
ORS G 18" 9g TRI we 09 40;}
Prn 28 9-CPD Epi 30
11x 26
mm c 9_5_ g; 1 4 H as _1;
Ors‘gg Com £1
ASL G 04 03
SOL G 31,-; 07 V10 3; VIC G _1_1_ 9;},
we 9.2. 9.9 we 9.5. 9.1
S-ASH AS ,1;
COM G 9}; 9;}, Q_2_ Com _3_(_)_
1 4 Viollg
wc g_§ 93 v10 G 45;? 9;
ASL G 03 Q; 11x 13
W0 27? AS 41” HRB G 08
Car 20
MEL G 37;? 51# cop we 04 g; wc 10
we 03 Tri- _23
AS 13 BEE G 10
0118 G ___1_ 91: we 31‘.’
Pm _2_2_ EPI G _;3_
wc _q__ we 12 409% lO-CAR
11x 27
PRN G 25': 30;? AS 07 1 4
we _0_;
GAU G 9_5_ 93, 95+, A110 G 17 241%
SOL c 12 319E we 00 20' WC 9.5. 9.7.
we 00 As gs AS 22 33'
Com 26 Epi 19
v10 G 12 417'; '-

AMSASg;

156
TABLE IX . cont.

CAR
lO-CAR(Cont.) l 4 H 1 4
1 4 H 11.x we 99 99 PTR G 92 _12
AS _1__8_ _19 hrb 88
COM G 12 20' 21' Epi _32 we '-
WC 10 15 16 bar 91; 15 09
shr 25 aoh 9_4_ hrb—- '71
AS 13 99 99 AS 24 1'5"
a-l 04 LYC AS 11 Epi 14
Mel _29; Ors 25
002 we 92 92 91 Pru 29 Pru 43
hrb 92 hrb 92
As 9.2 1. 11 MAI c 9.1 9.9.
Amo _l_._5_ hrb 91 R110 140 9_9_ _1_5_
Pru 22 we 99' L9) 1}, Com
a-l 21 AS 05 _19 99 Mai 9_5_
a-2 .94. Epi 91 29 hrb 9_
Pru _2 AS 19 13
COR we 09 92 A1110 02
hrb _3_9_ MEL G 91 99 Epi 03
AS 02 9_9_ WC 9__ 04 99 Pru 99
Amo 99 hrb 91 hrb 29
Cop 01 AS 15 15 959
Pru _12 1.9 Epi 02 TRI we 1.; _1__
hrb _3_7_ Pru 03 Mai 9]_._
a-l 29 Rho 05 hrb 99
gli _l_4_+. hrb 99 AS 05 05
Epi 99
1:91 G 00 023 G 99 91; 99,
we _1_1_ _12 _19 we 9_9_ _19 VIG G
AS 48'.‘ 553‘ hrb 01 02 99
Ors 67 AS 38" 49'.‘ hrb 29
hrb 34 Epi 63 WC
bee 65 Pru 94; 22' _1_7_
AS
GAU G 92 91 9_l_ PRU G lg 92 21 20
hrb 29 WC 07 00 hrb 10
WC )9 AS 62# 28 99
_12' _11 _;_3_ Cor 33 VYP G
hrb 29- Epi 40 29.! 22.1.!
AS 99 _1__ 99 Ors 08 WC
E91 32, _29 A5. 1.5.
Pru 99 shr 942
Vyp _2__l_1_ AS 11 08
hrb 2.: Epi 99 }_6_
a-l 08 Gau 21 21
Ptr 9;

O OHHIN
u memo

TABLE IXI cont.

lO-CAR(Cont.)

4

19'
lit 32
WC 12

lit 30
AS 05

TRE G

78#
Ptr 96
we 61#

Gau 72
Ptr 83
AS 56#
Epi 36

HRB G

99_
lip 23
WC ‘2;

119.32
A899.
hrb 99.
lip‘99
aoh‘99

MOS G

25?
1.1739
hrb‘ig
“C 2.9."

tre 99

hrb 9§_
As 99.
Epi 10
hrb 10
mos.99
lip‘lg

LIT G

LIP G

4

93.
lit 09
WC (9;

lit 07
A319.
a-l‘29

AOL G

06
hrb 99.
WC 07

hrb 19
lit 20
As 9.9.
Epi 06
hrb 11
lip 02

ADM G

AOH G .99
lit 10
WC ‘9;
AS ‘92'
hrb
mos

lip

A-l G 14
WC

BEE G ‘99
WC 08
hrb 91.
AS 29
Epi 49
a-l 39

157

CAGR, Ckfii
ll-CHM
l 4
CAL R '92
ERI R 50? 51?
shr 39
99.1119:
shr ‘99
GAU R 29_ l9_
19 14
Eri 30 26
Van ‘9;
Vox 36
shr ‘9;
hrb 04
ILX R 15 ‘99
Eri 26
Can 99.
KAL R 20 16
Eri 09
Vox 91’ ‘9;
991191
LED R 17
11.11
Vib.29
shr .29
lip 32.
a-2 ‘92
DC 92? 29#
shr ‘99
SPL R
DG
VAM R 06 36
Eri 19
Vox 68
shr 16
V18 R ‘21 .91

04

07

01

13
.2-2."

10

42'
Eri 31
Vam
DG 08
hrb

VOX R

99.
Vam
Vox
shr
tot
aoh
DC ‘29?
Led‘99
shr

VYP R

SHR R
Eri
Led
Vam
Vig
Vox
DG

HRB R
Gau
Vam
Vox
shr

Eri
Vox

.I.‘

O
H

I...

Oacchuar-RJ
wJ-‘Lnn-‘w

a.
g

N
01c4
om

Lo.)
BJFJUI
U10)

H
N

c~
has: unu>c>
£>-: rd0\r‘

ISIS :1:

03

W-

158

CHM, COM

ll-CHM(Cont.) 4
4 AM R 22
Vox 36
M03 R 11 tre 34
Eri 9}, tot 24
Vox 99 ""
tre _12 12-001-1
11c 99 """"
bar 23.3. 1 4 H
119 2.1.
80h 01 AMS AS 19 8
DG .22? Gan—30 '__
tre 91_ "-
8hr 1‘1 cm 0 99 9; 92
11? 5L6. we 99 92_
AS 41"
TOT R 41'
CAR G 12 20' 21'
BAR R 2.1 we 10 15 16
8hr 2‘1 shr 25
99 9!: AS 13 92 99
HP R 3.2. cop we 00 07 04
Eri 12 Rho — 70
Vam 32 AS 19 .-
Vox 08 2:; L9. 9.5!.
tre 05 Cor 03 04
shr l9. Vam 33 _-
DG 26’.‘ "’
tre 33 COR we 03 03
shr 10 AS T7 _-
m°8 £9. 25' 1
c _13 '—
AOL R 1.1. Rh: ‘3"
DG 21' '—
3'“ 07 DAN G 15 08 92_
AOH R 21 WC 4% _09
V0}! 1.1 EPI G 04
tre 0“ WC El 19'
8'2 31 20' 18 00
DC 3224' A3 06
521' 13
A-l R 27 '—
Vox 01

H

02
29
we 19'

GAU G

Rho
Vam
shr

AS 35 '
a-l

ILX WC

MAI G

WC

30 '.'
06

14
9_1_
25':
O6
01

Gau
mu
Tri

MEL G

W C

ORS G

12
15
25'.’

10
13
17

05
O4
38"
26

19'
06

00
11
03
11

08
10
05

08
32
10

45

TABLE IX.

12-COM(Cont .)

l

41#
28?
‘99
07
Cor 29
Mel l9

RHO WC

AS

TRI WC 29#
20'
Rho 09
AS 22
de 32
Mai

VAM WC 2813*
ll
Rho 99_
shr
As 91
ll
Cop 26
Vig 21

VIC G 05

WC

cont .

4

26#
18

26?
29?

19

06

03
10
05

9_5_

O3

12

353

07

22

06

07
23

15

47#

N
O\

b.)
I.»

12

12

36#
23
03

SHR

HRB

MOS

LIT

AOL

AOH

BEE

159

COM,

G 33%
Vyp 50

WC 38#
Vyp 51

AS 10

G 10
WC
_19

Ptr 03
119 3.9.

'P‘F‘CDCD
0000\1

00
00

N
00

mm
IO
t-‘D‘TJ

N
elol

COP

13-COP
l 4
AMD WC 00 ‘9;
AS 42? 07
Cor 19
Vam 52
CPD we 92 92
Tri ‘22
As 1.2..
CAR.WC 99_ .92
hrb '92
As 91 11
a-l 22_
a-2 ‘92
some 99 91
Rho
As 1.9.
2.9. 1.6.
Cor 99, ‘99
Vami99
COR NC 80# 74#
AS 99S
81# 69#
DAN wc 9_9_ 99
Rho 2;
EP1 WC 27? ll
Cor.99
Rho 17
Tri 14
AS 25
03 99
GAUWC 01 9;
Cor 99. ‘99
Epi‘l9
Tri‘ig
As 19 .1
a-l ‘92
a-2 10

05

05

TABLE IX, cont.

13-COP(Cont.)
1 4
ILX WC ‘93} ‘Q2
AS 3 15
Cor ‘92
LED AS ‘12 ‘12
aoh ‘29
LYC AS 24
Cor‘Qg
Rho 14
MEL WC 16 309
As 12.
17 18
Cor 2Q 03
Lyc O9
Rho-92 06
a-Z 31
ORS WC _0_5_ 94
As 1.. 12».
Pru‘gg
PRU'WC .Q§_ 19;
AS 10
16 16
Cor‘gg
Ors 31
a-2 3O
PIR NC 22' 24"
hrb 13
AS ‘gg'
12 .92.
Vam‘gg
RHO WC 42# 32#
Cor 26 18
Dan 25
AS 32
34' 22
Cor1<1 11
M21
Pru 17

65#

20

22

04

25

29?

15

02
08

160

COP, COR

1

TRI WC 30?
Cor
Epi 19
Rho 14
AS 29

Vam 14

VAM WC 0 6
AS 40
479
Cor 61

VIB AS 23
Cor 34

Tri 18
Vam O9

LCN WC
a-2
LIP WC

mos
a-2

43#
33

01

37'

33 tel: '5'?

25

04

08

O
(I)

ISLE

4
AOM WC .12'
AS 06
Cor 19
Vam‘QZ_
a-z‘gg

AOH WC 24'
Cor 13

AS 09

Led 25
a-Z‘lg

A-1 WC 11
mag
CorIlg

A-2 WC 14
AS 53#
Ams 63

GLI AS 06
Cor 32
Vam‘Qg
a-l 17
a-Z‘ig

14-COR
1

AND WC 92,
AS 38"
Cop 09

Rho

CAR WC ‘gg
hrb

AS 02
Cop
Pru
hrb
a-l

 

TABLE IX. cont.

14-COR(Cont.)

H

COM WC
AS

I l
0“!“
I‘D 'w

Cop'lg
Rho‘gg

80#
90#
81#

COP‘WC
AS

44#
25
15
11x 01
Rho 01
R 18
Vam 30
Vyp 06
1cn

EPI WC
AS

GAU WC 23"
Cop 37
Epi 04

A511
Cop 02
VYP
a-l

R19.
Epi
Vyp‘gl
hrb

11.x wc __
AS 469
R 9.9.

05

74#
69#
19'

03

67#

161

COR
1 4
LYC AS 34'
Mel 21
Rho 17
MAI WC 22' 30?
Epi 09
Rho 03
AS 09 27
MEL WC 10 22"
GOP 9.9.
As 11
35' 22
Lyc 22
Pru O9
Rho 01 93
NEMR 12 29
Epi 50
Vyp 23
1cn '94
ORS we 93 9};
As 12.. 12.
Pru; 19.
PRU we 9; 9;
AS 10
21 31
Ors 37 41
Rho 14
PTR WC 17 14
00139.}.
Epi 29
As 21' 15.
lg 12
Cop 08
11x 04
Mai 22
Mel‘gg
Rho 21

1

RHO WC 341'}L
Cop 01
Mai
AS 29
59#
Pru
hrb
R 13

TRI WC 22'
Cop 94_
Epi‘gg
Rho 07

AS
12

VAM WC 14
AS 49#
”15
Cop‘flg
R 11
Epi‘gg
Vypéé
hrb

VIB AS 04
CopZé.

Rho 14

R 9.1

VIG WC Q§_
As 19.
Amo‘gg
Cop 05
11x 05
Vib 08
R 57?

VYP WC 04
AS 00
Can
R 35
Epi
Nem
hrb
lcn

28?
O7
18

29

43?
33
31

29?

95+.

u>c> P‘P‘
(I‘M \ON

02

08

08
23

15

g;

26
07
25

4>

SI—IR WC 01
R 11

14"COR (canto)

4

AOH WC 21'
Cop 05

AS 13
Led 31

Rho 23

hrb 3O

11p 19

R 19.
Epi gg_

A91 WC 02

A519.
Cop‘Qg

R 06

A-2 WC 09
AS 18
Cop 39.
Rho 28
hrb 29
R 05
tot 21

BEE WC 01
AS 23

GLI AS .gg
Cop‘gg
Rho‘Ll
18-DAN

1

COM G 15
wc 40#

COP WC 9_4
Rho

CUP G 27#

TABLE IX, cont.

162

COR, DAN, EPI

OPU C 0

ORS G
WC

RHO WC

TRI WC QQ
Rho

MOS G
11p
WC

LCN G
bar
lip

WC

BAR G
WC

LIP G
mos
1cn
bar

WC

30# AOL 6

bar
WC

cs

AOH c ‘1;
lcn‘gl

lip 29.
mg},

A-2 G 10
WC 19.
Rho‘gg

ZO-EPI

H

04
33'
18

40#
27

;g_
55#
67
34
65

19'
18

11

00

05

W-

20-EPI(Cont.)
1 4
COR WC 44# 19'
AS 25
15 03
11x 01
Rho 01
R 18 67#
Van 30
Vyp 06
lcn 78
DAN G ‘9;
RC9. 99.
Rho .99
ERI R .99 “_9
mg; .21
GAU G 17
we 23" 25?
45# 25?
Tri 32
AS 21 19
Car 31 31
Vyp 02 ‘21
a-l 33
R 32 37'
Cor 55
11x 55
Vam 13
Vyp 21
hrb .1;
ILX WC 17 39#
de 25
AS 31 03
Car 45
R 08 ‘9;
Geu‘gg
Led 10
Vam‘gg
hrb 29

06

163

5P1

H

0

LED AS 1
R

[23!

Vam
aom

lb-

LYC AS .9;
Rho‘gl

MAI G 58#
WC 405')e

31#

Gan 17

Rho 20

Tri 10

MEL G 04
we 29#
36#
A391
14
Rho 01
VYP
hrb

NEM.R .ll
Cor

ORS C Q;_
WC 03
As 04

Car

PRU G ‘93
WC 16
03
AS 10
18
Carliz
Mel
Rho
hrb

4
99
99

1_.

329
23"

O
\l

b
0‘

H O
NNNUH w
‘1me

H

11

O6

1

PTR G 13
WC 22?
22!
Cor'gl
Gau‘gi
hrb
As1.2.
25
hrb

RHO WC 53%
27?
Mai 12
Tri 05
AS 29
23
Amo
Mel
hrb
R 9.1

TRI WC 48#
52#
de
Gan 42
le
AS 19
Vam 3O
Vyp O4

VAM WC 45#
12
As 17
gg_
9:1.99
R 69#
Cor
Gau
shr

VIB AS ‘99
R 09
Vam 21

53#
30?

19
09
06
13

51#
35#
22

23
03

36#
06

17

19
37

52'

HO
w“)

18

 

TABLE IX, cont.
20-EPI(Cont.)
1 4
VYP G 00
WC 11
21' 13
AS 37' 36'
R 38' 28
Can 17
Vam O9
hrb 10
SHR WC 12
As 91
R 9.9.
Cor 08
Can ‘Ll
Led O8
Vam ‘lé
Vyp .12.
hrb ‘99
bar Q§
HRB WC 06
Car 17
Ptr 23
AS 57#
Car 38
R 5 9
Can 40
LCN wc 1.2.
.99
As 9.7..
R 99.
Cor ‘24
Vyp .0_3.
AOM we 99
As 29.
hrb ‘99
R .2_?.
Cor ‘16

164

EPI, 5R1

4

H AOH WC 05
A392.

R 08
Cor 19

17 Vyp 21

A-l WC 07
A312.
Car 05
Gau‘gg
hrb‘gl

R 18
hrb 05

A-2 WC 26?

GL1 AS 91_
Amo 09
Car 10
hrb 22
aom O9

Zl-ERI
1 4

50? 51?
shr 39
9912. .2.9.'
shr ‘9;

CHM R

EPIR _1_9 lg

08

1
GAU R [l2
Chm‘gg
shr
hrb

ILX R .99
Chm‘gg
Can 01
shr

KAL R 25
Vox 99.

DC 12

05

LED R 02
Spl
shr
mos
lip

DG 13
00

SPL R
DG

VAM R 13
Chm
Epi 34
Can
Vyp 27
shr
hrb

VIG R 92_
Spl
shr

DG 11
21'

VOX R 35
Chm 18
Kal 25
Vam
DG
02

H
(ANO‘
\l

O N
\ON

N
H

02

09
99

02

OH
UlmO—‘t-‘N
UILJ‘IO‘

42"
3o

65

27
52

15

11

18
28

28

02

27
17

25
13

TABLE IX, cont.

21-ERI(Cont.)
1 4
WRR 11 .18.
Vam.§l ‘gg
shr ‘§§
Re 9.1 99.
TRE R ‘99
Chm SEQ
shr ‘lg
EDS ‘98
lip .11
DC ‘29
‘99
SHRR 40'
Chm 21
Vam 24
Vyp 51
DC ‘gg'
Chm 'yg
HRB R 511
Can 25
Vam .29
shr ‘il
DG 43#
Vox 10
MOS R 27
Vam 37
tre 17
11p 10
DC 13
TOT R 05
Chm ‘92.
LIP R 25
Vam '_37
shr '15
ms 5;
DG 15
shr .IC
hrb 57.
11:: E

165

 

ERI, GAU
4
AOH R 06
DC ‘12
hrb 23
A-l R {Q_
Chm‘gg
A-2 R '91
Chm‘g_
22-GAU
1 4
AMS AS 23 14
Com 33
Vyp 42 33
a-l ‘QQ
a-2 25
CAL G 20' 18
99911.6.
AS
R99
CPDGQ_5__Q_4_
WC 00 20'
As99
Com.g_
CAR G ‘99 ‘91
hrb ‘99
WC
1911
hrb 29.
AS1919
Epi._l .éé
Pru‘QQ
Vyp .29
hrb ‘99
a-1 08

35#

H
O

'O H
D Lo

'0 HH o
U: (DO H

1

CHM R '93
19
Eri 30
Vam
Vox
shr
hrb
COM C Q;
;Q_
19'
17
Rho
Vam
shr
AS 35'
a-l

WC

COP WC 01
Corigg
Epi.1§
M81
Tri‘lg
hrb

A518.
a-l
a-2

COR WC 23"
Cop 37
Epi O4

As11
Cop 02
VYP
a-1

R19
Epi
Vyp‘gl
hrb

_99

14
26
11
36
01
04

O4

TABLE IX. cont.

22-GAU(Cont.)
1 4
EPI G 17
WC 23" 25?
45# 25?
Tri 32
AS 21 19
Car 31 31
Vyp 02 ‘91
a-l 33
R 32 37'
Cor 55
11x 55
Van 13
Vyp 21
hrb .99
ERI R ‘l_ .99
3331.11
Vam ‘21
shr .99
hrb ‘99
ILX WC 12 15
AS 13 11
V13 23
Vyp 01
R 89# 46"
hrb 20
LED AS .9; .99
a-l .99
R 99
1999
11x .51
Vam 09
shr 19

H

06

166

24?
43#
WC 27#
38#
Epi 25
Tri 25
Vam
AS 11
Pru
VYP
tre

MAI G

21"
04
WC 13
A311
Rho
VYP
hrb

MEL G

ORS G 07
‘Qg
WC 05
shr
A3 11

PRU G 10
06

WC 18
22"

AS 20

Mai

29#
22”
WC 13
22'
Epi 35
hrb
AS 04
hrb

PTR G

21"
309

29?

H 1

RHO WC 32%
20'
Epi O9
mai.99
Tri 04
AS 06
M31
Mel
R 05
le‘gg

22H
52#
67

TRI WC 32#
38#
Epi 19
Mai 25
As 15
Mai
Vyp.1_

O
U!

H
BJCJ
\a

VAM WC 24?
13
03 As 00
R 34
Epi 17
Eri
Vyp 20
hrb

VIB AS .11
R 6
05 11x 24

VIC G 249

WC

35#
23"

12

30%
10

O4

31#

p—a
:vtvho
\Im>

TABLE IX. cont.

22-GAU(Cont.)
1 4
VYP G 23" 16
20' 21'
WC 27#
16 08
shr 29.
AS 53# 65#
R 38' 36
11x 48
Vam 26
tot 48
hrb 06
TRE G 09
WC 02
A3 99.
R 19.
Vam 07
hrb 15
mos ‘99
SHR G 10
WC 33%
Vam 22
Vyp 43
AS 16
Vyp 1.9
aom 30
R 31
Eri 41
Vam 99.
Vyp 20
hrb 03
HRB G 12
Car 28
WC 33#
Car 56
AS 14
Car 26
Epi 04
Ptr 25
aom O4
a-l 29
R 85#

16

99

10

O9
62#

167

GAU, ILX
4 4
mos G 17 LIP G 9;
13 WC 08
WC .1; AS 11
l9_ lit 25
AS 99 aom 00
Led 05 a-1 25
lit 99_ R 06
lip 13 shr 17
a-1.99 mos ll
R l_. lit 27
Vyp 9.2.
tre 91 AOL G 20'
lit _3_4; WC 9_§_
A3 13.3.
LCN G 11. Vyp 99
1‘1 R L3.
WC 91% hrb 10
19
AS 12 AOM G 08
Vyp 01 WC {99
aom 02 AS ‘99'
R _19 R _19
Nam 03 Epi 9;
hrb 22 Vam O3
shr 99
TOT R 22 hrb 14
Vyp 4O
a-2 10 AOH G 28?
WC 15
LIT G 13 AS 1;;
WC 99_ R 03
AS 29L 11x 15
R 19_ Vyp 20
Vamfigg
hrb 08 A-l G '19
mos 9; WC ‘9;
lip 39. AS 51?
R 18
11x 04
A-2 G 18'
WC 17
A3 99
Vyp .19.
aom‘lL
R 20

BEE

GLI

AMO

CPD

CAR

CHM

COM

COP

COR

4>

OO
(Db)

47-9
37

vac:
O‘R-

c>c>
c‘h‘

lo

TABLE IX. cont.

23-ILX(Cont.)

1

EPI WC 17
de
AS 31
Car 45
R 08
Gau‘fig
Led
Vamn99
hrb
ERI R '99
Chm.§9
shr

GAU'WC 12
AS 13
Vig 23
Vyp 01
R 89#
hrb
TED AS ‘99
aoh
R 12.

MAI WC ' 312’.é
Vam 21
a-2
AS 27
bar

PRU WC 03
AS 05
Car 20
Corl9§
Ors 16

RHO WC 14

AS 21
Cor‘99

R 29
Can 55

168

ILX, KAL
1 4
4 TRI WC 26‘.‘ 3812»L
de 24
39# 'Mai 14
25 AS 15 19_
03 bar 99_
91 VAM we 24" 16
Mai 06
10 AS 06 05
R 21 16
.12 Can El .11
shr 05
l9_ hrb ‘99
£1 v13 AS 00 99
R 9.5. 19.
15 Can ‘2_l_+.
11
VIC WC 04 02
AS .99 O4
46" R ._g 03
20 Led _2_1_
shr ‘19
‘91
46 vox R 10 1o
42:? a.]__ E
25Ll VYP we 07 18
shr 02
20 AS 22 12
16 Epi 12
00 bar 22
R 19 O7
04 Can 99 _1_1._
19, hrb ‘9;
29 TRE we 16
lit 26
A3 91
04 R [lg
19, aoh ‘9_
.251

4

SHR WC 22"
AS 01

R 22

Can 09

Led 02

hrb 09

HRB WC 04
A391

R 43"
Can 08

M03 WC ‘99
A391
bar‘ig
aoh‘gg
R 10
Can 20
Led 99_
11t‘99
aoh 23

LCN we 9;

4

AOM.WC 19
AS 04
bar 99.
R 9.2.
Led _1
hrb 08

AOH we 99
AS 33'

Led 54

bar 18

20
Eri O9
Vox‘ll
DG ‘99

CHM R

COR R 05
ERI R 25
Vox 99_
DC 12
05
LEDR 9;
mos
11p
DG 13
O3

07

02

TABLE IX. cont.

24-KAL(Cont.)

1 4

VIB R ‘99 99
VOX R 73# 23
DC 12 Ql

SHR R 04
DC 10
Chm 21

MOS R 34
lip 20

DC 12
LCN R ‘99
Cor .9;

DG 06

LIT R .99
mos O4

11p 02

DC 99

LIP R 29'
mos 04
lit 11

DC ‘9;

ADM R ‘99
lip 27

DC _9

A-I R ‘l_
A-Z R 20

25-KAN

1 4

VOX DG 20' 15
TRE DC 29?

H

169

KAL, KAN, LED

4
MOS DG 03
tre 21
LIP DG 99
tre 21
26-LED
1 4 H
AMS AS 99 02
aoh, 93
CHM R 17
11 19 01
Vib 99
shr £2
119 2.5.
a-2 ‘9;
DG 9&# 99# .29?
shr .99
COPAS _19 _1_2_
aoh 39
COR AS 1;; '12
hrb 99
aoh 99
R 11 1.8.
Epi 99
V1399.
EPIAS 99 19
R 91 99
Vam‘lg
aom ‘19
ERIR 02 9_2_ 27
Spl 17
shr '99
mos '99
lip 12
DC 13 18
00 05 03

GAU

ILX

SPL

SPP

VAM

VIB

VIC

08

TABLE IX; cont.

26-LED(Cont.)
l 4
VOX R 99. 19.
mos ‘9;
lip 9‘1
831117.
hrb SL1
VYP AS 99. ‘91
shr ‘99
R9991
shr ‘99
mos 10
DC 50# 42#
TRE AS ‘99
mos 99.
R 19.
shr ‘99
mos 12
lit ‘9;
11p 08
aom ‘99
DC 06
99_
SHR AS 18
Vyp 28
mos 08
11p ‘9;
aoh 01
R 47?
Chm 61
Vam 65
Vig 30
88 9.2.
Chm 55
Vyp 12

10

170

LED, LYC, MAI

LIT AS ‘99
mos 02
1199.2
aoh 99

R 2.9.
mos 02
lip 14

DC 09
mos 27
11p 22

BAR AS ‘9;
1191.9.
aoh‘gg

R 11
DC 92,

4

AOM AS 23
lip 07
R 44"
11p 34

DC ‘9;

AOH AS 72#
mos 58
lip 61

R 00
11x 10
V19 19
mos-99
lit‘ig
lip .12.
aom 17

DC ‘99
Chm.9§_

A-l AS 11
Can 23
mos 07

I"

CAR AS 1

NM
O‘H

COP

COR

EPI

MEL

ORS

PRU

PTR

R110

VYP

CPD

1

AS 24
Cor 99
Rho 14

AS 34'
Mel 21
Rho 17

A311
Rhol_l

AS 48?
Rho 36

A399.
Pru‘g9

AS 14
Car 27

A399
Mel‘gg

AS 36'
Cor 21
Mel 11

AS 09
Can 20

28-MAI

H

TABLE IX, cont.

28-MAI(Cont.)

1

COM G 14
01
2'5":
06
01

WC

AS
Gau
Pru
Tri

COR WC 22'
Epi O9
Rho 03
AS 0

GAU G 24?
43#
WC 27#
381‘-
Epi 25
Tri 25
Vam
AS 11
Pru
VYP
tre

4

99
11
03
11

30?

0‘0
IN H Ln

32.9
23"

15
27

08
10

21"
30?

29?

H

08

32
10
44
45

11

22”

52#
67

171

ILX WC 31#
Vam 21
AS 27
bar

LED AS 99,
aoh

MEL G 14
09
WC 24"
16
Epi 05
A391
shr

24?
05
WC 06
AS 24
Pru 14

ORS G

PRU G 18'

O
\l

O

9—. 59:73

1

RHO WC 52:7,:
57#
Tri 45
Van 45
AS 09
Can
Pru

TRI WC 48135
46#

Rho 29

AS 20

341%
62#
03

VAM WC
AS

25?
20'

VIC G

WC
19'
Vam 09
A311

19'
13

VYP G

WC
07
AS 10
shr

TRE G
WC
AS

SHR G
WC
Vam

Me 1
Ptr

VYP

55#
42%

32#
23”

04

36#
33#
9_s_

29#
06
14
06

32#

24
35
O6

60#
33%

00

04

O7

13
03

TABLE IX, cont.

28-MAI(Cont.)

4

21'
Car 42
Ptr 05
WC 05

Car 21

Can 99;

Ptr 15
A399.
Cor 99_
Epi‘gi
Mel 05
Ptr 19

HRB G

MOS G

mossy;
aoh 23

AOL G 04
WC ‘99
A311

tre 99.

AOM G 11
WC ‘99
A319

bar 07
11p 22

MAI, MEL
4
AOH G 06
WC 14
AS 12
Led 37
mos 24
A-2 G 07
WC 36#
A3 9.8.
BEE G 99
WC 16
a-2 04
A3 2..
29-9-IEL
l 4 H
AMD G 92, ‘92
RC 9.6. 99.
A3 19 91
Rho 9_. .99
hrb ‘99
AMS AS .99 ‘19
ASH G 37# 51#
WC 03
CAR G 91_ ‘99
WC ‘99’ 04 ‘99
AS 15 15 529
.Epi 02
Pru O3
Rho 05
hrb .9;
COM G 12
15 ‘99
WC 25? 10
O3 05 92.
AS 19
19 9.9. 15
Rho 28

1172

1

COP WC 16
3319
17
Cor 99.
Lyc O7
Rho‘99
a-2

COR WC 10
Cop
A311
35'
Lyc 22
Pru
Rho 01
hrb

O4
29#
36#
99
14
Rho 01

VYP
hrb

EPI G
WC

AS

21"
O4
29R
13
As 11
Rho

VYP
hrb

GAU G

WC

LYC AS 48#
Rho 36

14
O9
24"
16
Epi 05
A3 91
shr

MAI G

WC

30?

18
03

06
31

22"
OO

22

65#

20

O6

05
13
27

TABLE IX, cont.

29-MEL(Cont.)

l

0R3 C 16

hrb

RHO WC 33#
08
AS 12
58#
Pru

VYP
hrb

RUB G 42#
WC 99.

AS 17
Rho 29

TRI WC 293
14
AS 29
Rho

4

32'
48
O7

12
10

60#
48

46

45#

10
14
04

H

02

O
O

H
H

63#

51

roki

173

MEL, NEM

37#
31#
02
02
.99
Rho
hrb

VIC C

WC

v10 G 19'
Ash

VYP G 25?
WC
Epi
Rho
shr

hrb

SHR C
WC

hrb

HRB G
WC

Rho
MDS G
WC
hrb

lip
aoh

LIT G
WC

Rho
mos
lip

08
02

4

LIP G 10
WC 03

AS 24

shr 14

hrb 01

mos O9

aoh 11

AOM C
WC

TABLE IXJ cont.

30-NEM(Cont. )

1
VIB R 05

VIC R 93
shr
1cn

VYP R 24

IRE R
1cn
aoh

SHR R
Vig

VYP
1cn

HRB R
6811
Vam

VYP

MOS R
Vig
VYP
1cn
lit

LCNR

LIT R
lcn
lip

LIP R
Vig
lcn
lit
aom

AOM R
Vig

AOH R

VYP
Ian

0
H

H
N

l

1...:
7"

Pi“)
aaa>

I

F‘ \l
C>RJUI ¢~
N’UI 5?":

3 l

15
25

0‘0

hi no
C>C>~J Caox C>h3
P‘O‘ O\

174

NEM, OPU, ORS

31-OPU

DAN C 01
PTR G 15

BAR G
Dan

32-ORS

AMO G O 2

CAR G 06
WC
AS 38”
Epi
Pru 99

COM G 07
09
shr
WC 05
shr
A3 9..

27?
14

O4
23

MIRISIS a
O\

l

r:
h)

IS

0
H

ISIS

9&_

Z59
63

'0 O O
sqhaoflhard
u: H

RDP‘k) I

EPI G
WC
AS
Car

GOO GOO
$‘UDF‘

GAUG 07

99

WC 05
shr

AS 1..

LYC AS 06

Pru _9

MAI G 24?
WC 06
AS 24

Pru 14

MEL G 16
06
wc 93,
A3 12.
Car
Pru £1
hrb

28#
637?
shr
WC 9%
AS 641?
Mel
Rho

PRUC

¢~

1:319

[31515.3

N 0
coin:

[SIR

c~
O\

l—' N O\ N
lo

FDCD
hJ\J

TABLE IX. cont .

32-ORS(Cont.)

PTR C 05

RHO WC 02

SOL C
TRI WC 10
3

VIC C

VYP C 05

H

175

ORS, PRU

AMS AS ‘99
Car‘91

21"
39#
Ors 22
WC ‘99

ANM C

ARC C 99
Ors‘gg

251'
9;

ASH G
WC

CAR G 99_
WC 07
AS 629
Cor 33
Epi
Ors 53

22"

lo
C:

O
u

30#

N00
WON

40
08

1
COM G ‘99
03
shr
WC 09
11
AS 42?
03
Mai

COP WC 9%
AS 10
16
Cor{99
Ors 31
a-2

COR we 99
AS 10
21
Cop 13
Ors 37
Rho

EPI G ‘99
WC 16
03
AS 10
18
Car‘il
Mel
Rho
hrb

GAU C 10

O6

WC 18
22"

AS 29

Mai

ILX WC 03
AS 05
Car 20
Cor‘99
Ors 16

H

O
loo 08
Ix.)

22

176
TABLE IXL cont.

 

PRU, PTR
34dPRU(Cont.) 4 1 4 H
1 SHR G 4121E CAR G 9_9_ 1_1.
Ors 25 hrb 88
PTR G 99 we 20' we I"
.11 Can 09 15 09
we _19 VYP 32 hrb— '71
03 AS 9?. AS 24 13"
AS 04 Epi l4
l9 HRB G 9.}. Ors 25
Car 99 WC 11 Pru 43
Mel AS 20 hrb 93.
hrb Car 05
Epi 32 COP we 22' 24"
RHO G 04 Mel 9_9, hrb 13
91' Ptr 50 AS 36'
we 23 Rho 951 _11 9_5_
16 Van 99
Car 05 AOM G 07
Cor o4 WC 92. COR wc 17 14
Mel AS 29_ Cop 99_
Ors 27 Rho l9, Epi 29
AS 34' _2_5_
SOLG 04 AOHG 07 16' 12
Graig; WC 05 Cop-OB
we 99 AS 2.1. Mai. 22
M61 _O__9_ Mel 22-
TRI we 13 Rho 91
01 9]: A-Z G 14 hrb 12-
AS 11 99 we 03
Mai AS 16 EPI G 13
Ors 25 Cop 29 WC 259 ‘99
VAM we 09 GLI As 99 Cor _3_9
9_9_ 9; Ors }__6_ Gan _3_2.
Gan _2__2_ Rho 92; hrb g9
AS 11 AS _19
oo 9_9_ 35-PTR 25 17
hrb ‘99
VYP G 02 1 4
99 04 GAU G 299 329
shr _2_6_ AMOG 91 02 22" 18'
wc 95_ 99 we 13 93; we 13 05 __
shr _2_5_ AS 01 99 22' 16 99:
AS 05 99 oo hrb .2... Epi 35
hrb 04

1304911
hrb 22

TABLE IX, cont.

35-PTR(Cont.)
1 4
LYCAsg;
Me1.§g
MAI G 15 20"
23" 35#
WC 08 ‘1;
1112».
Gau‘gg
511.1.
Tri
shr ‘41
MELGggzo'
91. 05
WC .1; O8
13_ 06
AS 47#
32' 32'
Pru 48
hrb '91
0R3G9_§_
119.6.
w09111
AS1111
Car 24 ‘22
hrb .El
PRU G 32$
119.2.
WC 19.
03 05
AS 04
19.11
Car 40
MEI ‘4;
hrb _J;

H

H H
Ma) U1
1“

H O
|~ lo

177

PTR

H

RHO WC 1
AS

Cor 32
Mel 02
Pru
hrb

TRI WC .11
04
Cop
hrb
AS
29.
M31
Mel‘gg
VAM we 2;"
10

41"

_2__§

AS

27#
37#
we '
20 '
AS 10

VIC G

18
03
WC

VYP G

lol“
00 w

shr
hrb

TRE G
WC

SHR G

WC

Mai
hrb

.p

17
g;
14
08

14
21
42

04

13
O3
03

00

H
O

C: c>u>c>
U! ~1¢~$~

%

26"

F3 rd
ox u:

10
27

4

48%
Car 91
WC 60#

Car 82
AS 70#

HRB G

MDS G

WC

NOOOO
\lNChU‘l-d

AS an.-
hrb 02
lip 9.1.
aoh‘ig

LIT G 06
WC .1;
AS 07
hrb 25
mosugg
lip 9.6.

BAR G _13,
we 23"
As 12.

18
WC 09
AS 29
hrb‘gg
mos 11
aoh 18

LIP G

AOL G ‘9;

WC 07

AS 10
shr 21

hrb 29

23"
11
9g

AOM G
WC
AS

4
AOH G 01
WC 12
tre 22
As 12.
hrb 91_
mos‘li
1113.13.

TABLE IXL cont.

36-RHO
1 4
AMD‘WC 01 .91
AS 19 32
Cop 06
Cor.Q§
Mel 31
CAR WC '16 ‘1;
Com
Maifyi
hrb ‘31
AS 19 13
Ann 02
Epi 03
Pru 99.
hrb .gg
COM we 41%? 26%;E
28? 18
As 91
O7 ‘94
Cor 29
Mel 19
COP WC 42# 32#
Cor 26 18
Dan 25
AS 32 '
34' 22
Cor §Q_ '11
Mel
a-2 37
COR WC 34# 28'.‘
Cop 01 07
Mai 18
AS 29 29
59# 43?
Pru 33
hrb 31
R 13
DAN WC 269 45#

H
NH
H

01

15

02

178

R110

1

EP1 WC 53#

27?
Mai 12
Tri 05
AS 29
23
Amo
Mel
hrb
R 9.1
GAU WC 32#
20'
Epi O9

wig;

Tri 04
AS 06
Mai
Mel
R 05
11x 42

ILX WC 14

AS 21
Corlgg

R 29
Gan 55

LYC AS 36'
Cor 21
Mel 11

MAI we 52%6
571‘
Tri 45
Vam 45
AS 09
Gan
Pru

33#
08
12
58#
Pru
VYP
hrb

MEL WC

AS

53%
30?

19
09
06
13

35#
23”

12

55#
42#

12
10

60#
48

46

18

31#

12
29
27

32#

24
35
O6

06

63#

51

1

0R3 WC 02
AS213.
Pru‘gg

PRU WC 04
91_
23
16
Car 05
Cor 04
Mel
Ors 27

AS

PTR WC

l°|°l*‘
\l UI Ch

AS
20

Cor 32

Mel 02

TRI we 64;}
4%
Cop
Mai 25
AS
09
Mai
Mel

VAM WC 632’?
39#

Mai 06

AS 31'

10

R 11

VIB AS ‘1;
Cop 22
Mel‘gg

R 9.6.

59;}
24"
12

60#
12

20
02

33'

13

18
05
28

11

TABLE IX, cont.

36-RHO(Cont.)

1

VIC WC
06
AS 04
Mel 23
R 17

VYP WC
01
Com
As12
Epi‘gg
Mel

HRB WC
Car

Mel
Ptr

MOS WC

Mel
hrb
lit
11p

LIT WC

Cor
mos
11p

LIP WC

Mel
hrb
mos
lit

4

21'

01

19
31

16
00

Q

179

RHO, RUB, SPP, SOL, TRI

35#
22
47?

17

A-2 WC 25"
M31 12
A811
Cop‘gg

Cor‘gz

hrb 92.

23"
9_2

BEE WC
AS

GLI AS 33'
hrb‘gg

37-RUB

“

1

MEL G 42%
chg
AS 17

Rho 29

Mel 6

457%

39-SPP

LED R 22

H

AMO we 99
As 11
Cop‘gg

11313 2_s_

22"

O7

31#

O7

07
_qg

07

TABLE IX. cont.

4l-TRI(Cont.)
l 4 H
CPD WC 09 40#
Epi 3O
11x 26
M12.
Com‘_1

COM WC 2927 14

20' 95.
Rho 09
AS 22 09 ‘gl
de 32
M31 3
COP WC 30? 43%
Cor 33
Epi 19
Rho 14
AS 29 91_ 25
Vam 14
COR WC 22' 29’.‘
GOP 93. 9.9.
Epi 92.
Mai 24
Rho 07
As 12.
12 ‘12
Cop 29.
DAN we 9g 9g
Rho 29.
EPI WC 48# 512‘?
52# 35#
de 22
Gan 42
11x 23
AS 19 03
Vam 30

Vyp 04

180

TRI

l

GAU WC 325?6
38#
Epi 19
Mai 25
AS 15
Mai

Vypll

ILX WC 26?
de
Mai 14
AS 15
bar

MAI WC 48$
46#

Rho 29

AS .20

MEL we 29!?
14
As 29
Rho

ORS WC 10
As 11
Pru‘gé

PRU WC 18

01

AS 11
Mai

Ors 25

PTR WC ‘1;
04
Cop
hrb
AS
29_
Mai
Mel'gg

4 H
30#
10
O4 10
lg
38#
24
_1_5_
.2.
32#
23"
O4 60#
10
14
04 05
_2_2_
01
9_6_
9_1
_1___ 20
_2_§_
O4
13
O3
03
99
00 10
27

1

RHO WC 64#
45%
Cop
Mai 25
AS
09
Mai
Mel

RUB WC 04
AS 26

VAM WC 4411
10
Mai.21
Rholgg
AS
37'
Vyp 48

V13 AS 20
Vam 09

VIC WC

59#
24"
12

15
O4

41:1
09

21
O6

18
32
04

06
23

18
05
28

21

H
'N'O

TABLE IX, cont.

42-VAM

l

AMOWC 9_3
As .12
Cop‘gg

CHM R 06
Eri
Vox
shr

COM WC 2817
11

shr
as 9.1
11
Cop 26
V13 21

COP WC 06
AS 40
47?
Cor 61

COR WC 14

AS 49#
15
Cop‘gg
R 9.2
Epl‘gg
Vyp 29,
hrb

EPI WC 451’]E
12
AS 17
‘92
Tri‘gg
R 69#
Cor
Can

4

I°|°
on w

36
19
68
16

26?
29?
19

O6

01

12

12

O4

08

181

VAM

1

ERI R 13
Chm
Epi 34
Can
Vyp 27
shr
hrb

24?
13
AS 00
R 34
Epi l7
Eri
Vyp 20
hrb

GAU WC

ILX WC 24"
Mai 06
AS 06
R 21
Gau‘gl
shr
hrb

LED AS '1;
mos
R _2.
Epi 92.
le
shr

MAI W C
AS

NEM R
aom

PRU WC O9
‘92
Can
AS 11
00

42"
3o

65

27
52

32#
42#
99
60#
74

26

29?

33#
OO

PTR WC

RHO WC 63#

TRI WC 44 '.'
10
Mai'gl
Rho{gg
AS
37'
Vyp 48

V18 AS 32
R 21%.
Epi 29_

VIC WC
20'
shr
AS151
R12.
V7921

VOX R ‘11
Chm
Eri
tre

H
04>

0030
up?
I“ l°
on o

HO‘
‘41:

11

Pl“
o 4‘

TABLE IX, cont.

42-VAM(Cont . )
1 4
VYP WC ‘93
9.3. 11
AS 14 ‘12
Tri 36
R 47" 25
Epi 31
Eri 36
Can 05
shr 08
hrb 04
TRE WC 04
AS 15
R .22.
Chm .yg
shr .1;
mms ‘43
lit ‘3;
lip 91
SHR WC 351'?
Com 21
Can 24
Vig 20
Vyp 57
AS 01
R 51”
Chm 41
Eri 41
Led 67
Vig 64
hrb 41
lip 61
aom 62
HRB WC 04
AS 09
R 37":
Can 15

182

VAM, VIB

LIP WC 05
As 2?.

R 20
Chm 31

Eri 35

tre 39
shr 43

bar 30

4

AOL WC ‘96

AS 02

R 11
shr 02
hrb‘gg
mos‘gg
lip 24

AOM WC 1

A-2 WC 16
Mai 05
AS 24
Cop O6
aom 12
R 15
Gen 03

GLI AS 25
43-VIB
1

AMOAS 9_2
0092.1
7

CHMR 2

TABLE IX , cont.

43-VIB(Cont.)
1 4
COP AS 23 08
Cor 34
Vam 09
CORAS 04 9};
CopZé.
Rho 14
R 21 19.
EPIAS 9_5_ 9_9_
R 09 ‘13
Vam 21
GAUAS 1_7. _1_4;
R 06 ‘16
11x 24
ILX AS 00 98.
R 2.5. 11
Gau‘gg
KAL R 29_ .12
LEDAS _0__8_ 92
R _1 .21
NEMR 05 27
RHOAS _1_3_ _qg
Cop 22
Mel‘gg
R 9.1
TRI AS 20 16
Van 09
VAMAS 32 g;
R 9.1 17.
Epi 29.
VIGAsgg y;
R 2.. 9.1.

H

08

183

VIB, VIG

VYP AS 99.
Can
shr

TRE AS
lit

SHR AS
lit

HRB AS

MOS AS
hrb
lit
11p
a-2

lip

LIT AS

LIP AS
shr
hrb
mos
lit
a-2

AOL AS
lit

A-2 AS
lit

44-VIG
1 4
AMO G ‘Q1 06
WC 91_ ‘92
AS 25 29
ASLG 301.2 17
we 9_1_ 03
CAL G 32% 20'
WC .99
AS
R
CPD G ‘11 SE;
WC 92. ‘Q1
As 11
Com‘zl
CAR G
02 gg
hrb ‘gg
WC
21' 17.
AS
21 20
hrb 10
CIIMR 91 9;
DR 18. 11
shr ‘38
COMG _g
92' 03
WC 10
O9 05
AS 28 lg;
de 38
COP we 91 9};
As 11 17.
Amo{£§
vam9.1 9.9.
a-2 Q§_

NHO
NI—‘N

0
L11

00 lo
o~u> r:

H O
IN!»

'IABliiiflx.

44-VIG(Cont.)

l

CORWC _Q_8_
A3 9.9.
Amozp.
Cop 05
11x 05
Vib 08
R 57?

ERIR 9;

Spl

shr

DG 11
21'

24?
11

GAU G

WC

21'
Van
shr
AS 30
R 21
shr

ILX WC 04
A518.
R19.
Led
shr

LED AS 19_
R 23
Cor 4O

shr
DG 18
07
MAI G 25?
20'

WC

19'
Vam 09
A811

cont.

4

10
02

14

H

25

13

08

O4

O7

MEL G

ORS

RHO

RUB

SPL

SP?

184

VIG

1

37#
31#
WC
02
A311
Rho
hrb

Spl-2:2
DG 08

4

34#
20'
19'
02
1g

35#
26H
05
16
O8

21'

Ol

19
31

12

08
02

H

TRI WC
12

A511
Vam.03

Vib

VYP

VAM.WC

aom
aoh
DG

21

lc’lH
ON P

‘O
0)

O H l-‘D-‘H HO O
4.\ (SIM r—I-DH 00,... w
H

TABLE IX , cont.

44-VIG(Cont.)

4

SHR G 54%
WC 57#
Vyp 73

AS 19
R 56?
Led 45

Vam 67

DC 14
Chm 48

HRB G 15
Car 25
Ptr O3

WC 03

AS 20
Mel 30

R 01
DC

MOS G

LCN G ggfl
2.5.?
shr‘lg
WC ‘3§#
2.2-!

A3 25
a-2 36
R 9.1
Nem‘gg
shr 13
DG 01

AOH G

A-Z G

185

VIG, VIO, VOX

4
TOT R '92
a-2 28

LIP G 04
WC 04

AS 14

a-2 04

R 9.1
Led{14
shr‘ll
mos‘gg

DG ‘94

12
WC 07
AS 04
R 22
Led 03
Nem 33
shr O7
aoh 36
9991

AOM G

02
WC 18
A311
R 31

mos l9

aom 41
9917.

22"
shrllg
we 21'
As 22.
lcn‘gl
R 19.
tot.14

GLIAS _2_2_

ASL

CPD

COM

MEL

CHM

ERI

GAU

ILX

KAN

45-VIO

1
G 12
G 33%
G 45#
G 20'

G 19'
Ash

4

41%

0
VI

0
H

26?
07

46-VOX

70#

83
11
23

03

l°|°
0‘ L0

0
\O

0
La)

LED R ‘12
mos
lip

DG .11
hrb

POS DG

TRE R

aoh
DG

SHR R
Chm
Vam
VYP

DG
Chm
hrb

HRB R
Chm
Gau
VYP
a-l
DG
Eri

O8
42#

TABLE IX; cont.

46-VOX(Cont . )

4

22
tre 04
lit 12
lip 00
aoh 04
DC 15
hrb‘gg
lit 05
1n>93

MOS R

LmflR 91
DC 26
hrb 16

22
Chm'll
a-Z 47

TOT R

‘24
tre;§1
aoh.y&
DG ‘16
hrb 07
lit‘QQ

LIP R

AOM R ‘13
liplgfi

aoh 00
a-1‘g;

DG 1Q;
MM1R 40'
tre 29
DC 28?

A-l R 37
Chm 26
442 R g;
Chmhgg
tot 43
aoh IO.

AMO G
WC
AS
Epi

HOO
luwloo

AMS AS ‘21
Can

ANM G 15
20'
shr
WC 27?
CAR G

WC

H 1-- to
halo '0

AS

Epi‘Qfi

Can 21
Ptr

CHM R .93
Vam
Vox
shr
tot

DG ‘22?

Led‘gg
shr

COM G
WC
Rho

AS 18
Gau‘gl

186

VOK , VYP

N
O\
RR

1

IOOHH
xiv-{>0

H
CW

1

COR WC 04
AS 00
Can
R 35
Epi
Nem
1cn
EPI G 00
WC
21'
AS 37'
R 38'
Can
Vam O9
hrb

23H
20'

GAU G

WC
16

shr

AS 53#

R 38'
11x 48
Vam 26
tot
hrb

ILX WC O7
shr
AS 22
Epi 12
bar
R 19
Gau‘gé
Vam 10
hrb

O8

O8
23

15
05
26
25

17

16
05
10

O9
62#

TABLE IX. cont.

47-VYP(Cont.)

1

LED AS 29.
shr

R .29.
shr
mos

DG 50#

Chm-43

LXC AS 09
Can 20

19'
13

MAI G

WC
07
AS 10
shr

25?
22'

MEL G

WC
‘91
AS 00
Epi
Rho
shr
hrb

NEMR 23

ORS G 05
01
shr
WC 13
shr
A519.
PRU G 02
99'
shr
WC ‘Qé
shr
AS 05

42% 10

'187

VYP

PTR G

Mel

SPL AS
R
DG

TRI WC
10
AS 43?
Rho
Vam 53

VAM‘WC

VIB AS 99.
Can
shr

R 08

P‘C>C>C>C>
rgcahaLpL»\o.P -D

P‘F‘

OH
00

0
lo

19'
O7
O6
16

35#
22
479

17

1

49#
259
shr
WC
09
AS 17
Can 01
Rho
R 20
Cor 00
DG 99_

VIC G

VOX R ‘22
aoh

9393

TRE G
WC
As
Can
R

SHR G
Ors
WC

aom

Chm
Eri
Vig
DG
Chm

HRB G
Car
WC

Epi
Mel
Ptr
shr

Can

01

O H HI—‘H P'"
3* ano hd$~hi a)
H

O
N

 

TABLE m, cont.
VYP
47-VYP(Cont.) 4
4 BAR G _1_1
WC 9;.
MOS G 25'.‘ AS Q;
02 R ‘1;
WC 9; shr _1
22" mos._1
lip 08 lip 22
AS ‘il aoh g;
Led 9; DG 02
shr‘gg
R ;g LIP c 10
shr i1. WC 2}.”
bar 4g mos lg
DG 05 AS 10
shr 3O
LCN G 293' mos 93.
12;" R 17
shr Qg shr 31
we _2_y mos 3;},
‘12 lit 03
As .lé aoh 06
R .11 Do 29.
Nam 07
mos £1 AOL G 13
DC 331} we 1;
AS 14
TOT R ‘gg Can 34
Ohm 48 shr 01
R 19
LIT G 07 shr 02
WC ‘19 mos 27
AS {1; DG 06
Can 04
R 18 AOM G 16
shr 28 WC 9};
mosigg AS .Qi
11p 08 Can 9_2
aoh 07 shr 44
DC ‘Q;' R 17
Epi 27
shr O7
hrb 28
aoh 05

188

AOH G

A-l G

GLI AS 09
Can 20
aom 20
a-Z‘lg

CHAPTER VI
HIGHER POWER REGRESSIONS

From Chapter III, we know that the zero-order
correlation coefficient indicates by its sign whether or
not the Species or variables in question tend to co-occur,
i,g, whether high values of both are found in the same or
different places. By the magnitude and significance level
of these coefficients we get an idea of how rigorous and
sharply defined this relationship is, and what the proba-
bility is of its being due simply to chance. With partial
correlations (Chapter V) we can examine the correlation of
one pair in the light of other variables, thus discovering
to what extent a given correlation is influenced by corre-
lations of each pair member with these variables. This
suggests possible causative agents of the correlation.

But it became increasingly evident during the
course of the study that there may be certain subtleties
in the physioecological relationship between two species
which are obscured by simple correlations. Even straight
line regressions which show the magnitude of a relationship
and not just its consistency, are not always adequate. As

189

190
a simple illustration, let us imagine a case in which val-
ues of the dependent variable, y, are relatively low at
both low and high values of the independent variable x, but
high at intermediate values of the latter (Figure 11 C,
page 198d). Now, a simple regression which best fits these
points is a nearly flat line, and the correlation is cir-
cumneutral (-0.06) which.might cause us to pass over the
codistribution of such a pair as indistinguishable from
random. But now, if we calculate the quadratic regression
of the pair (see equation 2 below) the resulting parabola
fits the points much better (correlation +0.80), and is
certainly a much closer approximation to the curve which
we could fit "by eye". This is similar to several examples
from.my actual data for the DG gradient segment, of which
the pair vaccinium angustifolium var. gigggm (gig, no. 44
in Table VII) plotted upon Leggm_groenlandicum (Led, no. 26)
has a nearly flat linear regression with correlation +0.03
(Figure 14 C, page 209). Now, both these species seem to
produce little cover where the dominant shrub Chamaedaphne
calyculata Lhm, no. 11) is lush, _I_.._e_<_i_t_t_m_ (correlation with
Chamaeggphne -0.43) perhaps someWhat more so than the
vaccinium taxon (-0.15 with Chamaedaphne). In areas which

are relatively free of the near-ubiquitous Chamaedaphne

191

perhaps both flaccinium and Ledum can thrive, hence the

 

initial rise in the cubic regression line (equation 3) of
the first as the second increases. But aslggggm clones
become established and grow lusher, they may exclude
vaccinium (by competition, antibiosis or both) as effec-
tively as Chamaedaphne does, so vaccinium decreases with
further increasing‘Leggg. The correlation correSponding
to this cubic regression is +0.25, significant at PR<:.02,
but when we "average" the two tendencies with a linear re-
gression neither of them comes to light.

Ideally, we should examine scatter diagrams of
individual points in order to view all possible subtleties
in the correlation between two variables. But in a study
of the magnitude of this one,such a thing is simply not
possible. Even When we are dealing with a relatively few
variable pairs, moreover, such misplaced thoroughness might
lead us to place undue importance on a few eratic and aber-
rant points, whose positions may be due only to random
fluctuations, unique local conditions, or even measurement
error. The present chapter is, therefore, a search for a
workable solution to the prdblem.of the insensitivity of
straight-line regression, and for a.method which yields

valuable additional infermation within a realistic period

192
of time, while avoiding the pitfalls of overly detailed
examination.

Accordingly, I first plotted several hypothetical
sets of points designed to illustrate all imaginable situa-
tions. Because these points were purposely chosen to fit
certain kinds of curves, their mean distances from those
curves are considerably less than one would generally find
in a natural situation and the correlations are extremely
high, but most of the information derived from them should
be applicable to real data. Linear, quadratic and cubic
regressions, actually multiple regressions of the forms:

(1) y a a + blx,

(2) y - a + blx + bzxz,

(3) y e a + blx +be2 + b3x3,
respectively, were calculated for each of these sets of
points, and the resultant curves plotted (Figures 8-12, pages
198a-198e). For convenience the zero-order correlation
between the two variables and the values for b1 of equation
1, b2 of equation 2 and b3 of equation 3, are given with
their significance levels beside each figure. The quad-
ratic will always be a parabola or a part thereof, while
the cubic regression will be all or part of a symmetrical

curve in the form of an S turned on its side; the first may

193

be concave downward or upward and the second may have
either its right-hand or left-hand branch concave down,
and apices may be located anywhere (see illustrations).
Regressions of still higher power would be nearly impos-
sible to prove significantly different from their lower-
powered counterparts, because of the rapid increase in
standard error as one adds more parameters (cf. Ezekiel
1930). Moreover, I felt that curves of the fourth power
or higher would generally be meaningless biologically, or
else hopelessly difficult to interpret.

The results and conclusions of this hypothetical
study were then applied to data for the DG (Dawson-Green-
wood peat) gradiant segment. This segment was chosen be-
cause it is the lowest of all in number of species and
consequently, for an exploratory study, the least complex.
Data only from quadrats of intermediate size (1/4 m2), were
used so that environmental variables could be included.

In deciding which regressions to calculate and
plot, all those possible combinations which would logically
seem to be meaningless were first eliminated. For example,
it is unlikely that the small, relatively ephemeral
vaccinium oxycoccus (Egg) appreciably influences the dis-

tribution of the large robust clones of Chamaedgphne caly-

194
M (SEQ). X. oxzcoccus might conceivably influence
seedling germination of the Chamaedaphne but, by the time
the latter matured, would likely have shifted in its own
distribution). The reciprocal regression on the. other
hand, might be meaningful. For those variable pairs not
eliminated, quadratic and/or cubic regressions were plotted
only if the coefficients of x2 or x3, reSpectively, were
significantly different from zero at a level of P < .05 or
higher (t-test). This level was judged, on the basis of
experience with the hypothetical data, to be about that at
which the quadratic was sufficiently different from the
linear, or the cubic from the quadratic, to yield unique
information of possible importance.

In several cases, reciprocal regressions are
plotted for the DG data, 1.5. , Species B on Species A as
well as A on B. Although the reciprocal of a linear re-
gression is merely the same line rotated through 90 de-
grees, this is not true of the higher power curves. This
can easily be demonstrated by plotting a series of points
on two axes which would be well fitted by a concave-dawn-
ward or concave-upward parabola, then actually rotating
the paper through 90 degrees. It will be seen that if the

former x-axis were now allowed to be the y-axis and vice

19S
versa and Equation 2 above were again applied, the result-
ing parabolic fit to the points would be close to a straight
horizontal line. The biological significance of this is
that each such curve of a reciprocal pair may reveal as-
pects of the relationship between two variables which the
other does not, e.g. the nature of the dependence of the
distribution of A on that of B as well as the converse.

A. THE HYPOTHETICAL DATA. If we are to recognize
deviations from randomness in regressions, we should be
aquainted with the approximate range of divergence from a
flat line generally found in data obtained from randomly
co-distributed pOpulations. Accordingly, regressions, were
calculated for three sets each of 20 pairs of random num-
bers (Figure 8,page 198a). As it happens, 8 C has a cor-
relation for the straight line regression significant at
P (.05 (we would expect this in about one of 20 cases of
variables which are in fact co-distributed randomly). It
is in 8 A and 8 B, however, that the higher power curves
appear more interesting, and in 8 B the quadratic regres-
sion is, in fact, significantly different from the linear
at P < .02. If we now look at the scatter diagram of
Figure 8 B, we find a cluster of four points in the upper

left and another four in the middle near the bottom.

196
Flukes of this sort can so easily occur, eSpecially with a
relatively small number of observations, that we should be
forewarned not to treat any regression plot as if it were
inviolable. This is e3pecially true in the DG data to be
discussed below, in the cases of those bog species which
are present in relatively few quadrats. Just as with the
correlation coefficients, however, it is unavoidable that
we will observe a certain number of such flukes and perhaps
try to rationalize them.

Figures 9 A and 9 B (page l98b) are, reSpectively,
"near-perfect" positive and negative linear regressions.
As one would expect, the quadratic and cubic curves do not
differ appreciably from the line, the latter being for all
practical purposes the best fit in these cases. In Figure

10 (page 198c), on the other hand, we have what is intended
to be a geometrically increasing y. Here the quadratic
and cubic curves reveal this feature where the line cannot.
The cubic is probably the better fit but neither its down-
ward bend to the left nor the quadratic's correSponding up-
ward bend represent any real feature of the data.

The last points up another danger in these higher
power regressions; that whatever subtleties they may reveal

they are as inflexible as the line in having to assume a

197

particular shape. The entire curve on one side of the in-
flection point, probably generally the smaller half within
the quadrat between x and y axes where the data lie, may
be completely misleading. Still interpretation, in the
light of field experience, other statistics, and logic are
indispensible. In the case of Figure 10 (page198c) and in
some others, another curve could be drawn in, based on
"the eye" as well as the other curves, which is probably a
better fit than any of the computed regressions. In such
cases the latter do, however, suggest where the "best fit"
should lie. Figure 11 A (page 198d) is the reverse of
Figure 10, where y decreases more steeply at higher values
of x. Again the quadratic is necessary to reveal this.

Figure 11 C was mentioned in the introduction to
this chapter as a case where the straight line regression
shows no relationship, although the quadratic does. Figure
11 B is a similar case in which the parabola Opens upward.
Figure 12‘A is intended to represent a hypothetical case
in which there are two separate areas of increase in y.
Although the linear regression reveals a general increase
it does not distinguish between these two areas, which may
reflect two separate thresholds of influence for variable

x upon y, each representing a different aspect of x's na-

198
ture. This pattern might also result from the influence
of two different variables, each correlated with x, upon
y, or one of the steep rises might reflect the action of
x itself while the other represents a third variable.
Figure 12 B is similar but with a third variable, or excess
of x itself, actually causing an inhibition of y before the
latter begins to rise. This rise might reflect an inhibi-
tion of a competing species, which suppresses y, by x, or
a change in x's character as it increases so that it be-
comes favorable to y. As curves become more complex one
can imagine more and more possible situations to account
for them.

Figure 12 C was intended to represent a "sigmoid
curve", that is, a more rapid increase in y at intermediate
than at either high or low values of x. But as the t~val~
ues as well as the eye reveal, the quadratic and cubic re-
gressions are scarcely different from the line. Now, the
quadratic and cubic equations used here always yield curves
such that a line which is parallel to the y-axis through
the parabola's vertex or the cubic's inflection point will
cut the curve into two halves which are upright or inverted
images of each other. Hence they may be more efficient at

revealing certain types of peculiarities than others such

FIGURE 8. Hypothetical Linear, Quadratic, and Cubic
Regression Curves

Code: solid line - linear, dotted line
- quadratic, broken line - cubic.

A. Random numbers.

B. Random numbers.

C. Random numbers.

 

upco-

800-

600‘

400'

2001

 

198a

 

X
2

 

O

I'OOOH
0001
600

000'I

_[ 1 fi '7 f V T fl V fl
IOO 200 300 400 500 600 700 800 900 I.OOO

X
3
I

 

 

6004

400‘

200-I

I f 1 I V T— 1 fl ‘—‘
100 200 300 400 500 600 700 000 900 I900

 

X
3

 

 

I y I W , V I 1* T f
IOO 200 300 400 500 600 700 800 900 LOOO

FIGURE 9. Hypothetical Linear, Quadratic, and CUbic
Regression Curves, cont.

Code as for Figure 8, page 198a.

A. X“ increasing arithmetically
as x1 increases.

B. x5 decreasing arithmetically
as x3 increases.

 

198b

Q

“000-

80 O-

000-

400!

 

 

 

uooo- \\

800- \

COO— \

4do

200'

 

960 7300

 

o IOO zoo 300 450 560

Y
800

130

soo

960 I960

 

FIGURE 10. Hypothetical Linear, Quadratic, and Cubic
Regression Curves, cont.

Code as for Figure 8, page 198a.

x6 increasing geometrically as
x2 increases.

 

X
2,m°-

“000-

“800‘

LOGO-

“400-

[200‘

upco-

000-

600-

400-

200-

198C

 

 

 

G

I00

260

360

450

550

550

760

330

960

I300

 

FIGURE 11. Hypothetical Linear, Quadratic, and Cubic
Regression Curves, cont.

Code as for Figure 8, page 198a.

A. x8 decreasing geometrically
as :3 increases.

B. Parabola opening upward.

C. Parabola Opening downward.

 

 

198d

gooe-
soo-
000-
r

400‘

200'

 

. . X
.\ 3

 

 

I l I U 1 I I I l fiffi
0 IOO 200 300 400 500 800 700 000 900 1000

I900-
000-
_/
600-1 ‘. I
400‘ ' *2 I

200‘ \H‘. "c. . ’If“.

D
‘ -6“,

X
I

o :00 2'00 3'00 460 500 360 760 ado 960 n.0'oo

 

 

[poo-
soo- A¢-" a“

W \..\' a
600- f’ "__-_—_—"““—“‘fi:f————-—-——
400- 77 ‘5

200'

X
l

 

 

I t I l I I V I V ‘l
O IOO 200 300 400 500 600 700 000 000 1000

FIGURE 12.

 

Hypothetical Linear, Qggdratic, and Cubic
Regression Curves, cont.

Code as for Figure 8, page 198a.

A.

x7 increasing rapidly at
low and high values of
x , more slowly at inter-
mediate values.

x1 increasing at low and
high values of X3, de-
creasing at intermediate

values.

Regression fits to data

fitting a sigmoid curve;
x increasing more rap-
idly at intermediate
than at low or high val-
ues of x3.

 

 

L_._.

 

198e

“000-1
0 001

0004

   

4001

   

.
’ . '
7,~A'— '0’

"2

 

f 1

r I fl

" ‘ U 1*
0 IOO 200 300 400 500 000 700 000 900 ImO

600-
400-

200-‘

 

X
3

 

 

‘ T T fifi r 1 fl V I fl
0 IOO 200 300 400 600 000 700 000 900 1.000

1.000-
8004
GOO-R

400-

   
 

(3
200-1

X
3

500 soo 700 s00 9001000

 

199
as this sigmoid curve.

B. W. Graphs for
these regressions are presented in Figures l3-l6 (pages
208-211)..

Beginning with the shrubs we find that dominant
Chamaemhne calyculata appears to decrease somewhat more
markedly at low than at high values of m ggoenlgncdicmn
(Figure 13 A), according to its quadratic. If Chamaedaphne
is responsible for the reduction in cover of _L_e_d_t__m; as sug-
gested by Figure 13 of the reciprocal plot, so the reverse
may be true in microsites where _L_e_c_1__tgp_ has become estab-
lished. If so, then the threshold for inhibition of
W would seem to be at lower values of _I_.__e_c_l_u_:_n_, so
that Chamaedaphne's cover is reduced to near zero where
Qdum's cover is relatively low, and higher cover of the
latter makes little difference. Perhaps the curve does
not reach zero because parts of .I_._e_d_m_n and Chamaedaphne
clones, by chance, may sometimes be included in the same
quadrat. It is suggested in Chapter IX that perhaps such
inhibition is due to an antibiosis-like reaction between
shrub species or their mycorrhizal symbionts, in addition
to competition. On the other hand, Figure 13 A may merely
reflect Figure 13 E @5ng on Chamaedaphng).

200

ghamaedaphne plotted against tree cover Figure 13 B
yields a quadratic similar to Figure ll B, a parabola con-
cave upward. 0n the basis of all other observations, how-
ever, tree cover simply inhibits ggamaedaphne in prOportion
to the farmer's cover, and in.fact the mean of those values
of Chamaedaphne to the right of the vertex (low point) of
the parabola is considerably below the overall mean for the
Species. Therefore, in this instance, the upward-curving
arm of the parabola would seem to reflect simply a decrease
_i_n_ £13; at; _qf decrease in Chamaedaphne cover with in-
creasing tree cover, i,g,, Chamaedaphne decreases most rap-
idly at lower values of tree cover and approaches a sort of
lower asymplote at higher values, as represented by the
heavy line in the figure. I mentioned in the preceeding
discussion of hypothetical data that the higher power re-
gressions are limited by the shapes of the curves, and
throughout the curves of actual data one arm of a parabola
frequently curves in a direction Opposite the general trend
of the variables. 'With care in interpretation, however,
combination of such parabolas with the least square fitted
line can still facilitate the plotting of a more realistic

"fit".

Since shrub cover (shr) is composed of over 80%

201

Chamaedaphne, the regression of shrubs on tree (Figure 13 C)
tells us no more than that of Chamaedaphne on tree cover.
The regression of ghamaedaphne on peat (AOH) thickness
(Figure 13 D) is positive probably because of the former's
. partial replacement by Leggy; on the shallow peat near the
bog margin. In this case the downward-curving arm of the
quadratic to the right must represent an upper asymptote
(heavy line), Lg. the relative constancy of Chamaeg_aphne
cover beyond a narrow marginal zone.

The negative correlation of £31119. on Chamaedaphne
(Figure 13 E) probably reflects this also. The left hand
positions of the quadratic and cubic curves are probably
. the meaningful ones since most of the points lie here, and
they suggest that Lgdg cover essentially goes down to zero
in the lusher- than-average Chamaedgphne patches (heavy
line). Vaccinium boreale 0n Chamaed_aphne (Figure 14 A,
page 209) is probably essentially the same story as 3:19;
on W and like the latter presumably reflects
essentially complete inhibition of other shrubs by lush
W or its mycorrhizae.

Vaccinium boreale on _L_e_d__t_g (Figure 14 B), since

the upward-curving arm of the cubic curve to the right is

much longer than the left branch, seems to indicate a strong

202
tendency for these to grow together, most eSpecially where
£29.92. is relatively lush. But 1. anggstifolium var. £18119;
(_V_i_g) seems to have a tendency to be lush at low values of
Leggy; and to drop off at higher 0n98(Figure 14 C), as dis-
cussed in this chapter's introduction. Moreover, a check
of the raw data, easy to do in this case since there are
few observations beyond the means, reveals that all parts
of the curves of both y, boreale and 1. aggggtifolium var.
9.1-£9.11! on Lady. seem to follow real patterns.

The small ground shrub Vaccinium ox coccus , a

 

cranberry, seems on the basis of its quadratic regressions
to rise to its lushest value at intermediate values of
Chamaedaphne and to fall off at higher ones (Figure 14 D).
Its regression upon shrub cover (Figure 14 E) is very sim-
ilar and amounts to the same thing, since shrub cover is
nearly synonmous with that of Chamaed_aphn . Since this
agrees with a visually evident pattern in the field, namely
that 1. oxycoccus occurs with medium lush younger 93%-
daphne clumps or only on slopes sorrounding older ones, it
tells us nothing new.

1. oxycoccus and Erigphorum gpissum are not present
on the bare ground under trees or in the old patriarchal

Chsmaemhne clones. On the other hand 2. mcoccus, with

203
about three times as much cover as Eriophorum is nearly
absent from certain moss-covered microsites with little or
no shrub cover or bare ground which are probably recently

disturbed spots. Even.§;i0phorgm_is absent from the young-

est of these. It is probably these peculiarities which make

the cubic curve of herb cover, which is composed only of
these two Species, maximal at intermediate values of bare

ground (Figure 14 F), reflecting the two Species' occur-

rence near lush Chamaedaphne and other shrub patches but

not directly in them.

The cubic of moss cover (meg) plotted on M-
ggghgg (Figure 15 A, page 210) suggests that at very low
values of‘ggamaedaphne there is relatively little moss.

A clue to the reason is found in the partial correlations
between these two variables (Table IX, page 158), for we
obtain values of -O.47 and -O.46 with per cents of tree

cover and litter cover partialled out, reSpectively, as

against a zero-order of only -O.35. Thus both mosses and

Chamaedaphne are absent on the heavily litter-covered
ground under scattered bog trees, tending to make their

correlation less negative. MDss, by reason of this mutual

aversion to tree cover, is present only where there is some

Chamaedgphne (cubic peak), but it is also inhibited by the

204

shade or litter of Chamaedaphne as clumps of the latter grows
older and denser. Hence, the decreasing curve to the right
(the secondary slight rise is surely an artifact, like some
discussed above). The lower asymptote of moss cover on
spring litter thickness which one can infer from the linear
and cubic curves (Figure 15 B) is not surprising, for a
feW'milimeters of litter are probably as effective in in-
hibiting moss growth as a few dozen. The moss on AOL (fall
litter thickness) quadratic apparently reflects this
(Figure 15 C).

Lichen cover (lgg) seems to be associated with
high vaccinium boreale, high,y. oxycoccus and high herb
covers, (the latter being over 751,2. oxycoccus (Figures
15 D-F). Moreover, if one supposes that the shorter left
branches of the cubic regressions and the dips below the
x-axis represent a slower rate of increase here than on
the right, lichens must be more markedly favored where val-

ues of these variables are high (heavy lines), though it is

difficult to suggest why. The cubic regression of lichen

cover on bare ground (Figure 16 A, page 211) first goes up

rapidly and then down. But here the linear regression is
nearly flat (E - +0.07) so the cubic's down-curving arm to

the right represents a real decrease, not a slowing up of

205
the increase rate. Thus lichens must be relatively abun-
dant under the trees and senescent Chamaedaphne clones
where most bare ground occurs, and where conditions are
relatively xeric and similar to the upland. But even
lichens may be inhibited where shade is extremly great.
Alternatively, the most extensive bare Spots represented
on the right-hand side of the curve may occur in Spots
where litter has been present for so long as to kill both
lichens and mosses. Heavily litter-covered, lichenless
areas may result in such microsites, with many bare
patches exposed by wind removing litter from the once
moss-covered surfaces.

Figure 16 B probably represents a situation quite

similar to that illustrated in 15 A for moss on Chamae-

daphn . Per cent litter cover seems to be relatively high

where there is no Chamaedaphne,.ifig. under the trees,

reaching a low where young Chamaedaphne is relatively

sparse but increasing again in older, lusher patches, of
this shrub , which produce their own litter. Here, too,
the partial correlations seem to substantiate the hypo-
thesis. Litter cover and Chamaedaphne are correlated at
+0.26, (P <:.01), zero-order, but this rises to +0.38

(P'<:.001) when tree cover is partialled out (Table IX,

206
page 153), so as to mathematically eliminate the contrast
between microsites of high tree cover, no Chamaedaphne and
of no tree cover, low Chamaedaphne.

In this connection, perhaps per cent litter cover
on‘nggg (Figure 16 C) reflects in the left, downward-
curving arm of its quadratic a probable tendency for‘nggm
to become established where both trees and Chamaedaphne,
hence litter, are sparse. The rise to the right then
presumably reflects older, lusher Ledum's own litter pro-
duction. A check of the raw data seems to bear this out.

In the regression of per cent litter cover on
litter thickness (Figure 16 D), both the very steep upward
slape of the linear curve and the raw data indicate that
the cubic's slight decrease to the right represents an
Upper asymptote, a point beyond which there is little in-
crease in per cent litter cover with increasing thickness.
This stands to reason since 10 mm. or 100 mm. of litter
will completely cover the ground beneath them.whereas l or
2 mm. will cover but a thickness-related proportion, but
this is merely a sidelight from our vieWpoint and tells us

nothing about the biology of the area.

Rounding out the group of curves which it seemed

profitable to plot we have duff thickness on Chemaedaphne

207
(Figure 16 E) and on shrub cover (Figure 16 F), which is
about 80% Chamaedaphne. The cubic suggests that two sepa-
rate phenomena may be causing the positive correlation
here, as in the hypothetical examples illustrated in
Figures 12 A and 12 B (see Section A of the chapter).
The nature of these forces is a mystery, however, for
mean ADM thickness on DC is about 60 mm., mostly dead
Sphagnum. Since Chamaedaphne at best can.produce but a

minor fraction of that amount, its own contribution is

probably not one such phenomenon.

If judiciously used in conjunction with other
knowledge, higher power regressions do seem to reveal
facts and details about the microdistributions and co-
distributions unavailable from other sources. 0n the

basis of this exploratory study they show promise as a

tool in vegetation analysis.

 

FIGURE 13.

 

Regression Curves, Segment'QQ.

Code:

solid line - linear, broken

line - quadratic, dotted line - cubic
curve. The broad lines represent prob-
able "best fits” to the biological

situation.

A.

Chamaedaphne calyculata on
Ledum groenlandicum.

Chamaedaphne calyculata on
per cent tree cover.

Per cent shrub cover on
per cent tree cover.

Chamaedaphne calycula§g_on
AOH (peat) thickness.

Ledum groenlandicum on
Chamaedaphne calyculata.

208

OEHAMAEDAPHNE

 

 

 

 

B

O ' ‘ V ' I T I fl
0 I0 20 30 4O 50 60 70 BO 90
0/0 TREE COVER

3° % SHRUB coves

 

 

 

0 IO 20 ab 4'0 :0 6'0 Tb 3'0
0/0 T REE COVER

   

 

 

 

 

CHAIAEDAPHNE
30!
\
\
20! ‘
l0-l
D
o I r Y 1' V T
0 IO 20 30 4O 50 60

A0“ THICKNESS, IN.

 

 

20.94

 

FIGURE 14. Regression Curves, Segment'Qg, cont.

Code as for Figure 13, page 208.

A. Vaccinium boreale on
Chamaedgphne calyculata.

B. Vgccinim mboreale on
Ledum gr roenlandicum.

C. Vaccinium angustifolium
var. nigrum on Ledum

groenlandicum.

D. Vaccinium ogycoccus on
Chamaedaphne calyculata.

E. Vaccinium oxycoccus on
per cent shrub cover.

F. Per cent herb cover on
per cent bare ground.

 

 

209

VACCINIUH
l5-

BOREALE

   

\vaccmuvu BOREALE
3~ \

z-

 

 

 

 

    

C MAEDAPHNE

VACCINIUM ANGUST'FOLIUM
VAR. NISRUI

 

 

 

 

 

LEDUM

     

 

 

 

 

 

 

 

4..
3< ' ~
vaccmluu oxvcoccus
. '. 3- ’
.1
I 2..
' \
|< .I I \
. I \
C h I \
I \
Q I \
I I T *1
0 5 l0 IS 20 ’ ‘ D
LEDUM C r v . _-
l5 :0 45 so so
CHAMAEDAPHNE
avaccmluu oxvcoccus senses cove-n
q '51
lo-
5..
F
‘0
[ I I I o I T V -.l
0 lo 20 so 40 so so 70 so ‘so 0 ' 2 3 9

°/o SHRUB COVER

°/. BARE GROUNO"

FIGURE 15.

Regression Curves, Segment 29, cont.

Code as for Figure 13, page 208.

A.

Per cent moss cover on
Chamaedaphne calyculata.

Per cent moss cover on
litter thickness

Per cent moss cover on
AOL thickness.

Per cent lichen cover on
Vaccinium boreale.

Per cent lichen cover on
Vaccinium oxycoccus.

Per cent lichen cover on
per cent herb cover.

 

I ‘—

210

scuoss coves
%woss coves IOO

76- -

50'

 

A as-

 

 

 

 

o I ' ; fl 0 l I r '—l

O 20 40 BO .0 0 5 l0 IS 20

CHAMAEDAPHNE

96uoss coves
96 ucusu coves

 

 

 

 

904 \ ’.'

‘ N\\ l.5‘ .

_ \\ i.
so- ‘x\
\
\
\
30-
C
o I I r l fi]
0 5 lo IS 20 25

soL THICKNESS,MM.

 

Samoan coves
2.0-

% LICHEN c oven
Ls-

  
 

LO-

 

 

 

 

O.5- l
. .- F
'.o' " .1.
' fi' I '7‘ T_. I fl
'2 l‘ o a ‘3‘. ‘0 I5 2)

o o
. . .
.‘H-uo'. .I" "

VACCINIUM oxvcoccus %HERB COVER

 

sir/0%}

FIGURE 16.

Regression Curves, Segment‘gg, cont.

Code as for Figure 13, page 208.

A.

Per cent lichen cover on
per cent bare ground.

Per cent litter cover on
Chamaedanhne calzculata.

Per cent litter cover on
Ledum groenlandicum.

Per cent litter cover on
litter thickness.

A

0M thickness on

Chamaedaphne calyculats.

thickness on
per cent shrub cover.

_‘m -

 

211

96 LICHEN coves
2.0}
us-

we
A ”Stuff“: coves

0.61 ;

 

 

 

 

% BARE GROUND
CHAMAEDAPHNE ”fun“ coves

%LIIT:R coves
30W

20-

  

IO-

0

 

     
  
   

 

 

 

o F T *1 :I‘ cf a v v *v

0 6 IO IS 20 O 5 I0 I! 20
LE no N! ,0“ "Henna“. LITTER THICKNESS,MM.

I20 _,

90

   

 

0 2C To oh so
CHAMAEDAPHNE

A0“ THICKNESS. M H.
'20]

90‘

60'!
F

 

I I I I I T 1 —fi

0 no 20 so so so so 70 so so
°/o SHRUB COVER

 

CHAPTER VII
FIELD EXPERIMENTS

This phase of the study consists of two sets of
experiments, one begun during the summer of 1960 and the
other during the summer of 1961. Both were designed as
tests of certain of the results of Byer (1960).

The earlier set consisted of transplants and seed
plantings of some of the commoner species to all parts of
the soil moisture gradient. The plots were kept free of
natural plant encroaoMent during the course of subsequent
observations. The Species used were some which had showed
relatively strong association patterns, and all but two
were markedly restricted in nature to particular parts of
the gradient. I felt that performance in the absence of
natural interaction should yield clues to the causes of
both macrodistributional Q._e_. along the gradient) and
microdistributional L195. association) patterns. The seed
plantings were designed to demonstrate whether establish-
ment (Lg. germination to maturation) was the same across
the gradient as survivorship of transplanted, mature indi-
viduals.

212

213

The later set was designed to test whether inter-
action could account for the very strong correlations be-
tween certain pairs of species. Because of high transplant
mortality, most experiments of this second set did not
yield the hoped-for results, but survival and growth meas-
surements from it were nonetheless useful in supplementing
data from the earlier plots. Since the 1961 plots were
located in a clearcut strip with or without artificial
shade, it was possible to compare the performance of some
species there and in the 1960 plots, which were located
under the naturally-occurring tree cover.

A. METHODS. The same piece of land used for the
1958 sampling was employed. The dry to wet axis of the
100 meter long gradient was approximately North- South, the
dry and to the North. The 1960 plots occupied the eastern-
most 15 X 100 meter strip, while the 1961 plots were lo-
cated in the westernmost 30 X 100 meters. Both of these
included all of the soil types encountered in 1958.

1960 Plots. For these experiments, the strip was

 

divided lengthwise into ten 10 m X 15 m rectangles, so that
there was a sequence of successively more poorly-drained
blocks starting at the north end. The seven better-drained

rectangles, representing the soil types Grayling through

214

Kinross sands, were located under a relatively Open and
uniform giggg banksiana canoPy. Data for relative light
intensity from Byer (1960) however, show a steady rise
from a mean of 12.6% of full sunlight intensity on.well-
drained Grayling sand through 33.4% on imperfectly-drained
Au Gres sand (at noon on a late August day). This three-
fold increase is no doubt due to an "edge effect",,i.g. to
infiltration of light from the nearly treeless bog to the
south close to Au Gres. The difference is intensified by
the fact that the Open area is to the south and the
slanting rays of the sun, coming from that direction at
this latitude, can penetrate a considerable distance "Up-
gradient". This effect is reduced at midsummer but in-
creased in the early spring (when most growth occurs) and
late fall. 0n Saugatuck sand (about Block 7 from the
north end) the next more poorly-drained soil adjacent to
Au Gres, mean light intensity drops to 9.3% of full sun-
light because of the heavy band of bog hingeline tree
cover found there, but moving "down-gradient" to‘Kinross
(Blocks 7 and 8) it rises to 76.37.. Although the latter
segment contains much of the bogpmargin tree cover, the
Proximity of the Open bog and the similar open character

of the adjacent portions of Kinross itself apparently bring

215
about this relatively high figure. Bog intensity (Blocks
8-10) is 84.5% full sunlight. Thus, incident energy as
well as drainage varied considerably across the gradient,
as it did also locally within blocks due to the shadows
cast by each tree.

The data from replicates from.each rectangle or
block were later treated as a unit. Ten randomly-located
strip plots 15'm.X 30 mm., with long axes perpendicular to
the long axis of the lS‘m x.100'm. strip, were spaded up
in each rectangle, turning the sod under and working it
about with the spade until it could be manipulated easily
with a rake.

In each spaded strip, there was room for 23 planting
locations spaced 1/2 meter from one another. A.strip 15
meters long can.be marked off at 29 intervals of 1/2 m
starting 1/2 m from one end, but in some strips as many
as six such points were covered with tree bases or stumps.
Each experiment was performed by locating one replicate
at a randomly-chosen available position in each strip.
Thus there were 10 replicates of each experiment in each
rectangle, or ten groups of 10 total with each group re-
presenting a different "treatment".

Adults of 14 species and seeds of 9 species were

216
chosen for testing. The individuals to be transplanted
were chosen from.as restricted an area as possible, usually
from what appeared to be a single clone, and as close as
possible to that species' point of maximum cover along the
gradient in order to increase the probability of selecting
the "representative" ecotype. Individuals of as uniform
a size and appearance as possible were afixed with randomly
selected numbered tags, which made it possible to remove
them and plant them in random order. The order in which
the 100 positions in the rows were filled was also
randomized, so that bias due both to selection of specimens
and time of planting were minimized. Seeds, too, were al-
ways selected from as restricted an area as possible near
the maximal cover of the species along the gradient. In
these experiments also the positions were filled in random
order. At each location, a given number of seeds or fruits,
according to species, was chosen "blindly" from the con-
tainer in order to eliminate unconcious bias as to seed
color, size or appearance. Seeds or fruits were hand
planted in "hills" like garden seeds. The area around
each transplant or seed planting was weeded in the spring
of each year during which readings were taken.

Survival of transplants or number of seedlings

217

present were recorded in Blocks 1-7 (numbering starting on
the best drained segments or "dry end") in the spring and
fall of 1961 and 1962 and in the spring of 1963. From the
spring of 1962 onward, it was also noted whether the plants
were flowering or fruiting. These data are presented in
Table X (page 229). Since survival, flowering, and fruiting
data had yielded almost no significant differences across
the whole inorganic soil gradient by the Spring of 1963,
measurements were taken at that time of one or more para-
meters such as height, number of leaves, etc. of each
transplanted species. One species grown from seeds was
included; onicodendron radicans (L.) Kuntze, the only Species
studied not native to the 1958 sample area. It was planted
in conjunction with the research of Mr. William T. Gillis
of the Michigan State university Department of Botany and
Plant Pathology, but was useful also in the present study.
The characters measured for each species are listed and
the results presented in Table XIII, page 242 , and in
Figures 17-21 (pages 243-247 ).

Blocks 8-10, on the organic soil, were read at the
same time and in_the same fashion in the spring of 1961
and 1962. But the spaded peat had become compacted and

settled about six inches below the overall bog surface, so

218
that in the fall of these years the strips on these blocks
were flooded and impossible to read. In any case, it was
evident after the spring of 1962 that all but a few indi-
viduals, all of species indiginous to the bog, had died in
Blocks 8-10; accordingly they were abandoned.

1961 Plots. For the 1961 interaction experiments

 

only the central 10 m. X 100 m, band of a 30 m. X 100 m.
clearcut strip was used for planting, thus acheiving 10-
meterdwide "buffer zones" on either side. This eliminated
all natural shade during the growing season except for
about one hour after sunrise and again before sunset, when
only a small fraction.of total daily incident energy falls
on the earth's surface. The maximum variation in total
insolation over the whole central strip, as checked by a
foot candle meter, was less than 4%, so solar energy was
presumably eliminated as a variable of any importance along
with tree canopy-related differences in soil temperature,
transpiration, rainwater reaching the surface and other
factors.

This central 10 X 100 meter strip was divided into
four segments by soil type according to a soil.map plotted
earlier (Byer 1960), transitional areas between them re-

maining unused. Experiments were located only on G, WC,

219
and AS. The plots for each experiment, 1/2 m. X 1/2 m. in
size, were randomly located within the apprOpriate soil
block. In some experiments, the plots were covered with
saran plastic shade cloth, tacked to a frame above the
plots.8 The fabric used allowed light penetration of about
152 full sunlight, approximating the means fbund for the
Jack pine forest in 1958 (Byer 1960). Each plot served as
one replicate of an experiment.

These plots were not spaded, for it was desired to
leave the soil as undisturbed as possible. Rather, the
edges were cut down to a depth of about 25 cm. with a flat
shovel, and then all visable plant parts carefully removed
by hand in order to eliminate as much outside competition
as possible while keeping the duff and humus layers fairly
intact.

In planting, the model used by Sakai (1955) was
utilized. The species whose perfbrmance was to be tested
was located in the center of a plot, and it was surrounded
in a hexagonal pattern by (A) six individuals of another

species, (B) six of the same species, or (C) by nothing

8Obtained from Lumite Division, ChicOpee Manufactur-
ing Corp., Cornelia, Ga. Fabric no. 1817-002.

220
(10 replicates of each treatment, 30 in all per experiment).
(A) showed the reaction of the central plant to the prox-
imity of a suspected interactant, as compared to the
control (C). The (B) plots were designed to estimate how
much of a reaction of the center plant might be attribut-
able to intraespecific interaction, as compared to the
interspecific interaction tested by the (A) plots. Sur-
rounding plants were located 10 centimeters from the
central individual and from each other. As in the case of
the 1960 plots, plants were selected for uniformity of ap-
pearance and location, and the orders of planting and of
filling spaces were randomized.

Survival was recorded for these plots in the fall
of 1962 and the spring of 1963 (Table XIV, page 253). In
the spring of 1963, the morphological features which were
measured on the 1960 transplants were also measured on
survivors in the interaction plots (Table XVI, page 259).
The two species not included in the 1960 plots, Cgmptonia
peregrigg and Epigaea r ens, did not survive in sufficient
quantities to warrant such measurements.

Both survival readings and morphological parameter
measurements were taken at times when mortality was not

likely to be occurring, and after rapid spring growth had

221
ceased. Spot checks indicated that there was neither ap-
preciable death nor appreciable growth during the several
days required for the readings.

Analysis of Data. At the outset let us clarify a

 

feature of the tables which could lead to confusion. If a
transplanted individual in the 1960 plots gave evidence of
life at any reading, and was definitely established as the
specimen originally planted at that spot, it was counted
as living for the reading in which it demonstrated life
3nd for all previous readings (Table X, page 229). Indi-
viduals which were recorded as apparently dead in the
spring following tranSplant (of which there were many) or
even later (relatively few) sometimes gave rise to new
shoots. Thus, this retroactive revision increased the
accuracy of survival data, although undoubtedly a few
short-term survivors escaped detection. The survival data
for the 1961 plots were not summarized in this fashion,
however, because I felt that the relatively short time
elapsed did not permit the drawing of meaningful survival
curves. Thus the figures in Table XIV (page 253) include
only the number of individuals which appeared living at

the time of the reading, not those demostrating life in any

subsequent reading.

222

For the Spring 1961 survival data from the 1960
plots, the data for Blocks 1 and 2 (dry end, Grayling
sand) were treated as a unit, as were those for Blocks 6
and 7 (bog margin, Au Gres and Saugatuck sands) and Blocks
9 and 10 (bog, Dawson-Greenwood peat). Significance of
difference in per cent survival between each pair of these
units was determined by means of a chi-square test. For
the seed plantings, differences between mean numbers of
seedlings of each species per plot on Blocks 1-2 versus
6-7 were tested by calculating t-values. But here, fall
1961 data were used for all but Melampyrum lineggg @131);
the latter's spring 1961 data were used because this spe-
cies is a winter germinating annual with a normally high
mortality rate through the growing season (Cantlon and
Curtis 1964). Since the seeds of many other species ger-
minated during the 1961 summer and subsequent seedling
mortality appeared relatively low, I felt that the fall
readings would be more informative. Blocks 9-10 were not
included in the germination calculations because here, in
the bog, none of the planted seeds of any species germi-
nated.

The significance tests were not repeated on data

from later readings since there was little mortality any-

223
where on Blocks 1-7 (inorganic soil) after initial trans-
plant loss. The number of seedlings did change appreciably
from reading to reading but, with the exception of 125122?
dendron, their prOportions on different gradient segments
apparently did not. Results of significance tests appear
in Table XI (page 234).

For the measurements of morphological characters
taken in the spring of 1963, the following comparisons
were made by t-test: (l) for the 1960 plots, a comparison
of Blocks 1-2 with Blocks 6-7, (2) a similar comparison
between Blocks 1-2 and Block 7 alone (1 and 2, right-hand
columns in Table XIII, page 242), (3) for the 1961 plots,
comparisons between solitary central plants and those sur-
rounded by individuals of the same or another Species, for
the only four experiments where survival was high enough
to warrant it (Table XVII, page 264), (4) where the same
interaction experiment was performed on two different soil
types, differences between all plants (central and sur-
rounders combined) on the two segnents (Table XVI, page 259)

and (5) differences between sun-grown and shade-
BrOWn plants (regardless of other treatment or position)
of the one Species (Vaccinium,ggggggifgligm'var.1215222)

which was grown under both conditions, (Table XV, page 257).

224

In number (4) above, the comparisons should be
valid because both soil types contain about the same pro-
portions of individuals planted during any particular time
interval. But the entire sets of sun-grown and shade-
grown blueberries, number (5), were planted at different
times, and while the former contains some central plants
the latter does not. But all blueberries were planted
within the same 4-day period about a month before first
frost, and were thoroughly watered immediately after plant-
ing, then every three or four days (except when soaking
rain.made it unnecessary) for the next two weeks. Thus
differences in dessication or recovery time before frost
were not very great, and should have affected the results
but little.

B. W
Data summaries for the different readings are given in
Table X (page 229). In addition, the final (spring 1963)
survival readings are graphed (solid lines) in Figures 17-21

pages 244-248 ). Tests of significance of differences

in survival and germination on different segments, based
upon the 1961 data are given in Table XI (page 234 ).
(Since the transplant survival data after 1961 show no more

than what could easily be accidental changes, they are not

225
included here).

We see from Figures 17-21 that on the whole, trans-
plant survival is not markedly different across the por-
tion of the gradient on mineral soil, Blocks 1-7. The
tests for significance (Table XI, page 234) show that only
one of 14 species, vaccinium,gggustifolium'var.,gigggg
(gig), has significantly different survival rates (P <:.05)
in Blocks 1-2 and 6-7. Out of 14 sets of data one would
show differences significant at this level more often than
not, even though the fluctuations were merely random, thus
13 anggstifolium var. nigrum's case is most probably a
fluke. In fact, subsequent readings (Table X) revealed
that most of the plants of this species apparently dead on
Block 6-7 in 1961 were, in fact, alive, and the most re-
cent reading shows equal numbers of living plants of this
species on both blocks! Moreover, mortality since the
initial transplant loss in Blocks 1-7 has been.minimal.

So it appears that most of the species tested, whose cover
Optima ranged from G to DC, have fairly uniform survival
potential as adults in cultivated plots all the way across
the non-organic soils of the soil moisture gradient. A
comparison.between either of the aforementioned blocks and

Block 9-10, however, reveals much higher survival in the

226
"dry end" and "bog margin" blocks, most of the differences
being significant at high levels (for 20 out of 28 P‘<:.001).
This merely confirms the obviously less favorable conditions
for survival in the standing water in the spaded rows of
the bog segments.

The results certainly do not mirror distributions
observed in.the 1958 study. Two ygccinium taxa, which in
nature are ubiquitous with high frequencies across the
whole range of inorganic soil types and occur sporadically
in the bog, reflect this in their transplant survival
curves @513, m, Figures 21 A, 21 a, page 248).. But 1.
myrtilloides flag), in this study area a Species of the
bog margin exclusively, has a similar curve (Figure 20 C,
page 247)., Though it was not found at all in the bog
proper in the 1958 study, it has a higher survival rate
here than either of its gradient-spanning genus-fellows,
and is, in fact, one of the few species showing no signif-
icant differences between Blocks 1-2 or 6-7 and Blocks 9-
10. Chamaedaphne calyculata (93mg, a species naturally
restricted to the bog, has a high survival rate there, but
also survives at a rate not significantly different clear
up to the excessively well drained Grayling sand. Another

exclusively bog and bog margin species, Ledum groenlandicum

227
(Le_d) has high survival on the inorganic soil but extremely
_l_._o__w survival in the bog, just the Opposite of its natural
distribution pattern. This low bog survival, suggesting
flooding intolerance, may be reflected in Ledum's natural
tendency to be concentrated on higher hummocks than is
Chamaedaphne. Yet some of the most xerOphytic species
such as Danthonia spicata @_a_n) and £91335 pensylvanica
at) have higher bog survival rates over the first fall,
winter, and early Spring than does Ledum. In general, in
fact, the xerOphytes were more tolerant of these extreme
conditions than some slightly more mesophytic Species

(Pteridium aguilinum (gtr), Caultheria grocumbens (Gau)),
while three of four bog margin Species (CgptiS groenlan-

 

gig; (Q3), §9_r__n_u_s_ ganadensis @595) and 33393 higpidus var.

obovalis GM: the exception being Vaccinium mntélloides

discussed above) appeared to be the least tolerant of all.
Mere short-term tolerance ranges of mature plants

to soil drainage, available nutrients, or other gradient-

related environmental variables then, do not seem to be

the primary factors reSponsible for the observed natural

distribution of these Species along the gradient. It could

be that the block of native soil approximately 10 cm X 10 cm

inserted with each tranSplanted individual has kept it a-

228

live. Yet there has been only token mortality since the
first'winter, and there should have been a second wave of
mortality signaling exhaustion of this soil if this had
been the case. Possibly such exhaustion had not occurred
by the last reading. Probably, though, the Species tested
are prevented in nature from expanding up and down the gra-
dient to the limits of their tolerance to the physical en-
vironment by various interactions with other Species. The
relative efficiency of Species in occupying available
niches surely changes along the gradient in reSponse to
physical factors, but such changes seem not to Show up
strongly under the relatively Stress-free conditions of
the cultivated strips. Failure of ecesis on certain soils,
discussed in the following section, may be another natural
distributionrlimiting factor.

Flowering and fruiting (Figures in parentheses,
Table X, page 229) were noWhere common enough for the
testing of significance between segments. In most cases,
in fact, they revealed or suggested nothing of the reasons
for the natural distributions across the gradient which
were found, for on the basis of limited data, margin-re-
stricted and even bog species seemed to flower as well on

the best-drained segment as elsewhere, while Species na-

229

TABLE X

1960 plots; survival, flowering and fruiting. The non-parenthetical
numbers under columns 8 61 (Spring 1961), F 62 (fall 1962), etc. ,
indicate the numbers of survivors (out of 10 planted) in the
correSponding blocks at the times indicated; the parenthetical
numbers show the number flowering (first number) and fruiting
(second number) at these times. The columns headed "M" Show
mortality between the two periods surrounding them (tplt. M =
transplant mortality).

tplt. s F s F 5
Block _114 €3_1 M. 62 M §_z M 03

I...
IZ

 

Carex pensylvanica

 

 

 

 

 

 

 

 

 

 

l 0 10 0 10 1 9(0, 2) O 9 0 9(0, 3)
2 O 10 O 10 0 10(0, 0) O 10 0 10(0, 1)
3 0 10 O 10 0 10(0, 0) 2 8 0 8(0, 0)
4 0 10 1 9 1 8(0, 0) 1 7 0 7(0, 1)
5 0 10 O 10 0 10(0, 0) 1 9 0 9(0, 0)
6 0 10 O 10 0 10(0, 0) 1 9 0 9(0, 1)
7 0 10 O 10 0 10(0, 0) O 10 0 10(0, 2)
1-7w 0 7O 1 69 2 67(0, 2) 5 62 O 62(0,8)
8 8 2 1 1
9 10 0 O O
10 10 0 O 0
8-10 28 2 1 1(0, 0)
Chamaedaphne calyculata
l 1 9 1 8 2 6 0 6 0 6
2 3 7 0 7 O 7 0 7 0 7
3 1 9 2 7 0 7 0 7 1 6
4 2 8 O 8 O 8 O 8 0 8
5 0 10 0 10 0 10 1 9 0 9
6 l 9 l 8 O 8 1 7 O 7
7 0 10 0 10 0 10 O 10 0 10
1-7 8 62 4 58 2 56 2 54 1 53
8 2 8 4 4
9 4 6 5 1
10 2 8 6 2
8-10 8 22 15 7

 

 

 

 

230

Table X, continued
5

 

 

 

 

 

 

 

 

 

 

 

 

 

tplt. S F F 5
Block M 1 1 M 62 M 62 M (2
Co tis groenlandiga
1 0 10 0 10 0 10(O.1) 1 9(0,1) O 9(0,Z)
2 0 10 O 10 O 10(0,0) O 10(0,0) 1 9(O.2)
3 0 10 O 10 1 9(0,0) O 9(0.0) O 9(0,4)
4 0 10 0 10 1 9(0,0) 0 9(0,0) O 9(O,2)
5 0 10 0 10 0 10(0. 0) O 10 (O, O) 0 10(0, 0)
6 0 10 O 10 0 10(0.0) O 10(0,0) 0 10(0.0)
7 0 10 O 10 O 10(0,0) O 10(0,0) 0 10(0,1)
1-7 0 7O 0 7O 2 18(0, 1) 1 67(0, 1) 1 38(0.11)
8 9 1 0 1
9 10 0 0 O
10 9 1 1 0 _
8—10 28 2 1 1(0, 0) F
Cornus canadensis
1 2 8 1 7 1 6 0 6 1 5(0. 0)
2 2 8 1 7 2 5 0 5 0 5(0, 0)
3 1 9 l 8 1 7 1 6 1 5(0, 0)
4 3 7 1 6 1 5 2 3 0 3(0, 0)
5 2 8 2 6 0 6 2. 4 0 4(1, 0)
6 3 7 2 5 O 5 0 5 1 4(1, 0)
7 2 8 0 8 1 7 1 6 0 6(0, 0)
1-7 15 55 8 47 6 41(0,0) 6 35(0,0) 3 32(2,0)
8 9 1 1 0
9 10 O 0 0
10 10 O O 0
8—10 29 1 0 0(0, 0)
Danthonia spicata
1 0 10 0 10 O 10(9,0) 0 10(0, 8) 0 10(0,0)
2 0 10 0 10 0 10(9, 0) 0 10(0, 10) 1 9(1, 0)
3 O 10 O 10 1 9(5,0) I 8(0,7) 1 7(0,0)
4 0 10 0 10 l 9(5,0) 0 9(0,9) O 9(0,0)
5 0 10 O 10 0 10(9.0) 0 10(0, 10) O 10(0,0)
6 0 10 0 10 0 10(8, 0) 0 10(0.9) 0 10(0,0)
7 1 9 0 9 l 8(8,L 0 8(0, 6) 0 8(2, 0)
1—7 1 69 0 69 3 66(53, 0) 1 65(0, 59) 2 63(3, 0)
8 5 5 4 1
9 7 3 3 0
10 6 4 4 0
8—10 18 12 11 1(0, 0)

 

 

 

 

231

Table X. continued
S

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

tplt. S F F 5
Block M _1 M 6_1 M 62 M 62 M 63
Gaultheria procumbens"<
1 0 10 0 10 1 9(2. 0) 1 8(0,1) 0 8
2 0 10 0 10 1 9(1, 0) 1 8(0, 0) O 8
3 0 10 0 10 0 10(0, 0) 2 8(0, 0) 3 5
4 0 10 0 10 1 9(0,0) 1 8(0,0) 0 8
5 O 10 0 10 0 10(0, 0) 0 10(0, 0) 3 7
6 0 10 0 10 1 9(0,0) 0 9(0,0) O 9
7 0 10 l 9 0 9(0,0) 0 9(0,0) 1 8
1-7 0 70 1 69 4 65(3,0) 5 60(0, 1) 7 53(0,0)
8 9 1
9 9 1
10 10 0
8—10 28 2(0, 0)
Ledum groenlandicum
1 1 9 O 9 0 9(1, 0) 0 9(0, 1) 1 8
2 1 9 0 9 Z 7(0,0) 1 ’)(0,0) 0 6
3 1 9 0 9 1 8(0, 0) 0 8(0, 0) 1 7
4 O 10 O 10 0 10(0,0) 0 10(0,0) O 10
5 4 6 O t) 0 6(0, 0) 0 6(0, 0) O 6
6 3 7 0 7 1 6 (O. O) 1 5 (O, 0) O 5
7 1 9 1 8 0 8(0, 0) 0 8(0, 0) O 8
1—7 11 59 1 58 4 54(1.0) 2 52(0. 1) 2 50(00)
8 10 O 0 0
9 9 1 1 0
10 9 1 1 0
8-10 28 2 2 0(0, 0)
Maianthemum canadense
1 0 10 1 9 2 7 0 7 0 7(0, 0)
2 0 10 1 9 1 8 0 8 1 7(0» 0)
3 1 9 0 9 0 9 0 9 0 9(0, 0)
4 0 10 0 10 0 10 0 10 0 10(1, 0)
5 o 10 1 9 o 9 o 9 0 9(2, 0)
6 o 10 o 10 1 9 o 9 1 8(0, 0)
7 1 9 0 9 1 8 0 8 0 8(0, 0
1-7 2 68 3 65 5 60(0,0) 0 60(0,0) 2 58(3,0)
8 5 5 5 0
9 7 3 3 0
10 9 1 1 0
8-10 21 9 9 0(0. 0)

*First reading on segments 8-10 in the

Spring of 1962.

 

232

Table X, continued

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

tplt. s, F s F 5
Block M __1 M 61 M 62 M 62 M 63
Oryz0psis asperifolia
1 l 0 1o 0 1o 0 10(0, 0) o 10 0 10(0, 0)
2 0 1o 0 1o 0 10(0, 1) o 10 0 10(0, 2)
3 o 10 o 10 0 10(0, 0) 0 1o 0 10(0, 1)
4 o 10 o 10 0 10(0, 0) 0 1o 0 10(0, 1)
5 o 10 0 1o 0 10(0,0) 0 10 o 10(0,0)
6 o 10 0 1o 1 9 (o, 0) o 9 0 9 (o, 3)
7 o 10 o 10 1 9 (o, 0) o 9 1 8 (o, o
1-7 0 7o 0 7o 2 68(0, 1) o 68(0.0) 1 67(o,7)
8 8 2 2 o
9 10 o o 0
10 1o 0 o o
8--10 28 2 2 010,0)
Pteridium aquilin_u_r_rl
1 1 9 0 9 4 5 o 5 o 5
2 0 10 o 10 0 1o 3 7 4 3
3 o 10 o 10 2 8 o 8 1 7
4 o 10 0 1o 4 6 o 6 2 4
5 1 9 o 9 1 8 o 8 1 7
6 3 7 o 7 1 6 o 6 4 2
L 1 9 o 9 2 7 o 7 1 6
1-7 6 64 o 64 14 50 3 47 13 34
8 10 o o o
9 10 o o o
1_()_ 10 o o 0
8-10 10 o o 0
Rubus hisPidus var. obov.a.l_i§
1 o 10 o 10 o 10 1 9 0 9(1. 0)
2 2 8 o 8 o 8 1 7 1 6(2, 0)
3 0 10 o 10 o 10 0 1o 0 10(3, 0)
4 o 10 o 10 1 9 0 9 0 9(2) 0)
5 1 9 0 9 o 9 o 9 0 9(2, 0)
6 o 10 o 10 1 9 o 9 0 9(4, 0)
L 0 10 o 10 o 10 o 10 1 9(6, 0)
1-7 3 67 o 67 2 65(0, 0) 2 63(0, 0) 2 61(20, 0)
8 8 2 1 1
9 1o 0 o 0
_10_ 10 o o 0
8-10 28 2 1 1(0. 0)

233

Table X, continued
tplt. S F S F S
Block M 61 M 61 M 62 M 62 M 63

Vaccinigrg Eyrtpiggifiae-S‘

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 2 8 O 8 O 8 1 7 4 O 7 (0, 0)
2 O 10 O 10 0 10 0 IO 0 10(0. 0)
3 1 9 O 9 l 8 O 8 0 8(0, 0)
4 0 10 0 10 O 10 1 9 1 8(0, 2)
5 1 9 0 9 I 8 0 8 0 8(0, 0)
6 l 9 O 9 O 9 O 9 0 9(0, 0)
7 1 9 O 9 1 8 O 8 0 8(0. 1)
1-7 6 64 0 64 3 61(0,0) 2 59(0 0) 1 58(03)
8 6 4 4 O I
9 5 5 5 O
10 7 3 3 0 j
8-10 18 12 12 0(0, 0) l
Vaccinium angustifolium var. nigrum
l 2 8 0 8 0 8 0 8 0 8(0, 0)
2 O 10 0 10 0 10 1 9 0 9(0, 0)
3 2 8 0 8 0 8 0 8 0 8(0, 0)
4 0 10 0 10 1 9 0 9 0 9(0, 0)
5 0 10 0 10 0 10 O 10 0 10(0, 1)
6 0 10 1 9 0 9 0 9 0 9(0, 1)
7 O 10 l 9 0 9 1 8 0 8(0, 0)
1—7 4 66 2 64 1 63(0, 0) 2 61(0, 0) 0 61(0, 2)
8 9 1 l 0
9 7 3 3 0
10 9 l l 0
8—10 25 5 5 0(0, 0)
Vaccinium boreale
1 0 10 0 10 l 9 l 8 2 6
2 0 10 0 10 0 10 0 10 1 9
3 0 10 O 10 l 9 l 8 2 6
4 2 8 0 8 0 8 0 8 l 7
5 0 10 1 9 0 9 0 9 0 9
6 2 8 l 7 0 7 0 7 1 6
7 0 10 0 10 O 10 2 8 0 8
1-7 4 66 2 64 2 62 (0, 0) 4 58(0, 0) 7 51(0, 0)
8 6 4 4 0
9 5 5 5 0
10 9 1 1 0
8-10 20 10 10 0(0, 0)

 

 

 

 

234

TABLE XI

1960 transplant plots. Significance of differences
between gradient segments, 1961 data. A dash (—)
indicates absence of significance (P>0.05).

A. CHI~SQUARE TESTS OF DIFFERENCES BETWEEN BLOCKS
1-2 (DRY ENE, 6—7 (BOG MARGIN) AND 9—10 (13007, SUR-

 

 

 

VIVAL.
S ecies* Segments compared
1-2 vs 6-7 1-2 v5 9-10 6-7 vs 9-10

chi sq. E; chi sq. P< chi sq. P<
CAR 0.00 —- 36.10 0.001 36.10 0.001
CHM 2. 29 - 0. 00 -- 1.40 -
COP 1.44.— — 22.72 0.001 32 48 0.001
COR 0.00 -- 16.41 0.001 18.57 0.001
DAN 0.00 -— 16.41 0.001 13.29 0.001
GAU 0.91 — 0.52 — 0.91 _
LED 0.00 — 10.66 0.005 12.60 0.001
MAI 0.00 — 14.43 0.001 12.10 0.001
ORS 0.00 — 36.10 0.001 36.10 0.001
PTR 0.00 -- 10.80 0.001 6.23 0.02
RHO 1.20 — 16.41 0.001 26.18 0.001
VAM 0.98 — 0. 40 — 3.68 -
VIG 3.90 0.05 6.54 0.02 20.05 0.001
VYP 0.01 —- 8.18 0.005 8.18 0.005

B. Tm TESTS OF DIFFERENCES IN NUMBERS OF SEEDLINGS
BETWEEN BLOCKS 1—2 AND 6-7.

Species" t ES__ Species)" t I:_<~
CHM 2.000 — TOX 2. 154 —
COR 3.662 0.001 VAM 0.137 .—
DAN 2.599 0. 01 VIC 1.796 —.
LED 0.000 —— VYP 0.000 .
MEL 0.271 ——

’1‘ Index to species abbreviation symbols in Appendix B.

235
tive to that segment showed no decrease in flowering on
the margin. In only one or two cases, mentioned later
where appropriate, were these data useful in interpreting
the behavior of a species in nature.

C. 1960 PLOTS: SEED GERMINATION AND SEEDLING
SURVIVAL. The results presented in Table XII (page 234)
indicate that there was no apparent germination of any
species in Blocks 8 through 10, in the bog. This is un-
derstandable, for the periodically flooded rows were prob-
ably not conclusive either to germination on seedling
survival, even for normal bog species. These blocks were
therefore of no use in analyzing natural distributions of
Species.

With the two bog species,‘nggm groendandicum (Egg)
and Chamaedgphne calyculagg QM) , results do suggest that
inability to germinate everywhere may impose a major re-
striction upon their gradient distribution. ‘nggg pos-
sibly produced seedlings (not confirmed) in strips 68 and
70 of block 7 only, while Chamaedaphne seedlings appeared
in strips 69 and 70. These locations are on the Saugatuck
soil type, very close to the bog hingeline. One of the
locations is mossy (Polytrichum) and in all four moisture

can be squeezed from the humus with the fingers during much

236
of the growing season. Thus all of them approximate the
conditions upon which the seeds prObably germinate in the
natural course of events, on moist bog surface, probably
on §phag§um. The difference of’Chamaedgphneig seedling
number on Block 6-7 from total absence on Block l-2 was
not quite significant at the 5% level (Table XI, bottom),
but this species continued to produce new seedlings in the
bog margin locations every year up to the last reading
(Table XII). The possible‘nggg seedlings, on the other
hand, did not appear until 1963.

In one other species, Qgrggg canadensis (92;),
special requirements for germination are by contrast evi-
dently not reSponsible for the species' limited distribu-
tion. In fact in 1961, six seedlings were present in the
most xeric Blocks 1-2, far removed from the species' nor-
mal habitat, and none in Blocks 6-7, where lush‘gggggg is
ubiquitous between the planted strips. This difference
was significant at P1<:.001, but since the time of this
reading, as we see from Table XII, seedlings have appeared
all over the inorganic gradient. It is questionable wheth-
er'Qggggg can really germinate more quickly on dry, sandy
soil, as suggested by the earlier results, but at the least

it is not prevented from germinating at all there. The

237

other bog margin-restricted species tested, Vaccinium

 

mmilloides (1813;), also continued to germinate across the
whole gradient.

By the fall of 1961 fewer seeds of Danthonia
gpicata Q)__an_) had germinated in Blocks 1-2, on this spe-
cies' native soil, than in much more poorly-drained Blocks
6-7, near its range margin (difference significant at
P <.01, Table XI). This might suggest that Danthonia
germinates best on fairly moist soil in the absence of
competition from the lush vegetation of AS, but a glance
at Table XII shows that in the spring of the same year the
situation was reversed, and at the time of the last reading
the differences had leveled off. With the other five taxa
tested, germination was fairly uniform throughout Blocks
1-7.

With the three Vaccinium taxa (Kain, y_i_g and 3112) ,
seedling numbers seemed to reach a peak in the fall of
1961 and then decrease to only a few scattered individuals,
but with no appreciable differences between segments, which
in the case of the broad ranging 1. gngmtifoliym var.
21:51-12. and 1. boreale is not surprising. But Tgxicodem
radicans may whose seeds continued to germinate through-

out the period of the readings, shows increasingly lower

238

TABLE XII

1960 plots. Numbers of living seedlings present at each
reading. "561" indicates Spring reading, 1961, "F62"
indicates fall reading, 1962, etc.

S F S F S S F S F S S F S F S
61 61 62 62 63 61 61 62 62 63 61 61 62 62 63

 

 

 

 

 

 

 

 

 

 

Chamaedaphne Cornus Danthonia

Block calyculata . canadensis spicata
1 0 0 0 0 0 0 3 1 0 0 24 31 15 16 15
2 0 0 0 0 O 0 4 4 3 5 25 33 14 17 18
3 0 0 0 0 0 0 1 0 0 1 10 16 12 13 16
4 0 0 0 0 0 0 1 0 0 0 24 24 12 31 23
5 0 0 0 0 O 0 3 1 2 2 26 23 11 13 12
6 0 0 0 0 0 0 O 0 O 0 38 21 9 13 12
7 0 4 14 13 20 0 O 0 O 1 22 27 12 22 18
tot 0 4 12? 13 20 0 12 8 5 9 169 175 85 125 114

Ledum Melamgyrum Toxicodendron
agenlandicum lineare radicans

l 0 0 O 0 0 6 -— - ~— -— 13 12 13 9 7
2 0 O 0 0 0 ll -— - - -- 9 5 10 9 7
3 0 0 O 0 0 6 -— —- -— - 4 7 10 9 9
4 0 0 0 0 0 12 -- — -— -- 10 9 12 9 7
5 0 0 O 0 0 6 —— - ~ -— 7 4 11 9 8
6 0 0 0 0 0 6 - — - —— 9 5 5 3 3
7 0 0 0 0 6 10 - — — -—- 7 8 7 7 3
tot 0 0 0 0 6 57 — — -— —- 59 50 68 55 44

Vaccinium Vaccinium XacciniMLn

myrtilloides anggstifolium boreale

var. r_1_igrum

1 0 2 1 O 0 7 6 2 0 0 0 8 0 0 0
2 2 19 6 0 l 3 6 1 0 Z 1 4 1 1 1
3 4 4 1 0 0 3 6 1 l l 4 4 0 0 3
4 2 8 5 0 14+ 7 18 4 1 4 1 3 2 0 1
5 3 6 1 0 10+ 3 5 1 1 l 5 3 0 1 0
6 5 8 2 o o 1 13 2 o o 4 7 o o 3
__7 5 12 0 o 3 1 6 1 o 2 2 5 o 3 4
tot 21 59 16 0 28+ 25 60 12 3 10 17 34 3 5 12

 

 

239
figures on the more poorly drained Blocks 6-7 from the
first reading on, until in the final reading fourteen
plants on l-2 contrast with only six on 6-7. The random
probability of a difference in the mean numbers per plot
as great is slightly below 0.05. A check of the raw data
shows that both delayed germination and survival were lower
on the latter block. There is nothing in ggxicodendggglg
native habitats to suggest why, fer in those fewggiggg
banksiana stands where it is present this Species occurs
.ggly on the imperfectly to poorly-drained soils.

I. D. 1960 PLOTS: MORPHOLOGICAL MEASUREMENTS. The
results are presented in Table XIII (page 242) and Figures
17-21, (pages 244-248).

The number of significant differences between the
means of the characters measured on Block 1-2 and Block 6-
7 significant at the P.<:.OS level or higher is only about
what one would expect of randomly fluctuating variables.
However, if we include the P <:.l level, about 212 of the
differences have a random.prdbability lower than this,
about twice the proportion which we would encounter if in
fact no real differences existed. Such measurements may,
therefore, be somewhat more valuable indicators of‘diffenr

tial vigor, in cultivated plots such as these, than sur-

240
vival alone.

All of the criteria used were intended to reflect
rapidity of growth and hence size of the photosynthetic
surface, from which we may in some measure infer relative
suitability of different physical environments along the
gradient for a species. In areas where such measurements
yield relatively low values in the cultivated plots, one
might suspect the species of lower competitive ability in
the adjacent natural habitat than in areas where the values
are high. But it cannot be assumed, nor do the results
indicate, that vigor as estimated from these measured val-
ues will predict natural distributions across the gradient.
Ability to germinate, distribution of disseminating agents
and other factors will enter the picture. In addition,
relative competitive ability in an area itself depends not
only upon reactions to physical factors, but also upon in-
teractions with soil micrdbiota and other naturally-oc-
curring species, some of which may enhance growth.

Among those differences which are significant,
some do nonetheless reflect the natural gradient distribu-
tion of the species, suggesting reaction of adults to
abiotic factors such as soil moisture or organic matter as

at least major influences in this distribution. For in-

241

stance, mean clump size of 933.335 pensylvanica ((335) is
smaller on the bog margin where cover of this species nat-
urally begins to diminish rapidly (Figure 17 A, page 244).
Similarly, the largest shoots of 9231193 ganadensis (92;)
reach maximum size on Blocks 6 and 7 where this Species
"peaks" (Figure 18 A, page 245). The bog margin blueberry,
Vaccinium myrtilloideg Gag) also produces its longest
shoots in the same area (Figure 20 C, page 247),0n the
other hand, the bog margin M grtoenlfiagica C ac-
tually produces broader (though slightly fewer) leaves on
the "dry end", so that with this Species drought-related
'low vigor would not seem to be a cause of its exclusion
from Grayling sand (Figure 19 B, page 245).

There is a general trend, evident from Figures 17-
21 for values of all morphological characters to be lower
on Blocks 2 and 3 than on the adjacent blocks on either
81113. The cause of this may be relatively heavy tree
shade on these segments supressing growth, although shade
actually causes greater shoot elongation and leaf eXpansion
of most plants.

E. INTERACTION PLOTS: SURVIVAL.AND GROWTH ygggy

UREMENTS. These results (Tables XIV for survival, page

 

253, and XVI for growth measurements, page 259 ) reveal few

1960 plots.

and bog margin, t-test.

Species

Carex
pensylvanica

 

 

Chamaedaphne
9.3131991. 3.1.21.

CoEtis

groenlandica

 

Corn_u_s
canadensis

Danthonia

Spicata

Gaultherla

procumb;

*A dash (-—) indicates P>O. 1.

242

TABLE XIII

unless otherwise indicated.

Character

____ ._~_ .—

 

bdth. clump

no. of branches

longest branch

avg. Of all
branches

no. of leaves
bdth. largest If.

no. shoots
bdth.broadest
shoot

bdth. clump

no. of shoots

hgt. lngst. sht.

no of leaves,
longest shoot

79.

8.

114

77

12.
33.

55.

85

34.

5

000

57.

13.
26.

\O

0‘
0‘5

72.

.12.

39.

.s_-_7.

13.
28.

513

33.

51>

5%

Differences in growth measurements between dry end
Linear measurements are in millimeters

Significance

 

1-2 1~Z

vs 7 vs 61-?

P< * P<
0.1 0.1
O 1 0.1
0.1 0.02

243

Table X111 cont.

5218811124823
Mean on blocks 1-2 1—2

Mm”.— —-—— ~~-.——_—n

 

 

 

 

 

Species Character vs 7 ' vs 6~7
“*7 ‘ “ 1-2 7 6.7 P< "‘ Pg *
Ledum
gLoenlandicum no. branches 5. 6 7. 1 6. 2 —- —-
lngst. branches 69. 1 63.0 64.5 — —
avg. of all
branches 43. l 45. 4 45. l — -
Maianthem‘gLn
canadensis no. shoots 9 8 6 8 0.05 0 1

3. . .
bdth. lngst. leaf 49. 4 39.1 43.5 0.1 ——

 

 

OryonSis
asperifolig no. of leaves 31. 1 27.3 22. 9 — —
Pteridium
aguilinum no. of fronds l. 6 1.0 l. 4 - -
lgth. lngst.
frond 272 294 358 - -
bdth. " " 242 219 258 — —
Rubus hispid_u_§
var. obovalis no. of leaves 36. 3 44. 0 37. 8 -- -
Toxicodendron hgt. plant 68. 8 41.0 38. l -— -—
radicans no. of leaves 3. 1 2.0 2. 1 - —
Vaccinium no. of branches l9. 1 21. 9 18.2 — -—
myrtilloides lngst. branch 58. 7 87. 1 74.0 0. 1 --
W 422. no. of branches 28.5 34.1 31.8 _-
var. nigrum lngst. branch 57.6 85. 9 94. 4 0- 1 O. 01
meg; no. of branches 24. 9 23. 6 23.6 ._ ..
8322313 lngst. branch 92. 4 102.0 92.3 _ _

*A dash (—-) indicates P>O. l.

_—

 

FIGURE 17. Survival and Growth, 1960 Plots

Code as indicated above the vertical
axes. Block numbers along the horizontal
axes.

A. Carex pensylvanica
B. Chamaedaphne cglygulata

C . Cgtis groenlandica

 

244

 

 

 

 

 

 

 

CLUMP 9;,
savanna. suam
100- 1008
so~ 809
so« so-
40- 4o-
zo- 20-
J A
O 0 I v I I I I “I
1 2 3 4 a o 7
CAREX PENSYLVANICA
ave BRANCH L6TH.LONGEST no 96
LGTH. ma. BRANCHJIM. snmcues sunv.
15 - 1501 35- 100-
120« 120‘ 28- so-
so- 90‘ 21~ so-
so- so- 14« 4o-
30- 30- 7- 4o«
OJ 04 OJ 0

 

 

 

 

 

 

 

 

 

 

 

£6715: iv'n'o'é'éi' u .
LEAF, Mm LEAVES SURV.
40] 254 MW /__—.___
I: q q '.‘. '.'. A
32 20 80 -. ’.’ .V .........
I \\
I, \
24" '5‘ 601 \ I \\ ’
\\ ’I y”'
\\ ”F~--~/
Is- 10- 40- ‘v’
3- 5 - 20 -
1 1 C
0 V T V I Y I ;
a O I 2 3 4 5 6 7
COPTIS GROENLANDICA

 

01744-8

FIGURE 18. Survival and Growth, 1960 Plots, cont.

Code as indicated above the vertical
axes. Block numbers along the horizontal

axes .
A. Cornus canadensis
B . Danthonia gpicata

C . Gaultheria procumbens

 

BOTH. LONG EST
SHOOT, MM.

80«
45-

30-

 

NO. LEAVES HGT. TALLEST
“REST SHOOT SHOOTJIM.
6- 35.

4‘ 28‘

3. 2| -

2" "I

l" 7-

0-I o .

 

 

 

80'-

60'

4o-

 

245

 

96
sunm
Icon

80-
so-
40-

20‘

 

(30 RhlU S

%
suau
loo-

eo~
so
40-

20-

 

0

.L..

 

I r I a

2 i 4 s s 7
C A N A DIEN SI 5

 

 

D A N T'H O N I1A

405 "

 

 

BO-

60-

4o-

 

5 i 4'
SPICATA

mu
a.
q

 

 

G A U L 1'14 E R l

(I:
0‘1
NJ

I I
3 4

N.

A PROCUMBENS

FIGURE 19. Survival and Growth, 1960 Plots, cont.

Code as indicated above the vertical
axes. Block numbers along the horizontal
axes.

A. Ledum groenlandicum

B. Maianthemum canadense

C . Oryzgp sis ageri folia

246

 

 

 

 

 

 

 

1

 

 

 

 

 

 

 

 

 

 

AVOJRANCH LOT". LONGEST NO. NEW %
LGTK, MI. BRANCH, HM. BRANCHES SURV.
75' 76" 7.5- loo-1
‘0. 60" 6. 80"
43- 45- 4.5« so.
30‘ 30‘ 3'- 40..
15- lbu 1.5- 20-
1 1 1 A
o O O I I I I I I fl
1 2 3 4 5 6 7
LEDUM GROENLANDICUM
111232? '15 6'71}: 7.13 E87 716.” " 96
311007, an. suoors sunv
50' 7. - IOO-
40. s- so-
30- 4.59 ' 80-
20- 3- 404 V
IO- [.5- 20‘
1 1 B
O O 0 r r u v I fi
1 2 3 4 s 6 7
MAIANTHEMUM CANADENSE
no. 96
LEAVES SURV.
‘0‘ '001
. \
\
\\
329 80‘ \
\
‘\ ,.I\ .
\v,—-”‘ \\\ I,
2“ 60" \\ ’I’
h----/’
IB- 40"
a- 20-
“ C
O I r '7 I Y fl
0 1 2 3 4 a 1'5 7

ORYZOPSIS ASPERIFOLIA

 

.2484

FIGURE 20. Survival and Growth, 1960 Plots, cont.

Code as indicated above the vertical
axes. Block numbers along the horizontal
axes.

A. Pteridium aguilinum
B. Rubus higpidus var. obovalis

C . Vacc inium mEtilloides

RATIO
ust/BDTH.
1.5-

QB-

05-

 

BOT H. LONBBBT HOT. LONBEST

FgONOIM.

q

400-

300-

 

Loru.to§éésr
BRANCH, u u.
100.

80-

4o«

20-

 

7170740. U u.-
500-

400'

zoo.

IOOq

 

NO.
FRONOS
159

mu:

 

55-
40~
30.
ZCF

IO-

AVIS

 

.1

FiU BlJS

281

2b

14-

 

N0.
BRANCHES
a.

'V A (:0 l

247

 

SUR\(
100-

80‘

601

40«

 

 

 

P T E RI 0| U M

 

 

4i 5 8 1I

AlilJIL.IIIU M

 

 

 

 

 

96
13%?“ A
\
\
\ /
80- \\ 1’
\ I
\ I’
80-
40-
20'
B
0 I T T fl
4 B 6 7
Ell SI’IIDLIS 0'B()\/A Ll S
96
SURE
100‘
804
60" '.'... J, ’1”
If:*\\V//
40‘ ’I
\ I,
‘J
20‘
C
c J 5 s' 7
NI U If L L.0 I D E S

FIGURE 21. Survival and Growth, 1960 Plots, cont.
Code as indicated above the vertical
axes. Block nimbers along the horizontal
axes.

A. Vaccinium anggstifolium
var. nigrum

B. Vaccinium boreale

C. Toxicodendron radicans

 

248

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

£3717. 1.636287 ' ' 'Jo'.’ ' 96
anaemia. saaucuss sunv.
1001 35') IOO-
BO“ 28‘ BO'
60" 21‘ 60‘
401 ”1 4°“
20« 7.1 20- A
0‘1 0'1 0 fi' 7— I— —I T f “w
I 2 3 Q B 8 7
....... ”v Ac_c1_~ up M ANGUSTI FO LIUM VAR. N l GRU M
LBTN. Lanes" 110. 96
BRANCH,IU. snaucuts sunv.
IOO-1 35'] IOO}
80-1 28-1 801
601 21-4 BOT
40-1 141 40-1
20" 74 20'
8
OJ 0'1 c I I I 7F I I fl
I 2 3 4 B B 7
VACCINIUM BOREALE
3737147533? 1713:]: 'E'A'Vié's 11' .37:-
WIOEST LEAF PERJ’LAMT U H.
751 51 IOO \
1 i\
"A
60.1 4'1 309 \\\
.4 ,A
\l} / .‘\.
45-1 1 60-1 3‘
3 "'?>\":’ ,‘J'.'»fi'..-\m
. ,. v 41.}.
“' CV .....
307 2-1 40-1 ‘\, :5 K.---r
IB- l- 20-1
C
0.1 0-1 0 i 1 fl 7 I I fl
I 2 3 4 B B 7

TOXICOOENDRON RADICANS

249
surprises. 0n the whole survival percentages, despite
large differences due to the different sensitivities of
species to tranSplanting. reflect the approximate scale of
relative drought tolerances which one would postulate 0n
the basis of distributions along the gradient and natural
tendency to grow in open or shaded sites. In Table XIV we
find the Species which we would guess to be most xerOphytic,
Danthonia epicata (Egg) with by far the highest prOportion
of living plants (about 90%) in unshaded plots. Eggggflpgn-
sylvanica (gag), which also reaches its natural cover max-
imum on the best-drained soil, also survives fairly well
(about 70-75%) in similar plots. Oryzgpsis agperifolia
(figs), whose frequency and cover peak on Croswell and Au
Gres sands respectively but which is only slightly below
maximum abundance on best-drained Grayling sand, survived
in about the samc.pr0portion as gaggg, but in plots shaded
after the first droughty spring fallowing tranSplant.
(Judging from.the changes which have occurred in natural
vegetation since clear-cutting, drainage differences made
little difference in vegetation.here between G and WC al-
though they apparently did between the latter two and AS.
It is therefore assumed in interpreting these results that

such drainage-related differences on the two best-drained

250
segments, such as might occur between m on WC and
Orxzopsis on G, are trivial compared to those related to
presence or absence of transpiration-limiting shade
cloth).

Quitting the three blueberries (Vaccinium flag.)
for the moment, two other species have natural cover max-
ima on intermediate segment WC. They had quite low sur-
vival in unshaded plots on this segment, fieridium
aguilinum G’_t§) with but 15% and ngtonia peregrine (£4)
with only one plant in 90. Evidently some tranSplanting
intolerance is involved here, and in Pteridium's case this
seems to be related to the exposure to full sunlight and]
or the droughty spring following transplant since trans-
plants of this species exhibited nearly 1007!. survival in
the naturally-shaded 1960 plots. This is suprising, however,
since local farmers complain that this species' rhizomes
can produce new shoots even after cutting by farm imple-
ments, and Watt (1945) reports similar behavior for a
closely related taxon, a rank weed in British pastures.

Moving to the presumably more mesophytic species
(the foregoing are all classified in the xerophytic
groupings (SS) and (CX) in Chapter II!) , we have Maianthe-

m canadense (Mai) and Epigaea rgpens (Epi). Their natu-

251
ral maxima are on the bog margin but they are abundant on
G (hence, (MD) in the Chapter IX grouping). Haianthemum
in 1963 had 37% survival on well-drained G, actually some-
what better than more xer0phytic Pteridium's 15%, but the
shaded plots in which the first was grown probably contrib-
uted to this. Maianthemum was the only species tested to
exhibit substantial transplant recovery between 1962 and
1963. But of 20 Epigaea individuals on the driest segment,
also in shaded plots, none were able to survive, and even
onumuch more poorly-drained sites only 1 of 20 lived out
the first winter. Epigaea appears to be another case of a
species difficult to move successfully; according to
Parmelee (1965) this is c0nsistent*with its reputation.
Finally, £92513. groenlandica (922) and 9_9_r_n_u_§_ canadensis
(Cor), belonging to the (MN) grouping of "obligate" bog
margin species which do not reach the "dry end", managed
less than 25% survival under the most favorable moisture
conditions (shaded, an AS).

As discussed above, the 1961 experiments were
subjected to full sunlight at least through the first
droughty spring, and for a longer period for those not
shaded artificially. The moisture stress thus created

probably revealed the rather great physiological differ-

252
ences between species which were not evident under the
protective natural canopy over the 1960 plots. ggeridium
is noteworthy in this reSpect; unfortunately there is no
basis of comparison for two other very low survivors in
the interaction plots, ggmptonia and Epigaea. The much
smaller block of native soil transferred with the 1961 as
contrasted with the 1960 plants may also have contributed
to the lower survival in the interaction plots.

The Vaccinium taxa do not behave as expected. Un-
shaded y, myrtilloides, the bog margin Species, has higher
survival on the xeric gradient segment than does unshaded
X},§§gg§ti§glium var. ni rum,the gradient Spanner, while
the latter fares better than the former on A8. One might
exPect the latter, abundant on the "dry end", to be hardier
than the former, but the almost inevitable conclusion is
that 1. mtilloides is the more desiccation-tolerant of
the two, at least while they are recovering from trans-
planting. Yet the 1960 germination experiments indicated
that it was not specialized germination requirements which
in nature excluded‘y,'mygtilloides from the "dry end". The
present results, moreover, fairly safely preclude absence
of humus or a specialized soil microbiota as this species'

limiting factor, since the 1961 plants had little native

1961 competition plots.

253

TABLE XIV

Survival of shaded (*) and unshaded

individuals of all species tested, summaries for all experiments.
F 62: fall, 1962, S 63 3 spring, 1963, N: number of original plants
upon which figures are based (in parentheses).

Species

Carex
pensylvanica

 

Comptonia
pe r i g rina

*Qoptis

groenlandica

*Cornus
canadensis

Danthonia

apicata

*Epigaea
repens

*Maiagghemum
canadense

*Oryzopsis
asperifolia

Pteridium
ag uilinum

Vaccinium
wrtilloides

Vaccinium
angustifolium

var. ggrum

*XLaccinium
inflstifolium
var . Iligrum

 

 

C} lNCI .AS
livin living living
F‘ F‘ S F‘ "FS"'
02 63 1% 62 63 N 02 03 N
75.0 71.3 (240)
1.1 1.1 (60)
22.0 (150)
24.0 (150)
95.3 89.7 (234)
0.() 0.0 (20) 5.0 0.0 (20)
7.8 36.9 (103)
73. 70.0 (60)
22.8 15.0 (80)
21.3 21.3 (150) 40.7 45.3 (150)
14.7 15.3 (150) 64.0 04.0 (150)
45.0 40.0 (60) 78.3 73.3 (60)

 

 

254
soil transferred with them. This leaves differential com-
petitive ability of the two taxa as related to each other
and to environment and other Species as about the only
feasible explanation, and indeed since 1. myrtilLoides'
cover is much higher on relatively sparsely-vegetated
Saugatuck than in the luxuriant vegetation on Au Gres it
may be a poor competitor.

The morphological measurements for Vaccinium also
yield somewhat puzzeling results. In the data for the
1960 plots, no change in numbers of blueberry shoots
across the gradient could be detected (Figures 20 C, 21 B,
pages 247, 248). In the 1961 plots, however, both shaded
and unshaded 1. a_n_gustifolium var. £132.19; as well as X.
mygtilloides produce many more Shoots on more poorly-
drained AS than on G, the differences being significant
in all three cases (Table XV A, page 257). But again, as
in survival, it is the seemingly drought-tolerant y.
ang1_1_stifolium var. 2.1.8.412 which shows a stronger favorable
response to increased soil moisture than the natural obli-
gate mesophyte, g. myItilloides. me farmer's shoot num-
ber differences are significant at P < .01, the latter's
only at P < .l.. 1. mygtilloides, however, has longer

shoots on the wet end, the difference being much more

255
highly Significant than that for the 1960 plots. .2.
gaggstifolium.var. ni rum, which in the 1960 plots ex-
hibited a better-defined shoot length increase on the "wet
end" than did‘y. m tilloides, here shows only slightly,
non-significantly longer Shoots on AS than on excessively-
drained G. By this criterion, then, it is the bog margin
Species which, under the relatively great moisture stress
of the 1961 plots, seems to show the most striking mois-
ture-related increase in vigor, in conformity with its
distribution in the wild.

On both gradient segments, shading seems to Stimu-
late production of longer shoots in‘!. angustifolium var.
‘gigggm than does growth in the Open (Table XV B), the
difference being significant however only on A8 (there was
no evident difference in this character between the more
and less heavily shaded blocks in the 1960 plots, inciden-
tally). While this suggests that some shade is elongation-
stimulatory, as does performance in nature, we must remem-
ber that increase in shoot length with reduced light is
essentially universal among green plants. Number of shoots
may be a better criterion of vigor. The 1960 and 1961
plots give somewhat conflicting results for this character

in relation to shade, but in neither case are differences

256
significant.

For those species included in both 1960 and 1961
experiments, a comparison of morphological measurements
between the two sets of plots is informative (Table XVI,
page 259). The two sets of results are not truly compara-
ble, for treatment of the soil, Spacing, and time allowed
for recovery from transplanting, just to name a few fac-
tors, are totally different in the two experimental series.
Nonetheless, it is worth noting that all of the values are
slightly to considerably smaller on the cutover (1961) area
than they are on the same soil types in the naturally
shaded (1960) plots. This indicates a slower growth rate
in the open sunlight. Moreover, it is the apparent xero-
phytes which show the smallest, the more mesOphytic Species
which show the greatest difference. Thus, for example,
means for extreme xerophyte Danthonia from the 1961 plots
are roughly comparable to 1960 plot values on the same or
adjacent soils. Somewhat less xer0phytic 9_a_1_:_e_:_:_' clumps
average some 20% smaller in the cutover strip , although
23312213513, with its cover maximum "down-gradient" from
283933 , (though with cover continuing near maximal up to
the "dry end") shows as little difference as does Danthonia.

More mesophytic gteridium's various measurements are some

1961 interaction plots.

257

TABLE XV

Differences between Vaccinium taxa

grown under different conditions (significance by t—test).
A dash (-—) indicates P>.10.

A. A COMPARISON OF PLANTS GROWN ON GLGRAYLINQ) AND
ON AS LAU GRES-SAUGATUCK) SEGMENTS.

Species
and shading

X. angustifolium
var. nigrum,
shaded

V. angustifolium

 

var. r_1_igrum,
unshaded

X. mrtilloide s ,
unshaded

Character

no. of
branches
longest

branch, mm.

no. of
branches
longest

branch, mm.

no. of
branches
longest

branch, mm.

mean on

S?“*"'7E§ .1 1i5§_
11.1 17.4 2.25 0.01
57.7 63.8 0.80 .—
9.9 16.4 2.25 0.01
44.5 51.0 1.21 _-
12.3 17.2 1.91 0.1
33.4 45.2 2.37 0.02

B. A COMPARISON OF PLANTS OF VACCINIUM ANGUSTIFOLIUM

 

 

 

WWW
mean on
§_0_il type Character - _EQI. L P<
Shaded Shaded __—
Grayling no. of ' . ' ____._
branches 11.1 9.9 0.59 ._
longest
branch, mm. 57. 7 44.5 1. 65 _.
Au Ores-Saugatuck
no. of
branches 17.4 16.4 0. 43 _
longest

branch, mm.

63.8 51.0 2.78 0.01

258

30-502 smaller in the cutover strip, while margin-dwelling
Maianthemum though artificially shaded in the cleared
area, yields mean measurements there less than half those
under natural canopy. The xer0phytes predominate in clear-
ings and savannas, and it is likely that they are better
adapted to such sites and so less inhibited in them than
are such Species as Maianthemmn.

Vaccinium ngtilloides shows about the same degree
of growth supression in the cleared strip as does 3;. m-

tifolium var. ni , additional evidence that the first

 

is not necessarily the less drought tolerant of the two
as their macrodistributions would seem to indicate.

F. EMOTION ELDTS: INTERACTION. The only
interaction experiments in which survival was high enough

to permit usage for their original purpose were those on

WC involving Carex pensylvanica (Car) , Danthonia gpicata
02g) and Eteridium aguilinum (:tr) (Table XVII, page 264).

Carex and Danthonia occupy obviously somewhat different
microhabitats in nature, the latter being more abundant on

much more xeric sites, yet the negative Cole coefficient9

9A measure of association based upon presence or
absence, similar in principle to the coefficient used
here.

259

TABLE XV'I

A comparison of the means of growth measurements for the
same Species in 1960 and 1961 plots. measurements June. 1963.
All measurements of linear characters are expressed in milimeters.

 

1260 plots 1.9111 interaction plots
' segments segments (5 3 shaded)
$339.53 "‘ <21 3333 is. r. 1;? (3:1 9 G: W C _e_s 4.5.?
CAR bdth. clump 79. 5 61.0 -- - 53.9 r *-
DAN bdth. clump 85.2 68.4 —- —- 64. 2 -— --
MAI no. shoots 3. 9 6. 8 —- 1.2 -— ~— ——
bdth.
longest leaf 49. 4 43. 5 -- 19. 1 -- - -
ORS no. leaves 31. 1 22.9 - 29.1 - -— ~—
PTS no. fronds 1.6 1.4 —— - l. 3 - «a
hgt. long-
est frond 272 358 -— -- 211 -—— -
bdth. long-
est frond 242 258 -— - 130 -— --
VAM no. branches 19.1 18. Z 12. 3 -- ~ 17. 2 —
lgth. long—
est branch 58.7 74.0 33.4 -— «— 45.2 -—
VIG n0. branches 28.5 31.8 9.9 11.1 -- 16.4 17.4
lgth. long-
est branch 57.6 94.4 44.5 57.7 — 51.0 63.8

*Species names corresponding to these coded abbreviations are
listed in Appendix B.

260
between them is not quite significant at P‘<:.05. The pair
gaggg surrounded by Danthonia and its reciprocal were
tested in order to see if there might be either positive
interaction between them reducing the value of this nega-
tive coefficient, or a negative interaction not strikingly
evident in nature because of habitat differences, but more
dramatic when the Species are forced together. An effort
was made, in other words, to substantiate or reject a non-
significant coefficient by experimental means.

Danthonia was found to be highly negatively cor-
related with Pteridium on G (Cole coefficient = ~O.436,
PW<:.Ol), so it was surrounded by Pteridium to see if the
latter might inhibit it. Qgggg, which exhibited circum-
neutral or only slight negative, non-significant correla-
tion with Eteridium.on G and WC, was also surrounded by
Eteridium in order to compare its performance with that
of Danthonia. In.view of Qggggf seemingly "indifferent"
reaction to Pteridium in nature, it was supposed that
whatever growth suppression it exhibited in contact‘with
the latter might reflect ordinary interSpecific inter-
action, not a Specific inhibitory effect, and that such
an effect could be postulated for Danthonia only if this

grass showed substantially more growth suppression than

261
sees.

Only two of twenty-four comparisons, unfortunately,
yield differences Significant at the P < .1 level or
higher, which is slightly less than the expected number
among randomly fluctuating variables. Nonetheless, it may
be worthwhile to examine whatever trends are shown.

9333;, in sets A and B of Table XVII, is always
by far the largest when grown alone. It is smallest when
surrounded by D_:a_nthonia, intermediate in size when sur-
rounded by Pteridig or by itself. But Danthonia in set
C, and in sets C and D combined, is also smallest when
surrounded by individuals of its own Species. It is
largest not when alone, but when surrounded either by
29535 (set C) or by Pteridium (set D). Solitary plants
are in one case larger, in one case smaller than those
surrounded by Danthonia itself, and when the data are
combined (C + D) there is little difference.

If these differences reflect real tendencies, the
picture which begins to emerge is as follows. 9_a_1_:_e_x seems
to be a much poorer competitor than Danthonia, for it is
strongly inhibited by this grass, and to a lesser extent
by Pteridium. But Danthonia ac tually seems to thrive on

competition with the other two Species, and possibly cer-

262
tain interactions with these Species are beneficial to it.
It also appears to be a better competitor with itself than
is‘gpggg.

.In comparing the experimental results with those
of the present correlation analysis, we find that‘gpgpg
(no. 10 in Table VII, page 97 ) actually has a.moderate
(though non-significant) positive ; coefficient with Danthonia
(no. 18) in.1 m2 quadrats on.WC, the soil type of the 1961
experiments. But‘we find that on a 1/4‘m2 scale the cor-
relation of the two becomes Slightly, non-Significantly
negative, in line with the somewhat negative Cole coef-
ficient on 1 m2. Thus the two xer0phytes may have an af-
finity for the same or Spatially close microsites, but on
a smaller scale Danthonia may inhibit the nggp close to
it. In any case the negative shift is not strong, and
possibly any advantage Danthonia has over gppgg is reduced
under the more mesic conditions below the géppp_canopy,
where Danthonia's natural cover is lower than in the Open.
On G, where perhaps more of the area is available to both
Species or their relative competitive ability is different,
their correlation on all scales is circumneutral.

‘gp; g with.§teridium exhibits non-significant

though consistent Slight to moderate negative correlation

263
on G and WC, as in the earlier Cole analysis. It does not
seem as strongly inhibited by this fern when growing natu-
rally under Jack pine canopy as in the environment of the
interaction.plots. Danthonia, which in the experiments
does not seem.to be inhibited by Pteridium, shows about
the same degree of negativeness with it in the‘g coeffi-
cients on these two soil types. But the 15 coefficient
of the latter pair on G, 1 m2, in common with the Cole
value, is more highly negative than the‘gs and significant
at P <:.001. These coefficients based upon presence only
thus suggest that Danthonia and Pteridium occur together
rarely but, on the basis of the less negative 3, they seem
to have mutually high cover where they do. Thus, with a
combination of experimental results and two correlation
measures, we can suggest that the negative 15 and Cole
correlations are due not to an inhibition of one by the
other, but to differing microsite preferences. An exami-
nation of partial correlations further shows that none of
the other variables Studied strongly affects the‘gppgpppip:
Pteridium correlations, and that the two Species must dif-
fer in their preferences for some abiotic factor or a

factor not studied.

G. COMMENTS ON INDIVIDUAL SPECIES. In this sec-

264

TABLE XVII

Results of interaction experiInents, and tests
of significance between the differences.

 

 

 

 

 

mean , . .
clump Significance tests
bdth. be-—-
treat- sur- central tween
Exper- Center ment rounded plant, treat-
iment species no. by * in mm.ments t P<:<*
A Carex 1 Dan 41.1 1&2 1.02 ~—
2 Car 60. 7 1 g, 3 1.65 -—
3 ~— ' 89. l 2 8c 3 0.77 ——
B Carex 1 Ptr 65. 7 13: Z 0. 58 ~-
2 Car 78. 8 1 8c 3 1.04 —
3 — 93.8 2 8c 3 o. 55 —
C Danthonia 1 Car 88.5 1 & 2 1. 96 O. 1
'— 2 Dan 57.8 l& 3 0. 55 -
3 — 79.0 2 8c 3 1.22 -
D Danthonia l Ptr 87. 5 l 8: 2 0.61 -
2 Dan 77.2 1 8c 3 1.01 —
3 —- 67. 3 & 3 0. 55 --
A +B Carex 1 Dan 41.1 1&2 1.65 _—
" 2 Ptr 65. 7 1 a, 3 1. 70 _—
3 Car 70. 3 1 & 4 2.02 0.1
4 — 91. 3 2 3: 3 0. 22 —
2 & 4 0.87 —-
3 a, 4 O. 94 —
C +D Danthonia 1 Car 88.5 1 8c 2 0.06 —
2 Ptr 87. 5 l & 3 1.63 -—
3 Dan 67.5 1 g, 4 0.96 —
— 73.2 2 .5, 3 1.42 -—
2 &4 0.85 —
3 3‘4 0. 45

f . . .
"Dan = Danthonia, Car = Carex, Ptr = Pterid1um. A dash 1nd1cates

that the central plants were surrounded by nothing.

**A dash indicates p>o. 10.

265

tion, the information available about each Species from
the several approaches is tied together. The Species are
treated in alphabetical order, Qpppgppip and Epigppp being
omitted because they seemed to exhibit nothing but an in-
tolerance to tranSplanting. In each instance the Species
name, its abbreviation symbol, and the grouping in which
it is placed in Chapter IX are given in that order (the
code for the last two is given in Appendix B, page 375).

'ngpg pensylvanica (gag) (CX) (Figure 17 A, page
244). The very large mean clump diameter in Block 1 is
enough to cause differences between Blocks 1-2 and 6-7,
and between the former and Block 7, both significant at
the P <:.l level. This combined with an unusually sharp
drop on Shaded segments 2 and 3, suggests a competition-
independent preference for fairly open "dry" sites. In
such Sites, however, Danthonia ppicata may tend to out-
compete it in nature, or so the results of interaction
experiments (Section F) would suggest.

Chamaedaphne calypulata (Opp) (B). This bog Spe-
cies survived tranSplanting well in its native habitat,
but neither survival nor morphological measurements indi-
cated any diminution of vigor all the way "up-gradient"

t0 the best drained soil (Figure 17 B). Inability of seeds

266

to germinate on such soil, on the other hand, could be a
factor stopping Chamaedaphne at the bog margin, eSpecially
if vegetative expansion is relatively unsuccessful,
Failure of flowering and fruiting may be another, since
none was observed among the tranSplants while Specimens
in.their natural habitat were doing so vigorously.

‘Qppgig groenlandicp (gpp) (MW). A.narrow-ranging
Species absent from both "dry" and "wet ends", C tis
nonetheless survived nearly uniformly across the range of
inorganic soils in cultivated plots (Figure 17 C). The
greater mean breadth of largest leaves in Block l-2 is
different, at the P <:.1.level, from.the smaller values on
6-7 and 7 alone; thus gppgip in the absence of other green
plants is perhaps less vigorous in its native habitat.
Number of leaves shows no discernable trend, and both
characters hold steady on heavily shaded Blocks 2-3 which
c0rr0borates a field-observed shade tolerance.

,gpgppp canadensis (925) (MW) (Figure 18 A, page
245). This Species, whose gradient distribution curve
nearly duplicates‘ggpgip', has its Shoots' maximum breadth
increasing somewhat from the "dry end" to the bog margin
(Block 1-2 vs. 6-7 Significant at Pi<f.02, 1-2 and 7 at

P <:.1). NUmber of shoots seems also to increase slightly

267
(non-significantly) and these differences may reflect a

preference on the part of this Species for its native Au

Cree-Saugatuck soils with or without the usual vegetation.

'Qp£p g Shows a higher post-tranSplant mortality
rate than most Species, but no significant gradient-re-
lated differences in this (dieoff to 1963, 33% in Blocks
6-7, 42% in Blocks 1-5). Both‘gppplp’and‘gpgpp§_appeared
relatively non-resistant to tranSplanting in the cleared
strip, perhaps being more poorly fitted for withstanding
the combination of drought and tranSplant Shock than the
xer0phytes.

Special requirements for seed germination.cannot
be considered the decisive distribution-limiting factor
for‘gppppp, either (see Section C), but there is a sug-
gestion that inability of the plant to flower on drier
Sites may be limiting, for all of three plants which
flowered in 1963 were in Block 5 or further "down-gradient"
(Table X, page 230), Still, the fruits of this Species are
large, red, and fleshy, strongly suggesting that they may
be bird-distributed. It is difficult to imagine any bird
so limited in‘igg distribution that it would not occasion-
ally range 30 or 40 meters up the "dry end". Besides, the

bare mineral soil of the Spaded plots seems in every con-

268

ceivable reSpect more unlike‘gpppppf native soil environ-
ment than the undisturbed Grayling surface, so that occa-
sional natural seedlings and Sterile mature plants should
occur on Grayling sand. Yet I know of no record of them
on that soil in the study area or elsewhere. This would
suggest that interSpecific interactions plays a major role
in this Species"normal gradient position. .

{ngppp, like C tis, seems to be uninhibited by
the heavy shade on Blocks 2-3.

Danthonia Spicata (22p) (CX). Clump diameter for
this Species on Blocks 1-2 is not significantly different
from that on Blocks 6-7, (Figure 18 B), although its nat-
ural habitat range suggests that it is more xer0phytic
than.§p£p§,'whose difference between these two sets of
blocks is significant. A sharp dip in both clump size
and survival on heavily Shaded Block 2-3 is, however, in
harmony with Danthonia's observed predominance in sunny
spots, and the interaction experiments suggested a com-
petitive advantage of Danthonia over,§p;p§_in such dry
sites. Seed germination difficulty related to soil fea-
tures (other than humus) is evidently not a limiting fac-
tor on the bog margin.

For no explicable reason, 80-902 of the tranSplants

269
flowered in 1962 but very few did so in 1963, although the
wild plants flowered and fruited abundantly both years.

Gaultheria procumbens (Gail) (MD). This Species,
with its highest cover on the bog margin AS but presence
on the "dry end", shows no discernible trends in the three
morphological characters measured. Although survival de-
creases in Blocks 2-3, these characters remain constant
even here, which agrees with Gaultheria's observed shade
tolerance (Figure 18 C).

There has been a slow but steady dieoff in this
Species, on all parts of the gradient. If this trend con-
tinues, it will suggest that Gaultheria requires something
in its native block of soil, possibly a suitable medium for
its mycorrhizal fungus.

Le_d_u_n_ ggoenlandicum (Leg) (B) (Figure 19 A, page
245). This bog-dweller, like Chamaeda hne, can evidently
survive and grow about as well (morphological measurements)
on the driest soils as on the bog margin in the absence of
interaction with other vascular plants. Possibly, as dis-
cussed in Section C, inability of seeds to germinate or of
seedlings to survive on dry, predominantly mineral soil may
preclude its further advance upgradient. Suppression of

vegetative expansion by interaction with other species may

270

also play a role in this restricted distribution, however,
M's distribution, unlike Chamaedaphne's, does not seem
limited at the flowering-fruiting stage in the life cycle.

Maianthemum canadense (_Mpi) (MD). Increasingly
moist conditions seem to make for much more abundant shoots
of this Species with slightly narrower leaves (Figure 19 B).
The differences in mean shoot numbers between Blocks 1-2
versus 6-7 and 7 alone are Significant at P < .l and
P < .05 reSpectively, while mean breadth of larger leaf in
segment 7 differs at P (.1 from that in 1-2. This latter
difference becomes more meaningful when we consider that
on Block 6-7, there are about twice as many Shoots per
plot, and subjective observations there showed more leaves
per shoot. The resultant increase in leaf number per plot
probably increases the chance that an unusually broad one
may be found.

Visual observations indicate that though the wet-
end leaves are narrower than their dry-end counterparts ,
they are also longer. This, in conjunction with the far
greater nunber of shoots and leaves on the wet end, prob-
ably makes for a much larger photosynthetic surface among
the individuals located there, and suggests that the Spe-

cies is most vigorous on the bog margin. Yet the 1958

271
study revealed a cover peak for this Species on soils in-
termediate between Blocks l-2 and 6-7. Thus we may be
dealing here with a Species whose growth potential is
greatest on the moistest inorganic soils, but which in
nature is held in check on these soils by unfavorable
interaction with other Species.

W lineare Mei) (MD). Although it has a
fairly strong cover peak on Au Gres, .this chlorophyllous,
annual root hemiparasite, in the absence of natural vas-
cular plant interaction, germinated with seeming indif-
ference over the entire Spectrum of inorganic soils. So
factors influencing seedling survival and growth subse-
quent to germination must be reSponsible for the cover
maximum on imperfectly-drained soil. Now, frequency is
fairly uniform across the three better-drained gradient
segments, while cover changes. It is visually evident
that individuals are markedly clunped; the results of
Cantlon _e_t_ pl. (1963) suggest that this may be due to
short dispersal distance from parent plants plus the
storage of large batches of seed below the litter by ants
or other animals. Hence the nunber of plants in a clump
probably influences frequency but little. Cantlon and
Curtis (1964) further report that germination rates are

272
apparently uniform on all inorganic soils in any given
small area such as that occupied by the present study
plots, which would account for the uniform frequency.
But their seedling survival rates and size of individual
plants vary such that maximum nunber and size of plants
per clunp would fall on about the part of the gradient
containing maximum cover in my data.

M ppperifolip Qrp) (SS). Slight de-
creases in numbers of leaves per plant and survival as
the soils become moister (Figure 19 C) are non-significant.
If they reflect a real natural "preference" for the drier
soils in the absence of natural vegetation, then the ob-
served slight cover maximmn further down gradient may be
interaction- re lated.

Pteridiun pquilinun (135;) (CX). There is a marked
increase in Blocks 5 and 6 in the combined criteria meas-
ured (number, length, and breadth of fronds) and then a
sharp decrease in Block 7 (Figure 20 A, page 247). This
peak may represent greater vigor in the area of the natu-
ral maximum, but is somewhat down gradient from the band
of lush Eteridium found at about the lower Block 4 - Block
5 level in the 1958 sampling.

Almost all of the high post-tranSplant mortality

273
recorded for this Species occurred during the winters.
fieridiun, more than any of the other Species tested, mi-
grates long distances by rhizomes from year to year; the
missing individuals may not have died but merely "escaped
captivity".

m higpidus var. obovalis (.8112) (W). This bag
margin Species is absent in nature from best-drained Gray-
ling sand, but as Figure 20 B shows its transplants sur-
vived and grew as vigorous ly there as elsewhere. Although
transplants flowered in all blocks, the proportion was
somewhat higher in Blocks 6 and 7, suggesting that these
individuals tranSplanted into the Species' natural habitat
may be healthier. (Failure to flower in 1962 may probably
be attributed to tranSplant recovery lag) .

Vaccinium boreale (YER) (MD), 1. a_pg1_1_stifolim var.

nigrum (Vig) (MD), and .I- nggilloides (Vam) (MW) (Figures
20 C, 21 A, 21 B, pages 247, 248). Of the two taxa tested

in the 1961 plots, _V. m tilloides, which is restricted to
the bog margin in the wild, seems to be more strongly fa-
vored in its shoot elongation under conditions of moisture
stress than is the seemingly more drought-tolerant y_.
apgpgtifolimn var. 9.1-£119.: which Spans the gradient. This

is as one would expect. Yet in survival as well as in num-

274
ber of shoots produced it would seem to be the more xero-
phytic 1. apgpstifolium var. piggy; which is more favored
by greater soil moisture or more inhibited by its scarcity.
Growth suppression of the two taxa in the cleared strip was
about equal. Thus these closely related entities help to
point out that physiological differences expressed in a
"cultivated" situation may not predict performance within
the context of the natural counmmity.

Neither the 1960 nor 1961 plot results Show sta-
tistically demonstrable shade-related differences in shoot
numbers, but the 1961 plants of g. angpstifolium var.
piggy; produced significantly longer shoots in shaded plots
on the more poorly-drained AS segment. 1. boreale had much
longer shoots in the 1960 plots than either of the others
and, unlike the latter, did not have demonstrably longer
ones on the "wet end".

Sporadic flowering, fruiting and seed germination
showed no discernable soil moisture-related trends.

micodendppp radicans @335) (not indigenous to the
study site). All of the measured characters (germination,
seedling survival, height, number of leaves, breadth of
largest leaf) seem to present a picture of increasing in-

tolerance to more and more mesic conditions (Figure 21 C).

275
The majority of these criteria do not Show significant dif-
ference between the xeric and mesic blocks, but the overall
trend looks fairly convincing. Since the species is only
rarely found in Jack pine forests and then usually on the
' more mesic Au Gres-Saugatuck sands, the results once more

conflict with performance of the species in the wild.

CHAPTER VIII
ENVIROMIENTAL VARIABLES

A. SYNUSIA. In Chapter III we discovered that
synusia as a whole seem to play a relatively minor role in
species microdistribution, if indeed the low strength of
Species-synusia correlations reflects a low degree of in-
fluence of these vegetation layers. Significance prOpor-
tions in species-synusia correlations are lowest on AS,
where soil horizon thicknesses may play a more important
role, and highest on R, where the tension-zone heteroge-
neity makes for marked synusium cover variation.

Correlation of Species with tree cover Q:_r_e_, var-
iable no. 48 in the tables) must reflect almcst entirely
the influence of this synusium upon other species rather
than the reverse. Most of the trees in the area are about 60
years old. Now, Watt (1947), among others, has demon-
strated that clonal Species migrate about, so that around
1900, when germination and seedling survival of the present
trees could have been influenced by clones of Vaccinium,
9.229%: Gaultheria, etc., the positions of these clones
probably bore little relationship to their present ones.

276

277

According to field observations supplemented by
exact cover data (Table I, page 27 ) we have an Open my;
banksiana forest on the upland, a narrow band of dense,
taller £33393, Egg and _I‘aiig straddling parts of Saugatuck
and Kinross on the bog hingeline, and scattered, somewhat
stunted trees of these Species in the bog. The heavy tree
cover on the bog margin, as one might expect, seems to pro-
duce the largest nunber of high correlations with itself,
mostly negative (though none are significant because N is
small). But on the whole tree cover Shows the lowest pro-
portion of significant correlations of all the synusia.
We must conclude, then, that degree of canopy closure,
within the extremes found in this area, does not influence
species microdistributions very greatly. This is not un-
expected since, at these latitudes, light strikes the veg-
etation at a low angle. Hence the quantity of light
striking a patch may depend upon the arrangement of trees
or shrubs to the south which can block it, and bear little
relation to the can0py directly overhead. In Chapter V,
moreover, we showed that a large part of Species correla-
tion with tree cover might be vicarious and due to influ-
ence of the litter produced by trees, or to factors merely

associated with the trees such as hunus thickness.

278

In the bog, there is a positive correlation of tree
cover with _K_a__l_n_1_i_a angustifolig, and another shrub too rare
here to have been included in the calculations (Nemgpanthus
mucronatg) is definitely associated with the scattered
trees here. In all likelihood these reflect the "bog
margin-like" conditions, Lg. coniferous litter, low
m and somewhat more compact surface surrounding the
isolated conifers here. On the inorganic soils, a weak
tendency for other synusia to be negatively correlated with
tree cover suggests a general shade inhibition of plant
growth. This tendency might be more strongly evident in
correlations with light intensity itself, for as indicated
above the projection of tree canOpy onto a quadrat gives
but a poor estimate of the solar energy reaching the plants.
Per cent of the grmmd covered by litter is not positively
correlated with tree cover on the dry end. Bare patches
may be as abmdant in the deep litter beneath trees as in
the relatively shallow litter between them, and as much
ground may be covered by tufts of £553; and other plants
which hide the litter (and prevent its being recorded) be-
neath trees as elsewhere. Conspicuous high positive corre-
lations with hunus thickness on G (P < .01) and WC (P < .001)

may stem from the conspicuously much greater accumulation

279

of litter under trees than elsewhere, and a tendency for
this litter to remain in place while decomposing. The
circunneutrality of this correlation on A5 may then result
from the greater microrelief here, a tendency for redis-
tributed litter to accunulate in the pits catmter-balancing
its tendency to rot where it first fell. A negative coef-
ficient with got; on R may reflect this same tendency on
the even more severe microtOpography here plus ecesis of
trees only on the mineral matter of tipup mounds, or else
concentration of the trees on the shallower organic matter
nearer the hinge line .

§L1r_ub_ cover (_s_1_3_1_:_, no. 49) rises steadily from G
through DG (Table 1, page 36 ). Judging from its rela-
tively high prOportion of significant correlations even
after we eliminate those with its own major components it
is a somewhat better predictor of species microdistribu-
tions than is tree cover, although this need not necessar-
ily imply that it exerts a greater influence. Indeed the
distribution of this synusium could itself be influenced
by members of the remaining synusia, especially tall ones
like gteridium which might cast a shadow over shrubs be-
neath them or species forming dense tufts (e.g. 9933;,

mosses) which could mechanically inhibit deveIOpment of

280
shrub seedlings or young rhizome buds growing upward
through them.

Most of shrub cover's correlations with individual
species suggest nothing new and no general tendencies.
However, shrub, herb, and moss cover's correlations on the
bog margin, R, are mutually positive, prdbably arising from
a visually evident mutual aversion to the areas beneath the
hingeline trees. The non-significant negative shrub-herb
correlation on AS, on the other hand, may reflect competi-
tion for light and.mcisture between.the synusia here where
both have lush cover.

Shrubs' positive correlations with AOL (fall litter
thickness) on A3 and with A0” thickness everywhere but on
WC suggest that shrubs may act as collecting points for
litter (a negative one with spring litter thickness (lig)
on R.then merely reflecting the absence of shrubs beneath
litter-shedding trees). Or possibly most shrubs simply
tend to grow in natural litter collection.points (micro-
depressions).

.flggb cover (hgb, no. 50) is about even across the
two best-drained segments and Au Gres sand; it goes down
to about 1/3 of this value on Saugatuck and further de-

creases, less drastically, on R and DG (Table I, page 37).

281

After exclusion of correlations with those species
which maythemselves be an important part of the herb
synusium, we are left with a possible trend towards posi-
tive correlation of this synusium with other species over-
all, but no discernible tendencies with individual species
or groupings. Strong negative coefficients with spring
litter thickness (lig) on G and WC may arise along with
shrub cover's negative correlations with litter parameters,
from.a retardation of evaporation and thus of soil warming
in the spring prOportional to litter thickness. A conse-
quent inhibition of spring rhizome buds and seedlings in
the more thickly covered patches might result.

Herb cover has in addition several other signifi-
cant correlations with litter parameters, other synusia,
and soil horizon thicknesses. A negative correlation trend
between herb cover and horizon thicknesses on A8 suggests
either a tendency for herbs to occur with younger, shallower
and perhaps nutrient-richer profiles, or else simply greater
herb cover in the shallow-profiled, better drained portion
of AS which is less frequently saturated.

Throughout, moss cover's Qggg, no. 51) correlation
with per cent of litter cover (lip) is very highly and sig-

nificantly negative. Now, mosses occur on quite different

282
kinds of microsites, patches of the same species being
found on the high htmocks surrounding tree bases or on
flat ground far below the mean level of the area. It may
be that these seemingly quite different kinds of sites are
similar in being so situated that they are exposed to rel-
atively high wind currents, and so constantly cleaned of
litter. Several milimeters' thickness of dead leaves would
be enough to cover and exclude light from the small repent
gametephyte stalks of these bryOphytes. These different
sorts of sites would also be relatively high in surface
moisture which may favor mosses because of their shallowly
situated rhizoids, the tree trunk bases because of stem
flow and the low flat epots because of gravitational pud-
dling of water in them.

Cover of the mosses exhibits a bimodal curve with
maxima on the two extremes of the gradient and a low on
Croswell sand (Table I, page 37 ). The maximum cover
attained on DC is many times higher than the secondary
high on G and represents the nearly solid cover of Sphagnum
on the former. We might be inclined to assume that the de-
crease in moss cover from Grayling sand to Net Grayling and
the sharper one between the latter and Croswell was due to

the existence of "dry end" and "wet end" moss species, with

283
no species having Optima on the intermediate soil types.
Yet all three of the inorganic soil mosses contributing
appreciably to this synusium mirror this bimodality in
their individual cover curves! Caliergonella schreberi
(ggl, no. 8), constituting the majority of moss cover on
all inorganic soils, is maximal on Grayling with a second-
ary high on Saugatuck, minimal on WC like moss cover gen-
erally. lgrgghythecium salebrosum (ggg, no. 7) "peaks" in
cover on wet Grayling and Au Gres, and Dicranum undulatum
(gig, no. 19) on Grayling and Kinross. Cantlon and Gillis
(unpublished data), moreover, found the same to be true in
several similar £1.33; banksiana gradients within a 50-mile
radius of the present one, so it does not seem to be an
accidental peculiarity of this particular stand.

The causes of this distributional phenomenon may
be related to the afore mentioned intolerance of mosses to
litter. Pteridium gguilinum (25;) has a cover maximum
nearly coinciding with moss' minimum and bare area @113)
is decidedly minimal in the same place. Possibly the
luxuriant Pteridium.here acts as a windbreak, reducing the
velocity of air currents to the point where they can no
longer clear many spots of litter. This would account for

the scarceness beneath it of bare areas and consequently,

284
perhaps of mosses.

The moss cover-AOH correlations are very positive
on G and AS, not surprising in view of the visually con-
spicuous affinity of“many'mosses for organic matter. But
this correlation is missing on intermediate WC, where the
moss low occurs (Table VII, page 118). If this is related
to the moss minimum on the latter segment it is not clear
just how.

Since inorganic soil moss species and lichen taxa
treated individually are so few in number, they will be
discussed here rather than in Chapter IX. Sphagnum and
Eolztricum of the bog, however, fit in so naturally with
the discussion of the assemblage native to DC that they
are included with it (Chapter Ix, Section F).

The three common inorganic soil mosses are de-
cidedly positively correlated with each other on the two
best-drained segments (the corresponding correlations were
not calculated for AS). In the field it is evident that
Brachythecium salebrosum.(§£§, no. 7) and Dicranum_ggdglgr
Egg (gig, no. 19) patches occur inside of larger patches
of Caliergonella schreberi (93;, no. 8) or interdigitating
with the latter, as if all three had similar requirements

as suggested also by the aforementioned.peculiar distribu-

285
tions. Or else a successional relationship exists with
patches of pioneer Caliergonella once established, facili-
tating colonization by the other species.

With the xer0phytic clonal vascular species on AS,
mosses tend towards slight negative correlation, perhaps
reflecting the greater moss cover on more poorly-drained
Saugatuck. With the more mesOphytic ones on G, the i
coefficients of the moss species are decidedly positive,
suggesting moister microsite preferences on the part of
the bryophytes, but this is only weakly reflected in the
_r_ coefficients with moss cover as a whole. So these tra-
cheophytes apparently tend to occur with mossy patches but
not to produce their best cover there. This is especially
conspicuous with W lineare. The number of both
seeds and seedlings per unit area in mass patches tends to
be higher than in non-mossy areas, but the larger plants
are usually in non-mossy spots (Cantlon 1965).

The edges of flacciniun patches on the dry end seem,
on the basis of correlations, to be more conducive to mess
growth than those of lush 9333;. But this may be an arti-
fact created by the spatial impossibility of other plant
growth in the tight and extensive tufts of this sedge.

Negative correlations of moss (here mostly Sphaggmn)

286

with both per cent and thickness of litter cover on the
organic soil, DG, probably reflect the near-absence of
mosses there under litter-shedding trees and old M-
1.32922 clones. Negative moss-tree and moss-shrub corre-
lations seem to substantiate this interpretation.

Lichen cover (1.531, number 52) also has a decided
bimodal distribution somewhat parallelling that of 993,
with high points in this case on G and R and the low point
on the Au Gres of AS (Table I, page 37). It is even
clearer here than with mosses that this is not due to
changes in species composition of the synusium along the
gradient, for all three of the common lichen taxa have not
only bimodal frequency curves but "highs" and "lows" in
the same places as does lichen cover overall. In this
case the peaks in the curves coincide with the areas of
relatively sparse understory vegetation which these sym-
bionts seem to find ideal. There is also an abundance of
relatively xeric microsites on the two maximal segments,
on G the whole excessively-drained segment itself and on
R the taps of high tipup mounds. Their near-hiatus, more-
over, corresponds to the soil type having by far the
highest shrub-herb cover. The same lichen taxa continue

into the bog, and in fact produce several times the cover

287
here which they do on AS.S Here, though apparently on the
surface, they are attached to abundant dying Chamaedaphne
branches beneath the Sphagnum.

The lichen taxa have high positive 525 correlations
with each other leaving little doubt as to their high de-
gree of physiological similarity suggested also by the
frequency curves. With vascular plants generally, the
lichen taxa collectivelly and individually eXhibit an un-
usual negative correlation trend, very marked with the more
mesophytic species fOund on G and WC. The trend is most
marked on G; on this best-drained segment the best micro-
sites for lichens are "drier" even than elsewhere and hence
‘most unsuitable for meSOphytic plants.

Moss and lichen cover tend to be circumneutrally,
not negatively, correlated with each other even though the
Species of the former seem quite meSOphytic. But there are
some similarities besides the bimodal frequency or cover;
both have a predeliction for ground bare or nearly so of
light-eliminating litter. Lichen cover's relativelwaeak
negative coefficientS‘with.per cent litter cover (lip)
everywhere except on G can.prObably be attributed to the
relatively small aereal extent of lichen.patches relative

to quadrat size except on the "dry end". On G, lichen

288
cover also has a high positive correlation with bare ground
Q93) and a negative one with litter thickness Q15), both
reflecting the intolerance of lichens to covering of their
photosynthetic thalli.

Lichen correlations with ADM and soil horizon
thicknesses can probably be dismissed as macrodistribu-
tional coincidences, while a negative one with AOH on C
may be a reflection of lichens' association with bare,
hard-packed soil there. Unlike mosses, lichens seem to
avoid shrub and particularly Vaccinium cover, but are
circumneutral with the dominant sedge gage; which mosses
seem to shun.

B. LITTER PARAMETERS. It is the tree Species
rather than the shrubs and herbs which contribute the bulk
of the litter, although Chamaeghne contributes a great
deal of the total on DC. Thus most correlations of Species
with these parameters do not reflect the Species' contri-
bution to them. But Species may influence the litter pat-
terns in other ways, such as the larger clonal Species
apparently acting as wind-reducing litter traps (Section A
of this chapter). On the three best-drained segments the
great bulk of the litter is of my; banksiana, but on R

and DC Bicea, £13; and Chamaedaphne play major roles and

289
thus differences in initial chemical composition of leaves
could be related to changing Species microdistributions.

On the whole Spring-summer litter thickness (lig,
no. 54), fall thickness (32;, no. 57), thickness of older
compacted, fungus-impregnated litter, or duff ($22, no. 58)
and per cent of the ground covered by litter (lip, no. 56)
or by nothing (255, no. 55) seem to be less important than
the other "environmental" parameters as Species patterning
forces. This one can judge from the relatively IOW'pro-
portions of significant Species-litter correlations ("LI"
in Column 2 of Table II, page .55, as compared to "S-" and
"H0" in adjacent sub-columns).

Lé£,and‘§gl are measurements of the same parameter,
litter thickness, not only at different times of year, but
using quite different techniques. Therefore a comparison
between their correlation tendencies will serve as an il-
lustration of just how much difference such factors can
make. ILiE, recorded in the late Spring through mid-summer
(mid-June through late August), was measured by poking a
ruler down through the litter at five points within.the
largest quadrats, namely the center and the four corners
of the 1/4 m2 quadrat , and averaging them. There was thus

an unavoidable tendency to compact the litter Slightly,

290
and thickness of the litter hidden within patches of tus-
socky graminoids such as gigs); and Danthonia and of some
mosses was not included, both leading to underestimation.
In measuring 9_9_; (from mid to late September), a block of
soil (used also in measuring soil profiles) was cut with a
flat shovel in the center of the quadrat. This procedure
tended to loosen the litter along the edge of the block
slightly so that, in measuring each of the four block
sides in order to obtain an average, over-estimates of
thickness were made. Litter hidden within tussocks was
also brought to light, further increasing the estimate of
321. over that of Iii. To be sure, we do find slightly
higher mean thicknesses of 9_9_; than of .l_i_t_:_ on the three
best drained segments, while on R and DC §9_l_ thickness is
much the higher due to the inclusion in it of much litter
hidden in the near-ubiquitous Sphamun and not recorded
in the $1.3; measurements.

Also making for possible differences between the
two measurements are the Shedding of new L199}. banksiana
leaves during the spring and summer and the decomposition
of older ones during the same seasons. During the growing
season also there may be a redistribution of litter by wind

as postulated in Section A in connection with moss distri-

291
butions, and a differential accumulation related to tree
trunks, Shrub cover, and microrelief in the path of pre-
vailing winds. This may lend to an overall leveling,
whereas snow melt and emerging plants in the Spring may
lead to the heaping of litter in small piles. Coeffi-
cients of variation10 (Clark 1961), which measure the ab-
solute variability of factors regardless of means or units
of measurement, range on different segments between 1-1/2
and 2-1/2 times as high for the Spring-summer measurement
LL; as for the fall measurement 92;. This seems to bear
out the postulate of a leveling during the growing season,
the higher Spring variability reflecting the presence of
thick heaps and thin patches between them. Corroborating
the leveling postulate also are the correlations of both
spring and fall thickness with per cent litter cover (Lip)
and per cent bare ground gag) (Table VII, page 119).
Spring thickness, _I_l_i_t_:_, is circumneutrally correlated with
both as would tend to be the case if thickness were highly
variable and permitted extensive bare patches in the same

quadrat with deep litter piles. Fall thickness, 93;, is

10These were not calculated for any species in
Chapter IX because there the values would reflect primarily
the prOportion of quadrats in which a Species is absent.

 

 

 

 

292
very negatively correlated with per cent bare ground, quite
positively with per cent litter cover on G, the segment
where the bare patches are most abundant and extensive.
This suggests that here the litter patches are broad
enough so that their thicker centers are far from the bare
Spots. Some of the higher variability of Spring thickness,
however, may originate in the failure to record thickness
of litter hidden in graminoid tussocks at that time.

Both litter thickness measurements, eSpecially the
Spring one'lig, exhibit a decidedly negative trend in their
correlations with synusia. This was attributed in Section
A of this chapter to retardation of Spring soil warming and
consequent bud inhibition by thicker litter. Possibly,
though, some of the larger Species, eSpecially the dense
tussock-former.§§gg§, mechanically exclude thick litter
buildup.

{Aglrspecies correlations are not nearly so high or
significant as those with‘lig,‘very likely because the dis-
tribution of litter during the critical soil-warming period
was more nearly reflected in the Spring-summer.;;£.measure-
ment than in the fall reading,Ԥg;. This could, however,
be due merely to the lower variability of‘ggl.

The postulate changes in litter distribution during

293

the summer are really not suggested by the correlations
between the Spring-summer and fall measurements themselves,
which are highly positive and Significant everywhere but an
AS. But on the basis of various other correlation patterns
we can reconstruct something of the nature of such changes.
Understandably very significant high positive correlations
of litter with tree cover are considerably lower on G, WC,
and BC when the fall measurement, gal, replaces Spring-
summer‘lig as the thickness criterion. On AS and R, on
the other hand, there is a substantial increase in positive-
ness of this correlation in the fall, as if the sharper
microrelief of these segments acted as a series of wind-
breaks allowing litter to remain in place where it falls
below the trees. This conflicts, though, with the at-
tempts to explain moss distribution patterns in Section A
of this chapter, for the picture deve10ped there suggested
that the dense Eteridium cover on WC might be a more im-
portant*wind-reducing factor than.microtopography. There
is also a suggestion discussed in Section A that dense
shrubs on AS and R.may act as important litter accumulation
points.

Danthonia (252), to all appearances an extreme

xerophyte among common dry-end Species, is circumneutrally

294
correlated with Spring.li§'but negatively with fall gel on
G. Litter is probably easily blown off the dense clumps
of this grass that tend to occupy xeric micrOprominences,
hence the negative Shift in correlation with it during the
growing season. There is a correSponding change from pos-
itive‘with'lig to circumneutral with'ggl in the correlation
of hummock-dwelling Trientalis on AS, probably for similar
reasons. 0n the other hand, correlation changes in the re-
verse direction of 93555 and dry end mosses from negative
with'lig to circumneutral with.§gl.seem.attributable to the
'methods of measurement. These methods, as mentioned above,
reveal litter hidden in tussocks of these plants.

Generally positive correlations of _l__i_.£ and _agl with
duff, or fungus-impregnated older litter (egg) and even
with humus (52h), which deve10ps from.litter, furnish evi-
dence that litter falls and tends to decompose in.place.
This trend seems strongest on the "dry end" and is not ev-
ident on DC where ggm represents primarily dead Sphaggum
rather than leaves.

Species correlations with duff may reflect direct
influence of this horizon upon Species more than do those
with litter thickness. Cycles of ACM accumulation and de-

composition are more nearly equal in length to 3-4 year

295
cycles of alternating occupancy of microsites by two or
more Species found by watt (1945, 1947) and Kershaw (1958),
among others. Also, most Species' winter buds, as well as
seeds of W lineare (Cantlon 31; pl. 1963) and prob-
ably of most Species whose reproduction in the area is pri-
marily sexual, are located at the juncture between the AGM
and AOH horizons (Olsen 1963). Many have rhizomes and much
of their absorptive surface in the AOM! A thick duff mat
might act as insulation, deterring both freezing and desic-
cation, and along with litter retarding soil warming in the
spring.

A.tendency towards moderate negative Species-AOM
correlations on AS may be a consequence of the last, but it
could also result from macrodistribution phenomena which
place the lushest cover of most Species on Au Gres rather
than Saugatuck, where AOM is decidedly thicker. (The only
Species Significantly positively correlated with AOM on
this segment, Vaccinium.my;tilloides, is in fact one of the
few reaching maximum cover on Saugatuck). A slight posi-
tive correlation with moss cover and a surprising very
highly positive one (P < .02) with lichen cover on A5 sug-
gests that while loose litter may inhibit these plants, a

thick AOM horizon is actually favorable to their growth and

296
that they grow 93 522 of it.

Correlations with per cent of litter cover (lip)
and of bare ground (gag) seem to yield little information
not available from correlation with the other litter meas-
urements. It would seem logical that they should be nega-
tively correlated with each other but in fact they are so
only on G, the lusher vegetation on less well-drained seg-
ments perhaps taking up most of the space not occupied by
litter, or bare patches being small relative to quadrat
size. A.positive correlation on R results from relatively
small quantities of both litter and bare patches on low,
Sphagnum-covered areas here.

The cover of bare patches on WC is only a small
fraction of that on the segments on either side of it.
£55 is positively correlated on this segment with Pteridium
(25;), the large conSpicuous fern which reaches a maximum
here. This plant may, though it is associated with the
bare patches, cover most of them with its own litter.

Per cent litter cover exhibits an overall positive
trend with other Species, both xer0phytes and meSOphyteS,
on A8. It is difficult to explain why with our present
knowledge, eSpecially since plants occupy Space which would

otherwise be litter-covered. But a negative coefficient of

297
litter cover with Shrub cover as a whole, on AS and R, is
the only evidence here that plants might tend to exclude
litter in this fashion. A decided negative correlation of
the bog species with litter cover on R must surely reflect
the near-absence of plant remains from the Sphagnum-covered
portion of the segment, upon which these Species grow.

C. SOIL HORIZONS. There is little evidence for
Species-soil interactions. Those high and significant
coefficients which the horizons exhibit with Species can
mostly be explained simply by macrodistributional changes
in Species cover and horizon thickness within gradient seg-
ments. For example, we have the tendency discussed in
Chapter III, Section H, for xer0phytes to be negatively,
and meSOphytes positively correlated with horizon thick-
nesses. They tend to occur on better and more poorly-
drained macro- or microsites, reSpectively, so the trend
can be ezqalained by horizon thickness' being generally
inversely proportional to efficiency of drainage. The
experimental studies (Chapter VII) did demostrate that at
least one Species (Chunaed_a_phne calmulata) , may be limited
in its macrodistribution by inability to germinate without
sufficient organic material Q93) , but even in that case

the requirement could be for increased moisture or some

298

other factor itself generally associated with thick humus.

Horizons are on the whole positively correlated
with each other (although, strangely, not on WC), probably
because of a tendency for the soil profile to deve10p as a
whole. Therefore the soil on a recently created tipup
mound will generally have all thin horizons, soil in a less
recently disturbed Spot thicker ones. The material leached
from a thick A2 horizon on more poorly-drained soil must
form.a relatively thick B, etc., other factors being equal.

Starting at the tap of the soil profile, humus
thickness (59h, no. 59) may be important in that a concen-
tration of'winter buds, rhizomes, roots and seeds, and
hence of plant activity, is found within it, as well as in
the duff, or at the juncture between the two. The mean for
lynlthickness gradually doubles between Grayling and Sauga-
tuck sand, then increases much.more rapidly between Sauga-
tuck and Kinross sand and also between the latter and
Dawson-Greenwood peat, there becoming the peat itself.

Just'how important is humus thickness, and at what
point increasing thickness ceases to have much influence
upon Species macro-and microdistributions, are questions
for future experimental testing. At any rate there are no

particular trends in species-aoh correlations nor, for that

299
matter,an unusual prOportion of significant correlations,
so the simple amount of humus may be less important than
some other characters of this horizon. Correlation trends
'with tree cover and litter parameters were discussed earlier
in this chapter.

mean organic-mineral or A1 horizon thickness (§;;,
no. 60) is nearly the same throughout the gradient although
coefficients of variation (see Section B of this chapter)
indicate that it is about as variable within each segment
as the other horizons. Nonetheless, it exhibits a very low
proportion of significant correlation with Species.

The cause of a slight positive trend of all Species
with the leached horizon (§;g, no. 61) on C, may be that
both plant cover and this leached horizon's thickness are
increasing rapidly in the direction of moister WC. An un-
usual negative trend with somewhat meSOphytic Species (MD)
on.AS and a positive one on R seem to be macrodistributional
also; patterns similar to these have been discussed.

Since the B horizon is relatively far below the
surface, its thickness (Egg, no. 62) in itself may have
relatively little influence upon plant growth. Far more
important may be the degree of cementing of the humus and

sesquaoxides which the Bh contains, for the resultant range

300

in consolidation may run from slightly "crumbly" to liter-
ally rocklike. The B horizons of the study area represent
the whole gamut, with no cementing on best-drained Grayling
sand and a hardpan which must be chipped with a pickaxe on
bog margin Saugatuck. This cemented B horizon, or ortstein,
may tend to hold heavy rainfall in the horizons above, sat-
urating the soil and plant roots until it finally seeps
through or perhaps is removed by evapotranSpiration! But
it may also act as a barrier to root penetration, so that
during droughts these organs cannot follow the downward
retreat of the capillary fringe (cf. Joffe 1949).

Unforttmately, I had no objective way of measuring
degree of cementing in the field, so measured only B thick-
ness. There is no evidence that thickness and degree of
cementing are related. For whatever reasons, there is a
tendency on AS for xer0phytes to be positively and meso-
phytes to be negatively correlated with this thickness.

Gley is orange motteling caused by local sesuqua-
oxide deposition in saturated soil. The highest level at
which it is found (gig, nmnber 63) is thus a "bench-mark"
for the maximum level of persistent standing water in the
soil. Gley was encountered above the B horizon consistent-

ly only on AS, and on that segment it was nearly always

301
visible at the juncture between A1 and A2 horizons; it
could not be seen.in the A1 because of black organic
staining in that horizon. Thus 11, in practice, is a
total of the upper horizon thicknesses AOL,AOM,AOH and
A1, and since correlations with it are thus correlations
with the total upper part of the profile they are of lim-
ited value. A negative trend in gley-xer0phyte correla-
tions and a lesser one with Species in general can.be
accounted for by thicker upper profiles, hence deeper gley
on the moister, more Sparselyvvegetated Saugatuck, but

these are not unique revelations.

CHAPTER IX
CHARACTER OF THE VEGETATION

The preceeding chapters have examined the m-
Chamaedaphne transition in the light of various analysis
techniques, discussing the more unusual or striking char-
acteristics of Species and coununity distribution patterns
revealed by each one. We have thus been examining the veg-
etation structure from various angles as one might measure
and describe various facets of a crystal, and it now be-
hooves us to put these facets together so as to reconstruct
the multidimensional ecological situation which exists.

In the discussion which follows, the species are
placed into several convenient grouping which have been
mentioned in previous chapters. These groupings are based
primarily upon the part of the gradient where the Species
have their maximum cover, although in one case two groupings
are also separated on the basis of presence or absence of
strong clonality, and two taxa of Vaccinium which are abun-
dant throughout the gradient are each included in two sep-
arate groupings.

These assemblages, let us emphasize, are visualized

302

303

here neither as "basic units" _s__e_n_s_u Hepkins (1957) discussed
in Chapter IV, nor as any other kinds of formally-derived
groups, and indeed they resemble such units only remotely.
The demarcations between the groupings discussed in the
present chapter are arbitrary, like those between Hapkins'
or Goodall's units. Moreover, it is extremely doubtful
that there are in nature groupings of Species all of whose
probabilities of occurrence vary equally, or in the same
direction, with all possible kinds of Spatial or temporal
changes. Nor is it likely that the same single dominant
factor, such as soil drainage, is the predominating influ-
ence on these probabilities for all Species involved. But
the arrangements here are prObably much more natural than
those derived by the latter authors in that decisions are
based upon information from.several different sources, not
just a single Simplified formula. That reality is re-
flected in the present groupings is further suggested by
the similarities evident among their members in reSpects
not considered in the grouping procedure itself. This is
particularly true of their interspecific associations.

.A. SMALL SOLITARIES (SS). The Species here in-
cluded are characterized by:

1. highest cover on one of the best-

304

drained soil types,Grayling sand (in-
cluding "Wet Grayling") or Croswell
sand in all but two cases, £52325
serotina (Egg) reaching a cover max-
imum on An Gres sand and Amelanchig;
saggginea (gmg) on Saugatuck sand,

2. all but one, Amelanchier egggginea,
present on Grayling sand, the best-
drained soil of the gradient and all
absent from organic soil,

3. relative rarity,

4. relatively low cover even where they
do occur,

5. all but one lacking in strong clonality
in the study area; that one, Arctosta-
£11119; uva-ursi e.g.-3), with only small,
scattered clones here.

The (SS) are thusva rather heterogeneous assemblage‘
of small, mostly non-clonal and non tussock-forming Species
whose areal extent is not great, which do not constitute a
conspicuous part of the vegetation. Three tree seedlings-

are included, Englanchier oblggifolig ($9) , _Qelanchier
peggginea (Ame), and Prunus serotina (Pru), one shrub,

305
Arctostaphylos uva-ursi (Arc), a small shrub Rubus cana-

densis (Rub), the five perennial dicotyledonous herbs

Anemone guinguefolia (Lin), liaise; pings; Ql'_i_o_), A353;
"$5.151 M2" (L313): m M (4.11;): and __.__.3_S°11da °
"lanceolate‘lgpg_petiole" (S21) (the last three are come
posites), and two grasses, Oryggpsis pungens (922) and.Q.

gsEerifolia (Ors) . 11 .
On the basis of their "behavior" here and in sites

somewhat different from the study area, many of these are
dependent upon disturbance for their occurrence. The tree
seedlings may tend to germinate anywhere there is a local
disturbance in the soil. Composites which appear in pas-
tures often.require small Openings in the sod for ecesis
(Keever 1950, Caruso 1963). The three found here may have
similar requirements, although they are growing under giggg
canOpy. ‘§E§g§,canadensis is a con8picuous denizen of
burned or "brushed" areas and Arctostaphylos is a somewhat
pioneer dune shrub along the Lake Michigan shore about 60
miles west of the study area. In this particular study
area (though not necessarily in other communities where

some of them occur) these Species probably play the role

11Life durations of species according to Fernald
(1950), .

306
of opportunists. They are mainly sexually reproducing here
and prObably seed in where other species, the real main-
stays of the community which preempt most of the Space in
the models of Whittaker (1965), are by chance absent at a
given‘moment.

. It is probably because of this "semidweedy" char-
acter, rather than.physiologies which require perfectly
drained conditions, that cover Of most of these Species
"peaks" on the better-drained soils where vegetation cover
as a whole is relatively light. The tendency to occupy
Sparsely vegetated Spots, as well as the absence of large
clones which in themselves would act as a Strong pattern-
inducing factor upon other Species, may be reSponsible for
the (SS)' exhibition of relatively few significant corre-
lations (Chapter III). Since most significant correlations
are positive, possibly because of an overall lush-sparse
mosaic in the vegetation (see Chapter III), scarcity of
significant coefficients among the (SS) may indicate that
they produce high cover where cover of other Species is
low. Beaman and Andresen (1965) report that subalpine
Species, in the context of a true alpine community in
Mexico, seem to play the role of Opportunistic users of

unoccupied Space like that postulated for (SS) here. They

307
similarly exhibit the least positive or most negative cor-
relation trends among the members of that community.

The dependence of the (SS) upon conditions favor-
able for germination and seedling survival in general may
account for their overall more positive correlations among
themselves than with members of other groupings, as well
as their overall very high positive ones with W
lineare, (Me__;___l, no. 29). The latter is an annual, and
therefore perhaps also favored by such conditions.

From Table VII of the correlation coefficients
(pages 91 ff.) we discover that there seem to be two sets
of mutually positively correlated Species among the (SS),
one consisting of Ananone (no. 3 in the tables), Arctosta-
£11119; (no. 4), 2x25228218 aSperifolia (no. 32) and Solidago
(no. 40), the other of 5333; 'm 9333;" (no. 5), A.
.1351}; (no. 6) and y_i_._o_l_§_ (no. 45). Within the Anemone etc.
set, all of six possible pairs are Significantly positively
correlated on G on either 1 m2 or 1/4 m2 scales, as are two
of three on WC. Within gets; etc. all three pairs are
significantly positively correlated on one of these two
segments. The two sets are connected to one another by a
strong positive correlation (P (.001) on G, l m2 between

Aster 'm ovate" and Solidago. Prunus (no. 34) shows

308
affinities with both groupings, while the other four Spe-
cies, Amelanchier oblongifolia (no. I), _A. saggginea (no.
2), ngzgpsis pungens (no. 31) and,§gbg§ (no. 37) are es-
sentially uncorrelated among themselves or with members of
either set.

Correlations of the members of these two sets with
other Species and variables are of little help in suggesting
causes of their apparent affinities for one another. Nei-
ther's members can be shown to be consistently mutually
positively or negatively correlated with any other entity,
so we must conclude tentatively that one or more unmeas-
sured factors are reSponsible. Interactions between group-
ing members seems unlikely in this case, because the small
sizes of these plants does not suggest a large or strong
Sphere of influence.

These correlation sets more likely result from sev-
eral Species' germination and seedling survival being fa-
vored by the same combination of conditions. Thus certain
(SS), whose seeds were already present throughout the area,
might quickly genminate where a tuft or part of a clone of
one Of the major Species of the community had died. They
could then thrive there for a year or two until another

major Species had found its way into the Spot and displaced

309
them. The absence of other vegetation on the Spot mean-
while would account for the generally circumneutral corre-
lations which (SS) exhibit, tantamount in this area to
negative correlations because of the overall positive
trends. Or a very local disturbance phenomenon could be
Opening up microsites favorable for colonization by sev-
eral (SS) species at once; a possible one in this Study
area is the paw marks made by deer in attempts to unearth
fruiting bodies of the subterranean fungus Elaphomyces
(Gillis 1959). Seaman and Andresen (1965) report ecesis
of subalpine Species, which seem to play a role Similar to
the (SS) in their study area, on the loose earth at the
edges of rodent burrows. Or multiSpecific insect seed
caches, much like the uniSpecific ones of Me re-
ported by Cantlon _e_t_; _e_l. (1963), may serve as sources of‘
clumps of several (SS).

In comparing these two sets of (SS) with the
HOpkinsian basic units (Chapter IV), we find that on G,
members of both are included within a single large basic
unit along with many other Species with which they have
little affinity (Figure 7 A, page 137). On wc (Figure
7 B), MM and 11.91; of one set form a unit of

their own at the P < .01 level, not connected to other

310

units. But here, too, members of the other set remain
with their atrange bedfellows in the large basic unit.
Some of the species not in either set are in the large
unit on both soils, while Others constitute uniSpecific
units unto themselves or participate in small ones together
with one or two non- (SS) species. We have a further demon-
stration, then, that a growing technique based upon an
oversimplification and vicarious correlation does not yield
information comparable to that from one using information
from several sources and involving "subjective" judgement.

On the "dry end" of the gradient there is a washer
of Significant positive correlations of (SS) with both
Vaccinium angpgtifoliun var. piggy; and with shrub cover
as a whole (no. 49). This could be a reflection of the
field-observed affinity of these (SS) Species for the more
"hmmocky" microtOpography where Vaccinium is lushest, and
also to a Spatial exclusion from 9_a_§_e_x pensylvanica Car,
no. 10) tussocks, for only one (88), Amelanchier oblgpgi-
$1.15, is Significantly positively correlated with 9933;;
on G or WC and only at P < .05.

A tendency of the (SS) towards positive correlation
not only with the clonal xer0phytes (CX) , but with the more

mSSOphytic (MD) and (MW) as well on the more poorly drained

311
AS, may be the result of mutual exclusion of all of these
Species from the more poorly-drained portions of this soil
type. This discussed in Chapter III.

Beginning with the individual species alphabeti-
cally, Amelanchier oblggifolia (.139) is positively cor-
related with 9933; £093) on G. This suggests that
germination or escape from seed predators of the former
may be favored in heavy patches of that sedge, which seems
inhibitory to other (SS). _A_.. sminea's (Am_s_) positive
correlations with soil horizons on AS and slight negative
trend with other species do little but reflect its cover
maximum on the more poorly-drained and poorly-vegetated
Saugatuck sand of this segment. Anelnone in the field seems
to be favored in mossy areas of buried logs and Stumps and
Sparse litter, along with such meSOphytic Species as _M_§_i.§_n_-
£2929. canadense Qgi, no. 28) and Gaultheria procumbens .
(Gan, no. 22). But the general absence of positive _r_;s
with these same Species must mean that Anemone grows best
with Sparse vegetation like other (SS).

Agctosggphylog and the two Agtg Species are fairly
typical for the grouping in their correlations. 9_ryzgpsis
228292. (9333), however, is somewhat unique in its positive
and quite significant correlation with Danthonia _Spicata

312
(Rag, no. 18), a consequence perhaps of the extreme xero-
phytism of both. 2. apperifolia (9g), is peculiar in
having cover "peaks" on both G and AS, with a low on the
intermediate WC. This low represents a depauperateness
Of individual plants or small clones, since frequency was
found not to diminish there. It seems not to be due to an
intolerance to the heavy Pteridium cover on WC, for this
W is not substantially negatively correlated with
that fern. Nor, on the basis of available evidence, does
it seem to be connected with the similar bimodal curves
for moss and lichen cover (Chapter VIII, Section A). On
G, the only segment where its correlations with the bryo-
lichen Species were calculated, 9. aSperifolia is circtmn-
neutral with all but one of them and negative with that.
80 causes of this peculiar distribution remain a mystery.
Moreover, this rice grass, unlike a few (SS), is not posi-
tively correlated with the more moisture-tolerant meso-
phytes on AS and is highly negative with A1 horizon
thickness (no. 60) and gley depth (no. 63) there. Both
of these suggest that, near the bog margin, it surely oc-
cupies the most xeric microsites and does not really have
a broad tolerance with reSpect to soil moisture.

Much of anus' shrub and herb-sized cover on AS

313
actually represents root Sprouts from larger and older tree
sized plants. Thus this Species' high correlations on that
segment possibly reflect mostly the influence of these
trees and their litter upon other Species. Elsewhere,
m' negative correlations with several lichens on WC
suggest either that it cannot ecese well on the hard packed,
Sparsely litter-covered soil favorable to these symbionts,
or that lichens grow poorly under g_r_up_u_s_ cover. My;
litter, in fact, may contain inhibitory chemicals, Since a
number of vascular Species also exhibit poor growth near
these trees.

A very high positive correlation of m with
W LM_e_l_, no. 29) on G is noteworthy. Solidago's
correlations are predominantly non-significantly negative ,
which suggests that its marked restriction to the Sparsely
vegetated "dry end" may be due to an even stronger intol-
erance of lush vegetation than that found in most (38).
119;; is also thus restricted, but its high positive corre-
lations with _Cégeépedmlculata (9251, no. 9) and W
suggest somewhat different reasons, possibly a selective
relocation by insects of seeds of all three of these Spe-

cies in the same microsites (Cantlon 1965).

B. CLONAL XEROPHYTES (CK). All of these Species

314

are characterized by the following:

1. maximum per cent cover found on one of

the three best-drained soil types
(Grayling, "Wet Grayling", or Croswell
sands) , although cover on Au Gres sand
may nearly equal that on Croswell,
cover on the best-drained soil, Gray-
ling, is at least more than half that on
any other soil type,

all Species are markedly and obviously
clonal with large and extensive clones,
cover of all Species exceeds 0.28% on
the soil type where maximum and three
Of five have over 1.0%. the (SS), by
contrast, never having over 0.48%
(discomting 33.—WEE. abundant root
Sprouts on AS), and half of the latter
having less than 0.1%,

per cent frequency, on the segment
where maximum, is on the whole much
higher than for the (SS).

In Short, then, these are the more conmon Species

having xer0phytic microdistributions , generally more abun-

315

dant, clone-forming and con3picuous than the (SS). Two,
_c_a_r__e_§_ pensylvanica (95;, no. 10) and Pteridium aguilinum
(11;, no. 35) are dominants in the herb layer on those
soil types where they are most abundant. The criteria
delineating the assemblage from the (SS) on the one hand
and the "margin dry" (MD) Species on the other are undo-
niably arbitrary, a Species like Danthonia Spicata (_D_a_n_,
no. 18) approaching the former in its low cover and ab-
sence from A8, while Cgtonia pereggina (992, no. 12)
approaches the latter in having close to maximal cover on
Au Gres sand.

Of the five Species included here, Q3335 Began-
2213; (egg, no. 9) is a tussock-forming sedge, _C_. m-
m a Slightly smaller one which resembles it but
contributes many times the amount of cover, mtonia an
odoriferous myricaceous Shrub called "sweet fern", 25p-

thonia the xer0phytic "poverty oat grass" and Pteridiun

 

the large bracken fern.

Physio-ecological heterogeneity within the assem-
blage is suggested by the correlations between the included
Species themselves. Of ten possible pairs of species, only

two, Carex pensylvanica-Cgptonia and Cytong-Danthonia,

are Significantly positively correlated with each other on

316

either G or WC, the grouping's Optimsl segments. In fact,
the (CX) seem to have less in common than do the (SS).
Throughout the various soil types and quadrat Sizes, the
former scarcely exhibit higher prOportions of Significant
correlations than do the latter, in contrast to the group-
ings of more meSOphytic Species (see Table II, page 55 ).

In Chapter III, Section C, the predominant negative
shifting between WC and AS of correlations between these
(GK) and Species reaching maximum cover on the bog margin
was discussed. It was suggested that a positive trend on
WC might result from members of both groupings here being
near their range limits, therefore being mutually restricted
to microsites in which neither competition nor abiotic con-
ditions are severe, whereas on their Optimal AS the meso-
phytes themselves act as a competitive restrictive force on
the (CX). Now, for _C_a_r_e_x pedunculatg, ngtonia and
Pteridium either all or most of these Shifts are negative
(in each case about half of them with (MD), half with (MW)).
9333:; pensylvanica, however, exhibits three positive shifts
and no negatives, suggesting that this sedge may be more
tolerant of soil saturation or of competition from the
moderately meSOphytic species than the other three. The

latter seems more likely Since 2. penszlvanica decreases

317

sharply on the more frequently flooded part of AS. There
are no correlations for Danthonia on A8.

On AS there is a tendency in the correlations of
(CK) towards negativeness with A011 (no. 59), A2 (no. 61)
and moss cover (no. 51) . These variables exhibit higher
values on the more poorly-drained Saugatuck part of AS, so
this is compatible with the cover curves of these Species
and their postulated occupancy Of the more xeric microsites.
Perhaps for the same reason gag-£5 and Cgmptonie are also quite
positively correlated with thickness of the B horizon (no.
62), which becomes more compact and thinner in wet Spots.

993535 p:e_dunculatg is poSitively correlated with B
thickness on WC, however, which on that soil suggests a
decided preference for the more poorly-drained portion.
This is in harmony with the cover curve for this Species,
which is noteworthy for its puzzling secondary cover in-
crease on Au Gres deSpite its primary maximum on G. AS
with O__ryzgpsis asgerifolia of the (SS), the low of g.
pedunculata on WC does not seem to be due to inhibition by
Pteridium or connected with similar curves for mosses and
lichens, and its causes remain obscure for the present.

Dominant gm pensylvanica seems, to the eye, to

constitute one of the primary components of the vegetation

318
on the "dry end", the blueberries vaccinium'gpgustifOIium
var. nigrum and‘y. boreale comprising the other. On each
of the two best-drained segments,‘§§5§§ is Significantly
negatively correlated with a different one of these two
blueberries but not highly so. Evidently, then, the‘ggrgg
and Vaccinium components are not arranged in a distinct
‘mosaic, but blend and combine with one another almost
imperceptibly.

Overall, 95525 has a tendency towards Slight (non-
significant) negative correlation with most other Species
on these two soils, as if its large and extensive tussocks
are inhibitory to their growth.

Qgpptonia has an ill-defined cover "peak" on Cros-
well, but its cover is relatively high from Grayling through
Au Gres. In other areas it frequently colonizes disturbed
microsites such as roadcuts, and in this it is capable of
being a pioneer like some of the (SS). Of all the (OK) it
exhibits the highest number of Substantial positive corre-
lations with other (GK) and with those (MD) which appear
relatively drought tolerant on the imperfectly drained AS,
suggesting some moisture intolerance and maximum growth on
mdcrohummocks. Eggptonia's demonstrated intolerance to

transplanting under full sunlight conditions (Chapter VII,

319
Section G) at present suggests little relating to its nat-
ural patterns.

The 1960 experiments showed that Danthonia, like
most species, can both survive as an adult and germinate
about as well on AS as on G, provided that natural inter-
action is eliminated (in fact a few plants survived trans-
planting in the bog!). 0n the basis of quadrat data,
however, it is one of the Species most restricted to the
"dry end", and positive correlation with bare ground and
lichens, Slight negatives with litter cover and thickness
are in line with a xer0phytic microdistribution also. This
must be due, then, not to Simple moisture intolerance but
rather to Danthonia's possession of the competitive edge
over other Species only in xeric situations. Such an edge
‘was suggested, though indecisively, by the 1961 interaction
experiments (Chapter VII, Section G). This grass' broad
physiological amplitude in cultivation.must therefore
shrink remarkably under natural interactive Stress. Prob-
ably its production Of'maximum cover in dry, Open areas is,
then, a result of inability to outcompete less drought tol-
erant Species in the physically choicer microsites.

Pteridium is noteworthy on WC fer its prediSposition

 

towards slight negative correlation with other Species (see

320
Chapter III, Section I), contrasting with most Species'
positive trends. Watt (1945) found that the corresponding
taxon. in British pastures severely inhibited other species
beneath or near it. Seemingly it produces an antibiotic
substance which builds up slowly in the soil but remains
for some time after the bracken clone has died in a Spot
and migrated elsewhere. As it happens, clones of 355;;-
flgg produce their densest, most extensive cover on WC
(although overall per cent cover of this fern is not higher
on this segment than elsewhere), and so this negative cor-
relation pattern may stem from strong antibiotic inhibition
here. It would almost surely be more marked if bracken
cover of a year or two ago were correlated with present
cover of other species in the same quadrats, for Watt found
inhibition most apparent in spots where the bracken had re-
cently died.

The Pteridium situation points up a need for tem-
poral as well as Spatial study of the present area, but in
the absence of such a time dimension it is possible to de-
tect some evidence of a cycle of bracken invasion, bracken
death, inhibition, and gradual recolonization. For in-
stance, the ¢ coefficient (based upon presence) of M’

dim-Danthonia on G is negative, significant at P < .001,

321

yet the.; (based upon cover) is only slightly and non-sig-
nificantly negative. Thus luxuraint Danthonia tussocks may
be favorable new colonization ground for‘gteridiug, which
would then move in vigorously and produce lush cover in the
same.Spots with good growth of this grass. But later per-
haps Pteridium kills this Danthonia, causing the negative
¢ and making 1; circumneutral rather than positive.

C. :uggggn DRY (MD). This grouping of six taxa is
characterized as follows:

1. it is similar to the (CX) in being pre-
sent on G,

2. but, unlike the latter, the cover maxima
of its members occur on Au Gres sand or,
if on Croswell sand, then their cover
on Au Gres is only slightly lower,

3. cover of all species decreases between
Au Gres and Saugatuck sand, appreciably
so in all but Maianthemum canadense
Qflgi, no. 28), although it may rise
again on Kinross,

4. all are markedly clonal, and

5. all are either dominants or conspicuous

components of the vegetation, especially

322
near their cover optima.

Thus this assemblage is also one of clonal Species,
but with Optima shifted "down-gradient" from those of (CK).
They differ from the "margin wet" or (MW) species (Section
E of this chapter) in at least being scattered throughout
the bestedrained soil type Grayling, not absent or merely
sporadic.

Members of the grouping are Epigaea m Gigi,
no. 20) and Gaultherig procumbens (_(_;_a_u, no. 22), both
rhizomaceous ericaceous ground shrubs, W lineare
Q43}, no. 29), a delicate annual in the Scraphulariaceae,
Maianthemum canadense Qiai, no. 28), a smallish liliaceous
herb, and the low bush blueberries flaccinium angustifolium
var. aim Qig, no. 44) and 1. boreale @152, no. 47).

In their negative trend with lichens and in their
high positive correlations with other (MD) and (MW) on the
"dry end" of their range, where they are probably mutually
restricted to mesic or low-competition microsites, these
Species "behave" as expected for meSOphytes. With soil
horizons, which increase in thickness with poorer drainage,
the (MD) are much more positively correlated on WC than are
the (OK), but become negatively so on A3 as befits Species

favored by moist, but not overly-saturated soils. Yet con-

323

sidering the relatively small changes between means of two
of the horizons from Wet Grayling to Saugatuck, this pat-
tern suggests a real preference for thick profiles on WC
combined with an aversion to them on A8.

A count on G and WC of positive and negative coef-
fic ients with the two flacciniun taxa and with 9553 222.31]!
M is revealing. About half of the coefficients of
other (MD) with the two blueberries are positive (+0.1 or
greater), the other half circmmeutral (-0.09 to 40.09),
with none more negative than -0.09. With $1335, on the
other hand, the pattern is essentially reversed. No (MD)
are more positive than +0.09 with it, and almost half are
more negative than -0.09. The (MW), discussed in Section
E, is another mesophytic grouping. With Vaccinium, seven
of their eight coefficients which are 0.1 or over in ab-
solute value are positive, the other only slightly negative.
The pattern with m is negative, so the (W) 's affini-
ties seem similar to the (MD)'S. By contrast, the more
xer0phytic (CX) have slightly more negative (15) than pos-
itive (10) correlations with Vacc inium (32 circumneutral) ,
if we use the same breaking points. With gage; they have
6 positive and only 3 negative (13 circunneutral). Thus

on the well-drained end Carex and its associates seem in a

324
rough sort of way to represent the xer0phytic element in a
vegetational mosaic and Vacciniun the meSOphytic one.
(The aversion of the mostly xer0phytic ($8) to M, dis-
cussed in Section A above, could stem either from supres-
sion of germination or low seedling survival of these
Species in the sedge tussocks.

All of this, incidentally, is only very slightly
reflected in the Hapkins basic unit calculations of Chapter
IV, yet another damaging bit of testimony against mathemat-
ical oversimplification. .

Epigaea, of all the (MD), has the least of its
total cover on G and WC (and thus comes closest to the
(MW) grouping), but also is relatively low on fairly poorly-
drained Saugatuck sand. This would suggest a rather narrow
physiological amplitude for this Species, except that its
cover strangely rises again on R. Noteworthy among its
individual correlations are a very highly positive one with
gianthemum on G, which is still quite high on WC. The
latter is the other (MD) Species most nearly absent on
better drained soils, so both may be very severely re-
stricted to the same mesic islands on those segments.

Gaultheria also has very low cover on G but, unlike

Epigaea, much higher cover on Wet Grayling. 0n the "Wet

325

end" of its range it also seems more tolerant than m
by virtue of relatively high cover on Saugatuck, and is
even found in one DC quadrat. Gaultheria's only slightly
negative correlations with bog Species on R are in line
with this relative abundance unusually (for an inorganic
soil Species) far "down-gradient", and suggest relative
saturation tolerance. This Species is positively corre-
lated with Epigaea and Maianthemum on the "dry end" except
with Epigaea on C. It seems better able to withstand lush
cover and possible antibiosis by Pteridium than some other
Species, and positive coefficients with thickness of AOH
(on G) and of Al (on A8) fit with an observed affinity for
organic matter such as rotting logs. The latter is not
likely macrodistributional in view of Gaultheria's nearly
equal cover on both AS soils. Circumneutrality with both
shrub and herb cover (nos. 49, 50) on AS, contrasted with
high significant positive correlation with them both "up-
gradient" and "down-gradient" from it, suggest that Gag;-
£12213. may be a relatively poor competitor in the closely-
Spaced vegetation of Au Gres and Saugatuck sands.

flisnthemm's cover curve looks like a combination
0f Epigaea's on the better-drained segments and Gaultheria's

on the more poorly-drained ones, except that it is very low

326

on R; the Species thus must tolerate moderate soil satura-
tion better than drought. Some of‘ygianthemum's correla-
tions seem contradictory. A.positive one with tree cover
(no. 48) on AS suggests that shade may be favorable but on
G, where ggianthemum presumably would be even more in need
of this tranSpiration-reducing factor, the correlation
actually becomes non-significantly negative. Yet quite
positive correlation with herb cover (no. 50) suggests
that Maianthemmn might be favored by the shading and wind-
reduction benefits of this synusium. Significant positive

S'with three of twelve (SS) are not necessarily in con-
flict with this, since the three are not the most xero-
phytic of the group, and the correSponding circumneutral
_r_-s indicate that Maianthemum's cover maxima are not with
them. The same positive g -circumneutra1 3,; pattern with
mosses suggests similarly that Maianthemum is abundant in
mossy patches but produces lower cover there than visual
evidence would suggest.

W produces relatively high cover on the
well-drained soils. It is the only annual Species among
those regularly encountered in the study areas and probably
the only one abundant in undisturbed areas in this vegeta-

tion type in general. It also has the distinction of being

327
a hemiparasite (Cantlon _e_t_ 51. 1963, Piehl 1962, Malcolm
1962). This last may not be particularly reflected in
‘Mglggpyrggfs correlations, however, since the roots of
Bing; and other woody plants into which this annual sends
haustoria could, as often as not, originate outside the
quadrat containing the above-ground portion of this host
plant. W does, nonetheless, exhibit an unusual
nunber of very highly positive é coefficients, capecially
on the well-drained end of the gradient where vegetation
is Sparse. In interstices between lush patches of vaccinium,
Qaultherig, etc. the concentration of suitable host roots
near enough the surface for this hemiparasite to attack
would be relatively low, and desiccation-induced die off of
fine mycorrhizal roots in these interstices may further re-
duce their number, resulting in low probability ofgyglgg-
pyrgg.occurrence.

Cantlon,g§ugl. (1963) and Curtis and Cantlon (1963)
demonstrated that seeds offlyglggpyrgg germinate in the mat,
or duff, and Cantlon and Curtis (1964) fbund empirically
that both seeds and potential host roots were Sparse where
the surface was nearly or quite bare of litter, duff, or
humus and also in the 25-30 emacircle around tree bases

where litter is deep. It is not suprising, then, that

328
W exhibits a slight but definite negative corre-
lation trend with the measurements of litter and 'duff thick-
ness $15, no. 54, 99;, no. 57, 933, no. 58) and also
moderate negative correlation with AOH (no. 59) on AS.
Its moderate negative correlation with per .cent of ground
not covered (_1_3_a_r, no. 55) moderate positive one with per
cent litter cover 01.3, no. 53) are also as might be ex-
pected from these observations, as is a slight negative
correlation with tree cover (no. 48), which may reflect
the absence of seeds and host roots near the tree bases.
0n AS, moderate negative (though non-significant) correla-
tions with shrub and moss cover (nos. 49, 51), on the other
hand, do not corroborate the findings of the aformentioned
workers that lush cover of these two synusia seems to re-
duce bird and insect predation on Mg}: seeds.

M's annual nature seems to show up in its many
highly Significant positive correlations with other Species
which also reproduce in this area mainly from seed, and may
therefore have similar requirements for seedling survival
(Section A of this chapter).

[accinium taxa, because of the peculiar taxonomic
problem which they present and the desirability of discus-

sing their taxonomy and ecology together, will be given a

329
section unto themselves.

D. VACCINE, AN ECOTAXONOMIC PROBLEE, The two
distinguishable blueberry taxa which Span the Study gra-
dient have recently been called two varieties of Vaccinium
matifolimn (cf. Fernald 1950). The taxonomy, however,
has been in a state of confusion, partly because both dip-
loid and tetraploid forms of the blue-berried, non-glaucous
leaved entity (abbreviation symbol, 122) seem to exist,
while only tetraploid material of the black-berried, gla-
cous-leaved entity Qig) has been found (Camp 1945, Hall
and Aalders 1961, Aalders and Hall 1963). The two cyto-
logical conditions are apparently morphologically dis-
tinguishable in some areas, but unfortunately material from
the study area has neither been examined cytologically nor
compared with properly annotated herbarimn Specimens at the
time of this writing. Nonetheless, the voucher specimens
of the blue-harried form fit Camp's (1945) description for
diploid material only, as most plants in the field appear
to also. An occasional more robust, larger-leaved blue-
berried plant seems to vary in the direction not of tetra-
ploid material, but of y. m tilloides, which is also found
in the area. These aberrant plants probably represent in-

trogression with the latter such as seems to occur in other

330
regions (Camp 1945, Hall and Aalders 1961). The two papers
just cited both report that _V_. mxgtilloides is a diploid;
if so then the gradient-Spanning blue-berried form must
also be diploid in order to readily exchange genes with it.
This evidence strongly suggests that this blue-harried
gradient Spanner prOperly belongs to diploid _V_. boreale
Hall 6: Aalders (the authors retaining y_. angugtifoliun Ait.
for tetraploid material since it matched the type Specimen).

The black-berried entity remains 1. aggygtifoliun var.

 

 

 

 

piggy; (Wood) Dole. A comparison of recent nomenclature
follows.

ABBRE- Fernald Hall 6: Aalders
VIATION 2n Cg 1945 1250 1961

212 24 1. amsti- V. 523385}: 2. boreale Hall

folium Ait. folium Ait. & Aalders

 

48 1. lamarc- 1. £52.19." 1. mstifolium
k_:l_._i_._ Camp folium var. Ait.

laevifolium
House

113 48 y. britonii 1. M- g. wtlfolium

 

 

Porter ex folium var. var. nigrun
Bickn. nigrmn (Wood) Dole
__a flood) Dole
leg 24 _V_. myrt - 1. mEti- 1. mEtilloides
nee; 11.91429. Michx.

Michx. Micbx.

331

Aalders and Hall (1963) furnish corroboration for

the above postulates. They found that the "nigrum factor"
segregated like a single dominant gene in crosses between
1. gastifolium vars. aim and gngstifolium from tetra-
ploid populations of the Canadian Maritime Provinces. In
both the laboratory and in the field the authors found
occasional intermediates with only some characteristics of
the nigrun form, such as might occur within a freely inter-
breeding pOpulation containing modifier genes. But they
mention that such intermediates do not occur in the Lake
States, and indeed I have encountered none during extensive
field work in the study area and elsewhere in Michigan.
The two entities thus appear to be reproductively isolated
pOpulations and hence separate Species (cf. Davis and Hay-
wood 1963), in this region, very possibly by virtue of the
ploidy differences just postulated.

In the grosser SSpects of their distributions, the
two taxa which Span the gradient in the Study area do not
violate one's expectations for a single Species, for their
frequency and cover curves (Table I, pages 35, 36) coincide
alm0st exactly. In Spanning the gradient, moreover, they
differ from all other species save the three my; Species

found here. Yet if we examine their correlations, they do

332
not appear much more similar than any two phyletically re-
mote Species might.

To illustrate this last point, let uS score each
of the blueberries as positively associated with another
species or other variable if it exhibits a positive corre-
lation coefficient _1; Z. 0.25 with the variable, or an 5
smaller than this but significant at P < .05. Negative
coefficients satisfying the same conditions will be ac-
cepted as evidence of negative association, and variables
whose correlations with a Vaccinium taxon fall between
these two limits will be considered unassociated with it.
Then, of the 40 species and other variables with which at
least one Vaccinium entity is positively or negatively as-
sociated, there are only 137. with which the other exhibits
the same kind of association (on at least one of the same
gradient segments, irreSpective of quadrat Size). With
777. the other is massociated and with 10% it exhibits the
Opposite kind of association (_i_.g. negative as opposed to
positive or vice-verse). The degree of similarity demon-
strated is scarcely greater than, for example, that between
the black-berried form and m _c_anadensis @515, no. 14,
of the (MW)). Using the same criteria, these two phylet-

ically remote taxa sxhibit 7% similar association with other

333
variables, 79% association of only one of the two with the
variable and 14% dissimilar association. By contrast
99532; and‘ggggi§.g£oenlandica (932, no. 13, (MW)), taxo-
nomically remote both from one another and from black-
berried vaccinium, but suSpected of a positive interaction
because of the inordinately high positive correlations be-
tween themselves, yield the reSpective percentages 50, 50
and 0. This is about four times the prOportion of similar-
ity exhibited by the two blueberries!

Thus, the two blueberries certainly Show no more
ecological similarity than can be found among separate Spe-
cies. .At least, they exhibit more ecological difference
than might normally be exPected of two entities differing
by a single gene, such as seems to separate two similar-
appearing entities in Eastern Canada. Although the similar
distributions of both across the gradient may reflect close
phyletic relationship and recent isolation of the two pOp-
ulations, the demonstrable uniqueness of their microenviron-
mental relations is consonant with the taxonomic evidence
that they must constitute two separate gene pools. Each
may contain a fair number of alleles not possessed by the
other, or, extrapolating from the findings of LBve and LUve

(1943), a ploidy difference could itself account for the

334
observed ecological difference.

.AS for specific features of the two blueberries,
their cover is maximal on Au Gres but is nowhere really
law. They contribute sufficient cover in prOportion to
the total vegetation throughout the inorganic soils to be
very decidedly the shrub dominants.

If we score each taxon according to whether it
appears the more mesophytic oerore xer0phytic of the two
on the basis of its correlations, we find'!. anggstifolium
var. m the most meSOphytic-appearing on the "dry end"
(G and WC) on the basis of fbur out of five correlation
patterns. 0n AS, however, the two appear tied with three
correlations or correlation groups each in the more meso-
phytic category. 2. boreale seems to have much more aver-
sion to the dominant shrub __Chamsedanhne (9.19;: no. 11) in
the bog, and mmch.more affinity for areas occupied by
Erigphorum Spissum,(§£i, no. 21, "cotton grass") than does
I. aggggtifolim var. Big—Inn. 0n R, X. boreale is quite
positive with herb cover (no. 50) and seems to occur more
on the Open, "peat-like", non-litter-covered patches than
does 2. mtifolium var. 111-3122.

E. MAQQIN WET (ya). This assemblage includes nine

clonal Species of inorganic soils. They are the delicate

335
evergreen herbaceous goldthread, ESE-1?. groenlandica (9311,
no. 13), the robust, codominant, herbaceous bunchberry,
£93993; canadensis (99;, no. 14), the large occasional
shrubs Swamp holly, 1:133; verticillata @135, no. 23), many-
berry, yiburnum gagsinoide; (2L, no. 43) and Mountain
holly, Nemganthus mucronata (Neg, no. 30), a lowbush blue-
berry, yaccinium ggtilloides Vam, no. 42), an occasional
small herb, Trientalis borealis (_T_‘_r_i, no. 41), a repent
blackberry, 3.29.113 hisgidus var. obovalis @g, no. 36),
and the clubmoss nggpodium obscurum 0:19., no. 27).

In their maxima of cover as well as in their "dry
end" range limits, the (MW) on the whole appear more meso-
phytic than the (MD). They are characterized by:

l. absence from G, or of too rare occur-
rence on this best-drained segment to
have had correlations calculated for
them there (thus distinguishing them

‘ from (Min).

2. having their highest cover between Au
Gres and Kinross , inclusive (whereas
the (MD) sometimes peak on Croswell

and never below Au Gres) ,

3. being otherwise like (MD), except that

336
some are not dominant or conSpicuous.

Although (MW) are thus more meSOphytic in distri-
bution than (MD), some (MW) drOp more drastically on Sauga-
tuck and Kinross than some (MD). Although the groupings
used in this chapter seem somewhat natural, they are none-
theless arbitrary, and in this case one could as well use
"wet end" distribution to separate out two groupings of
mesophytes about as physio-ecologically homogeneous as the
(MD) and (MW).

Some overall correlation trends of the (MW) none-
theless do suggest that they have a general affinity for
moister microsites than do the (MD). unlike the latter,
for example, (MW) tend to remain mutually positively cor-
related on AS as well as WC, perhaps tending to be unsuc-
cessful in relatively xeric microsites even on their
optimal segment and so being clustered where moisture is
more plentiful. They are even more negatively correlated
with lichens than are (MD) and even slightly so with mosses.
Like the (MD) ard to some extent the (CX), the (MW) are
more abundant on those portions of WC with thicker profiles
and more poorlybdrained soils, and like them shun the
heavily litter-covered areas Shaded by dense tree canOpy

along the Saugatuck-Kinross border.

337

Those Species which peak on Saugatuck and Kinross
(11.33;, Nam anthus, yiburnum, Vaccinium mgtilloides) are
predictably much less highly positive in their coefficients
‘with the (MD) and other (MW) on AS than are the remaining
(MW) that reach.cover maxima on Au Gres (Cgptis,'gggggg,
Lycgpodim, 3212.9.» Trientalis ). Evidently the former must
occupy less droughty and/oerore frequently flooded micro-
Sites than either the latter or the (MD). The overall
absence of high positive correlation between the (MW) theme
selves on R also fits the picture; these Species are prob-
ably nearer their Optima here and so not so mutually
restricted to certain portions of the area as they are on
AS, or as are the (MD) here on R.

Starting with those (MD) Species having the most
xeric gradient distributions , we have the codominant Ruby;

higpidus var. obovalis, maximal on Au Gres. Its cover is

 

only about one-third less on Croswell, but drops to a

small fraction of this on Still better-drained soils. 0n
the "wet end" its cover draps drastically also (relatively
high frequency on Saugatuck, however , indicates abundance
but depauperateness there). Significant positive correla-
tions with Danthonig gpicata and m serotina, discussed

reSpectively in Sections B and A of this chapter, suggest

338
that Rgbgg is a poor competitor, but reSponds favorably to
the usually inhibitory litter of W. Dr possibly low
competitive Strength merely relegates high 31.1.1939. cover to
relatively unfavorable sites like those under _13__r_t_1n__1_1_s_.

Trientalis, a very minor biomass component, also
occurs on G and follows §_u_l_>l1§_ ' general pattern on the
better-drained segments, but its mean cover scarcely drops
on Saugatuck and Kinross. Along with its positive corre-
lation with gley depth, no. 63 Qigbgsf is negative) this
seems to mirror a prevalence on profiles develOped under
moister conditions than those most favorable to 3213.11.51-

It has relatively low (for (MW)) positive _r_-_s and {5 s with
other margin species on A3; whether as a result of encoun-
tering its optimal moisture regemin a bit farther "down-
gradient' than they or of poor competitive ability is a
question.

Mg and m, with the very high positive cor-
relation between them discussed in Section D of this chap-
ter, reach up the "dry end" only as far as Croswell and
reach their maximun in cover on Au Gres but in frequency
on Saugatuck. Cover drops rather drastically on Saugatuck
and m' does not but otherwise their distributions are

essentially the same, as are the majority of their corre-

339

lations with other variables. Evidence suggests that their
coincidence in these reSpects may be a result of a positive
interaction between them rather than a cause of their cor-
relation, for the partialling out of other variables changes
the high positive M'ng coefficients but little, on
the whole (Table Ix, pages 159-162). On the other hand,
the tranSplant experiments (Chapter VII) seemingly demon-
strated that each could grow separately, but insufficient
time may have elapsed for diapersal of metabolic byproducts
in the soil blocks transferred with the plants. 9.952% was
shown to be independent of 9325i; at the gemination stage
of the life cycle, but the converse was not tested. At the
moment it is impossible to suggest, then, just what might
be the mechanism of such an interaction if it exists.

Rounding out the species maximal on Au Gres is
Lycgpodium. It is relatively abundant on Croswell through
Saugatuck and absent elsewhere. Correlations with this
clubm0$s were calculated only on AS, 1 m2, which makes it
difficult to say much except that it seems typical for the
grouping.

Whether Vaccinium mxgtilloides' high survival as
tranSplants in the bog indicates a true flooding tolerance,

masked in nature by competitve inferiority with Chamae-

340

9.321512 and other bog shrubs, or merely a greater ability
to survive tranSplanting under more adverse conditions
than most Species, one cannot be sure. But positive cor-
relations on R with Chamaedaphne (no. 11) and the bog grass
Erigphorum (no. 21) , are unique among the (MW) (even in-
cluding Nemgpanthus and _I_l__e_x which have their maxima far-
ther downgradient) and certainly suggest the former. Yet
to complicate the picture, _V_. mygtilloides on R also demon-
strates greater affinities with the (MD) than with the
other (MW), which one would interpret as an indication
of some xer0phytism or flooding intolerence (on optimal AS
it _i_s_ highly positive with (MW)). The 1961 experiments
(Chapter VII, Section E) suggested that this Vaccinium is
relatively drought-tolerant in the; absence 3f competition,
and now it appears that the same might be true for imm-
dation tolerance. Possibly, then, y_. ngtilloides has a
wide moisture amplitude but is a poor competitor. Its
relatively low cover in the lush vegetation of Au Gres
sand is compatible with this interpretation.

Viburnun' a cover curve is similar to Vaccinium
mntilloides' , but the former first appears on Croswell,
while the latter actually reaches Grayling. Viburnum ex-

hibits a slightly negative trend with other Species gen-

341
erally on A5, and is barely positively correlated even
'with other (MW), possibly a consequence of greater abun-
dance on the more sparsely vegetated Saugatuck here. Yet
1. m tilloides, also with a maximum on Saugatuck, shows
the usual overall positive correlation trend with other
variables, and this continues even on R, where Viburnum's
correlations are predominantly weakly negative. This
leads one to suSpect Viburnum of ability to take advantage
of microsites unfavorable to most Species, or possibly of
some growth-inhibiting ability. It does not occur out on
DC prOper in any amount and would seem intolerant of "bog-
like" conditions on the basis of its negative trend with
organic soil species.

Nemgpanthus first appears on Au Gres, going "down-
gradient", reaches a maximum on Saugatuck and Kinross and
is present Sporadically on the peat. In certain Open bogs
this Species constitutes a distinct zone along the hinge-
line, and a very high positive correlation with lichen
cover on R suggests an Optimum on the relatively lichen-
rich, better-drained side of this segment where such a
zone would occur.

flg' maximum coincides with W' but the

former has a broader amplitude on both sides though less

 

 

 

342
cover on Saugatuck. It, too, forms a bog margin zone in
some areas and occurs in this bog on "high, dry" microsites
near Sporadic trees, associated with Egg m (with which its
correlation is positive on R). 0n the better-drained side
of its range, ngx is astonishingly positive with 93g;
pedunculatg (no. 9) and ngptonia peregrina (no. 12), two
relatively quite xer0phytic Species. A high positive coef-
ficient‘with herb cover, (no. 50), which predominates on
inorganic soils, and a non-Significant negative one with
humus thickness (no. 59), which tends to be directly pro-
portional to soil moisture, suggest that,;lg§,joins‘gggg-
panthus in bordering the "dry side" of the hingeline. Yet
the two would seem not be most luxuriant in the same micro-
sites, for they are slightly negatively correlated with one
another.

_I_1_e_x_ has a high positive correlation with bare
ground (no. 55) on AS. Since this arborescent perennial's
mature plant distribution could hardly be much influenced
by small bare patches, and since the ones present‘whengglgx
was a seedling have in all probability disappeared or
shifted Spatially, one wonders whether 119; may not itself
Play a role in their production.

F. BOG SPECIES (B). The dist1UEU19hin8 features

_ 343
of members of this growing are:

1. sufficient abundance in the bog segnent,
DC, for correlation analyses (Lg. pres-
ence in 5 or more DG quadrats),

2. cover maxima on one of the two most
poorly-drained segments, DC or R, or
else present in at least 14 DC quadrats,

3. cover on each segment better drained
than R characteristically ($.33. in all
but two cases) only a small fraction of
that on the adjacent, more poorly-

drained one.
These are thus Species characteristic of the bog, dominant
or abundant there but usually not elsewhere on the gradient.
All but the two, gradient- Spanning Vaccinium taxa decrease
rapidly in abundance "up-gradient"; these last however Span
the whole gradient and thus are considered in both the (MD)
and (B) growings.

Here are included the unidominant bog shrub leather-
leaf, Chamaeggphne cglyculatg ((219, no. 11) , the prominent
cotton grass Erigphorum gpissum 05;” no. 21), the locally
abundant shrub Labrador tea, £9932 groenlandicum (Lgd, no.
25). Bog laurel, 33131.; M ($31, no. 25), the

344
blueberries Vaccinium _a_:n_gustifolimn var. 31.3323 (11.3,
no. 44) and _V_. boreale Q12: no. 47), the creeping cran-
berry Vaccinium oxycoccus (3195, no. 46), and the inconSpic-
uous small woody laurel, Elm-i; polifolia (_K_a_l, no. 24).
Also included are the dominant moss Sphagnum plumosmn (Spl,
no. 38), other Sphagnum Species ($22, no. 39) and another
moss, zolytrichmn Strictum (figs, no. 33). These mosses
together form the loose, fluffy wper substrate of dead
moss known as "peat moss", which covers virtually the whole
bog surface, and in which the roots, rhizomes, stolons and
rhizoids of most bog Species are found.

Affinity for "boggy" conditions among the (B) is
seen in a decided aversion to the heavy tree cover and
litter cover found on the portion of R resembling inor-
ganic soil, and positive correlations with the high moss
cover typical of the peat-like portion here. But a posi-
tive trend with AOH thickness on this hingeline segment is
relatively weak for some reason, although this horizon is
decidedly thicker on the portion of R adjacent to DC.

With other Species, the (B) Show fairly Simple,
straight forward and predictable trends; a possible Slight
negative tendency with the (MD) and (MW) on R and AS , and

. mutual positiveness among themselves here near their "dry

345
end" range limits are traceable, probably, to mutual areal
restriction. There is an unusual (for this gradient) num-
ber to high negative correlations between the (B) Species
on DG, where there is a wider variety of microhabitats
open to them than on R so that interSpecific physio-eco-
logical differences are manifested. This is probably re-
lated to the strongly-develOped hummock microtOpography
here, and an ecological differentiation among the Species
relative to it. It is precisely with this individuality
that the part of Chapter IV dealing with these Species was
largely concerned, and for discussion of the bog community
from the "basic unit" angle the reader is referred to that
chapter. The distinct microtOpography does induce a rather
distinct mosaic patterning on the bog vegetation, so that
this is the one segment where the Hapkins basic units seem
to provide a reasonably accurate picture of nature.

The predominantly negative correlations between
shrub species on DG, which cause them to form distinct
basic units, may involve an amensal reaction between them.
Such a reaction is susPected also from the appearance of
clones of these Species in the field, for these clones al-
most never interdigitate with one another. Possibly estabe

lished clones or their mycorrhizal fungi liberate soil-

346
diffusible substances Similar to antibiotics or auxins in
their action; potential invader shoots from a neighboring
clone of another Species might then be inhibited if they
began to grow against a gradient of such a substance.
Such a mechanism would have a selective advantage, con-
serving the energy of the inhibitor-liberating plant which
it might otherwise expend in mechanical stresses with
shoots of other species, or lose by having its photo-
synthetic surface overshadowed by these shoots. Rhizome
and fungal distribution maps of sectors might be helpful
in corroborating or rejecting these postulates.

59.19113 polifolia is circmnneutrally correlated
with almost everything. This Small Shrub seems capable of
sending w shoots within clones of other Species, and it
may be the only DG Species here which is resistant to an
amensal inhibition such as that just postulated.

Between Chapter IV on basic units and Chapter VI
on higher power regressions, the patterns shown by the bog
Species have been fairly well documented. In addition to
these an affinity of Chamaedéplhne and Vacciniun ongoccus
for deeper peats, mentioned in Chapter VIII, must result
from differences in peat age, distances to nutrients in

the mineral soil, or thickness-related prediSposition to

347
desiccation.

A possible temporal vegetational cycle in the bog
is discussed by Byer (1960). At that time, I was too
eager to Speculate won details on the basis of rather
flimsy evidence, but it does seem clear that such a cycle
is taking place. It consists in the buildw of hmocks
either, in conjunction with or under the influence of ma-
turing clones of Chamaed_§phne or other shrubs, and appears
to be a counterpart on non-permafrost soil of organic soil
cycles described by Johnson and Billings (1962).

These clones possibly become senescent won death of areas
of the clones; at any rate they are knocked over now and
then by animals or frost action, possibly creating a new
"microvalley" for lush growth of the ever-present Sphagnum
and Erigghorum. Perhaps the Shrub clones in time reinvade
these and new hammocks form.

The bog moss taxa defined here reveal in their
coefficients a fairly distinct ecological series coincident
with the postulated bog cycle. m plumosum predomi-
nates in the hollows, other Sphagnm Species on the sides
of humocks, and golfirichun on the taps with fairly
ltmuriant Chamaeghn .

These mosses also find their way wland onto R and

348
AS, but their circumneutral correlations with Species of
inorganic soils suggest either that they do not occur in
"boggy" or "peaty" pockets there, or that such pockets
are very small relative to the size of smle units.
Either Sphamum must have some probability of occupying
microsites also available to mosses such as Caliergonella
and Brach thecium, or else rhizomaceous plants, solidly
anchored in typical inorganic soil, can send shoots into

the Small pockets in which M occurs.12

12Moss Species with cover maxima on inorganic soils
and lichen Species, because of their Small numbers and be-
cause they do not readily fit into discussions of any of
the above growings, are included with the discussions of
the synusia containing them in Chapter VIII, SectionA.

CHAPTER X
SUMMARY AND CONCLUSIONS

1. This Study, part of a projected continuing se-
ries on organization in biological communities, is an effort
to describe the patterning of a vegetation by the use of
correlation techniques, to put forth.postulates concerning
the causes and ramifications of this patterning, and to
test a few of these postulates experimentally. The present
paper is primarily concerned with the first of these, the
descriptive phase, but attempts not to lose sight of the
eventual goal of explaining biological organization on a
functional basis.

2. A.re1atively undisturbed vegetation transition
between.ging§,banksiana fbrest and Changedaphne calyculata
bog, located in northern "lower" Michigan, was used in the
study. This vegetation occupied a drainage-related soil
taposequence ranging from.excessively dry Grayling sand to
Dawson and Greenwood peats, and embracing several intermed-
iate types. For analysis, this essentially continuous
sequence was divided into five segments, each relatively
homogeneous unto itself.

3. Observation of differences in the patterns ex-
349

350
hibited by Species throughout their local ranges brought to
light possible characteristics of Species which study in a
‘more homogeneous area would not have revealed. The same
may be said for the examination of several quadrat sizes as
opposed to only one. .

4. The existence of small-scale, or microdistribus
tion pattern is amply demonstrated by the prOportions many
times greater than expectation of both _r_ and f correlation
coefficients. This is true of individual gradient segments
and quadrat Sizes as well as for the Study area as a whole.
It is hardly suprising, in view of the seeming universital-
ity of non-random.diSpersion as inferred from diverse
studies.

5. The more abundant, more clonal Species, on the
basis of the relatively high prOportions of significant
correlations which they exhibit, seem to be more strongly
patterned than the more isolated, scattered Species. 0n
the same basis, meSOphyteS would seem.more Strongly pat-
terned than xerophytes.. Since most significant correlations
are positive, this suggests that clonal Species and meso-
phytes are relatively Strongly clumped together with other
Species. In the case of the former, this may be a con-

sequence of clonality itself acting as a patterning agent.

351

6. An exception to this trend (5), the distribution
of the highest correlations primarily among the rarer spe-
cies, suggests high Specificity in the requirements of these
Species.

7. Species often exhibit stronger correlation, both
positive and negative, at the margins of their ranges across
the moisture gradient than on their Optimal segments. Meso-
phytes, on the other hand, are more strongly mutually posi-
tively correlated on their optimal segments than elsewhere.
Both trends may stem from restriction to certain portions
of a segment and from greater susceptibility to microenvi-
ronmental differences near a Species' tolerance limits.
Mutual positive correlation among physio-ecologically sim-
ilar Species under relatively poor conditions, making for
overall high correlation, is common, and probably a con-
sequence of mutual restriction to similar more favorable
microsites. The opposite tendency on the part of the meso-
phytes may reflect more unfavorable reaction to one another's
competition on their part under adverse than under favorable
conditions.

8. Judging from.their relatively weak correlations
where their values are low, certain Species and other vari-

ables appear to have threshold values of cover, thickness,

352
etc., below which they do not greatly influence Species
microdistribut ions .

9. The two "soot-end" segments R and DG seem, in-
ferring from their relatively high prOportions of signif-
icant correlations, to be relatively heterogeneous, perhaps
because of their himock-hollow microtopography. WC, with
more mesophytic elements , appears more heterogeneous than
better-drained G.

10. Large changes in the correlations of variable
pairs as quadrat size decreases are predominantly in a
negative direction, and the proportion of highly positive
coefficients decreases. Perhaps this is a consequence of
increasing intensity of Spatial exclusion and competition
with decreasing Spatial distance. But slight changes are
most frequently positive, suggesting that some positive-
association-producing phenomena themselves exhibit small-
scale patterning.

11. High positive correlation greatly predominates
over high negative correlation in the study area, but slight
negative coefficients are more cannon than slight positive
ones. The former may reflect the small-scale patterning
forces, the latter the competition and Spatial exclusion,

both mentioned above.

353
12. The trend towards predominantly positive cor-

relations (11) is most marked on the well-drained gradient
segments, and modes of all coefficients are also least
negative there. Probably these reflect reSpectively the
large, visually evident, Sparsely-vegetated patches on the
better-drained soils, which are rare on the mesic to wet
segments, and a reduction in the intensity of negative in-
terSpecific interactions among widely-Spaced plants. Within-
segment heterogeneity involving several ecologically dif-
ferent groupings on R may account for that segments' exhib-
iting the lowest proportion of positiveness among its sig-
nificant correlations.

13. The more poorly-drained the soil where the max-
imum cover of a Species is found, the greater in general
seems to be the reduction in the prOportion of its coef-
ficients which are highly positive as quadrat size is re-
duced (10). No cause-effect relationship can be suggested
at present. There is no consistency in the part of a Spe-
cies' range (optimal or marginal) where it exhibits its
greatest quadrat size-related change in proportion of cor-
relations which are highly positive.

14. A predominant lush-sparse patterning in the

vegetation may account for the relatively high proportion

354
of significant and high value é coefficients, which are
based upon presence and absence alone. A more pronounced
plastic response to interSpecific interaction gi__t_:3_1_ip_ the
lush patches may offset the association tendencies result-
ing from co-occurrence, substantially reducing the magni-
tude and significance of the 5 coefficients which are
based upon per cent cover.

15. But on DG there are similar prOportions of
Significant fis and _r_s, perhaps resulting from a combination
of strong inhibition by shrubs, ubiquitous presence of most
Species, and absence of extensive Sparse patches on this
segment.

16. A number of different sets of circumstances
can be hypothesized to explain canbinations of positive é ,
less positive or negative 5, or the inverse combination.

l7. Parameters of thickness and distribution of
litter cover, thickness of soil horizons and per cent
cover of synusia, all arbitrarily designated "environmental"
parameters, seem to exert somewhat less influence on spe-
cies distributions than do other Species. The litter pa-
rameters seem even less influential than the other two.

18. Strength of Species correlations with "envi-

ronmental" parameters often seems proportional to the local

355
variability of these parameters, as one might predict.

19. Since roots and rhizomes seldom penetrate the
lower layers of peat, Species-peat thickness correlations
on DC may result not from any direct effect, but from dis-
tance to mineral nutrients, greater drought susceptibility
of thin peat, or variation in recency of peat formation.

20. Correlations between "environmental" paramem-
eters themselves are quite strong, possibly as a result of
strong reSponses of these parameters to relatively few
other variables. In the case of synusia there is, then,
an apparently stronger "environmental" influence upon syn-
usia as a whole than upon their component species. Perhaps
strong infrasynusial competitional and Spatial stress tends
to Obscure strong "environmental" associations with these
Species.

21. In general, the more xer0phytic, rarer and less
clonal Species are more weakly correlated with "environ-
mental" parameters.

22. Trends among large numbers of non-significant
correlations may be helpful in generalizing about micro-
distribution patterns, even though the small coefficients
considered singly are not.

23. In attempting to group Species within a com-

356
munity in order to arrive at useful generalizations, the
desire for objectivity and simplicity has sometimes led
biologists to use methods which do not accurately reflect
biological situations. Theoretical objections to specific
grouping techniques are discussed.

24. A particular grouping technique, the basic
unit method of napkins (1957), was investigated, and was
found to reflect the microdistributional patterns which
had been brought to light by other methods very imperfectly.
This is in accord with McIntosh (1957), who found marked
inconsistencies among several methods. Therefore, grouping
techniques which attempt to deal with complex natural sit-
uations solely in terms of a single exact formula might
best be used only as guides.

25. Partial correlation analysis is used to deter-
mine which other variables may be reSponsible for correla-
tions between particular variable pairs, and to what extent,
and to infer the influence of these other variables upon
members of the pairs. An example is given in which some
Species and synusial correlations with tree cover are shown
to be largely due to their correlations with litter meas-
urements, which depend upon the trees, and horizon thick-

nesses, which probably helped determine tree distribution.

357

26. Analysis by partial correlation can also reveal
otherwise obscure patterning and association tendencies.

As an example, both positive and negative correlations are
revealed when, reSpectively, differences and similarities
in strength of reaction to apparently inhibitory £53323
serotina litter are mathematically eliminated.

27. Plots of quadratic and cubic regression curves
seem, for the poorly-drained DG segment, to reveal biolog-
ically consequential subtleties of physio-ecological rela-
tionships and patterns which linear regressions are too
insensitive to bring to light. They Show promise as a tool
for pattern analysis, but are quite time-consuming, subject
to larger random error than linear regressions, and like
the latter are liable to be misleading owing to the limi-
tations imposed by their mathematically simple form.

28. Fourteen Species tranSplanted to all portions
of the gradient having inorganic soils and subsequently cul-
tivated fairly convincingly demonstrated that it is not
differences in ability to survive as adults, flower and
fruit, or grow vigorously under natural canopy,‘igwgh£
abgggggugg natural vascular 21535 interaction which deter-
mines natural distribution limits and optima along this

gradient. Seed plantings did clearly Show, for one of

358
nine Species and doubtfully for another, that specific
germination requirements are a limiting factor in their
Spread to better-drained soils, but other Species showed
no clear-cut differences in germination or seedling sur-
vival in cultivated plots. Thus interaction between Spe-
cies, perhaps ig;combination'with abiotic environmental
influences, would seem to play the major role in their in-
organic soil gradient distributions, although Specialized
germination requirements are a decisive factor in a few
cases.

29. Measurements of morphological characters on
transplanted species and seedlings, by'means of which rel-
ative vigor may be inferred, do Show somewhat greater dif-
ferences between portions of the gradient than do simple
‘survival counts. This suggests, then, that such measure-
ments may be better indicators of the abiotic component in
gradient distributions than is survival.

30. The frequent inundation of transplant and
seed plots on the poorly-drained organic DG segment, due
to faulty engineering, limited the usefulness of trans-
plants on this segment. Nb seeds germinated, and all
transplants died after a year or so. Noteworthy’were the

high short-term survival of one bog Species as apposed to

359
that of the other one tested, the seeming independence of
relative meSOphytism-xerophytism.in natural distributions
and ability to withstand flooding, and the fact that,!ggr
gigigg m tilloides, absent from the bog in nature, sur-
vived in these plots much more frequently than 3. boreale
and _V_. gnggstifolium var. ni run, which are naturally
abundant there.

31. Survival in the interaction plots, which were
not covered by natural canopy nor artificially shaded
during at least the first droughty Spring following trans-
plant, was considerably lower than the near 100% survival
in the 1960 transplant plots (28). under these conditions
of moisture stress, the seemingly more xer0phytic species
‘were generally the best survivors, although Species 81m!
ilar in their natural gradient distributions sometimes
differed widely in transplant hardiness.

32. On the basis of survival data and morphological
measurements for both transplant and interaction plots,
gradient Spanning Vaccinimn anggstifolium var. $59.3.“
seems actually less able to withstand drought than.bog
margin-restricted‘yg myztilloides. These two blueberries
also performed contrary to expectation'with regard to

flooding (30), and they seem a good illustration of per-

360

formance under the influence of natural interaction being
different from.that in cultivated plots (28).

33. Interaction experiments suggest inconclusively
that both Danthonia Spicata and Pteridium aguilinum_may
be capable of outcompeting,ga5§§_pensylvanica in Open
(non-tree-covered) areas, and that Danthonia is not in-
hibited by Pteridium. The last seems reflected in the
correlations under natural canOpy, which suggests a pro-
pensity for high cover of these Species to occur together,
but‘ggrgg would not seem to be competitively inferior to
the other two under this canopy. Naturally greater abun-
dance of Danthonia in open and 93555 in shaded areas may
reflect these trends.

34. Correlations with tree cover suggest weak in-
hibition of other Species by canoPy shading, and both tree
trunks and shrubs may act as litter collection points which
in turn influence other Species. There may be an inhibi-
tion of new Spring buds of shrubs and herbs beneath heavy
litter.

35. moss and lichen cover are noteworthy for their
bimodal distributions across the gradient, and each.major
component Species of both synusia reflects the bimodality.

In the case of mosses, no reasons can'be suggested, but

361
the lichen minimum is coincident with a maximum of possibly
inhibitory vascular vegetation. The common.mosses of the
study area are decidedly more meSOphytic than the lichens,
but both have in common a "preference" fer microsites
entirely free of litter or only sparsely covered and
having a thick, partially decomposed mat (A0M horizon),
so their odd distributions may possibly be similarly ‘
caused. The physioecological similarity of Species within
each of these two synusia is suggested by their mutual
positive correlations QB‘WBII as their individual bimo-
dality.

36. Correlations suggest that litter is redistrib-
uted, perhaps piled and puddled, over the winter. The
spring measurement is more highly correlated with Species
than the fall one, possibly because thick litter, along
with the partially decomposed organic mat (ADM), retards
Spring soil warming or protects the surface from_freezing
and drying, thus influencing bud deveIOpment. Yet posi-
tive correlations between litter thickness, AOM thickness
and humus (A03) thickness suggest at least some tendency

for litter to decompose in place.

37, AOM thickness may influence Species somewhat

more directly than does loose litter thickness, because AOM

362

decomposition cycles more nearly correSpond to life cycles
of many perennial Species and because many Species are
rooted in or just below this layer. At any rate, thick-
ness of both loose litter and.A0Mlseem.much better pre-
dictors of Species microdistributions than are the propor-
tions of the ground covered by litter.

38. Thickness of humus (AOH) and mineral-organic
(A1) horizons seem less associated with Species microdis-
tributions than is thickness of the leached (A2) horizon.
This seems somewhat suprising, considering that A2 is not
in direct contact with many plant organs, but perhaps the
leaching thickness is merely indicative of the complete-
ness of drainage to which the Species reSpond. Correla-
tionS‘with thickness of the deposition (B) horizon seem
to reveal little of note; the degree of cementing of this
horizon and its influence upon drainage and root penetra-
tion may be much more important biologically.

. 39. The Species were separated into five conven-
ient groupings for further study. These were based prima-
rily upon macrodistribution across the soil moisture gra-
dient but take into account all available information
about the Species. They are:

($3) - small, relatively rare, solitary, predominantly

363
xer0phytic, not markedly clonal Species;
(CX) - Species with maximal cover on better-drained
soil, forming large clones, generally much more
common than (83);

(MD) - flourishing on the more poorly-drained inorganic
soils, but occurringoften on the best-drained
segments;

(MW) - Species similar to the (MD), which however sel-
dom occur on the best-drained soils;

(B) - Species native to the bog proper.

40. The relatively small physical size of the (SS)
and their general non-correlation with more clonal Species
may result from their playing the role of temporary colo-
nizers of unoccupied Space in the study area. This role
would imply that here they were primarily young plants
growing from seed, not vegetatively reproducing ones;
their positive correlations among themselves and with an
annual, W, do suggest that they have Similar re-
. quirements which might indeed be sites generally favorable
to germination and seedling survival. They seem to be
favored by relatively hummocky microtOpography, in shrubby
patches and generally not with tussock-forming Species

which might inhibit normal seedling growth.

364

41. The (CX) are apparently relatively depauperate
on the locally thicker-profiled soils and on mossy patches
on the imperfectly-drained AS segnent, which tends to
corroborate the relative xerophytism of which their dis-
tributions across the gradient lead one to suspect them.
_C_a£g_x_ pensylvanica and _g. edtmculata, however, seem able
to coexist with mesophytic elements or mesic conditions
better than other growing members. Tussecks of these
sedges seem to constitute a distinct element (not in the
sense of a Spatial mosaic) in the "dry-end" vegetation,
one which is inhibitory to certain other species. 225-
m aguilinum is noteworthy for its apparent antibiotic
inhibition of other Species.

42. The more mesophytic (MD) seem to have a greater
affinity for a vegetational element dominated by Vaccinium
.222. than do the (CX), which appear more strongly associ-
ated with the aforementioned 9_a_r_e_x_ element (41). Micro-
distributionally as well as macrodistributionally, the
former are associated with the thicker inorganic soil pro-
files and therefore, seemingly, are favored by relatively
imperfect drainage. W lineare is unusual for its
annual character and its hemiparasitism. The absence of

reflection of this hemipa'rasitism in the correlations per-

 

365
haps results from generally large distances of both host
and parasite aerial parts from the root connections be-
tween them.

43. Fairly great differences in microdistribution
patterns corroborate morphological evidence that the two
gradient- Spanning Vaccinium taxa in the study area are
tetraploid 1. anggstifolium var. 9.1.832 and diploid 1.
boreale, for these differences suggest prior reproductive
isolation of the two and evolution along different lines.
The latter seems to exhibit occasional introgression with
diploid X. mntilloides.

44. The (MW) demonstrate an affinity for less
complete drainage than do the (MD) not only in their near
absence from the best-drained gradient segments, but also
in an apparent restriction to the more imperfectly-drained
microsites on the AS segment and in relative aggressiveness
on Still more poorly-drained R (where the (MD) seem at a
disadvantage). Outstanding among them are the pair gop-
g._s_ groenlandica and 9m canadensis; the very high
positive correlation between them and their similar cor-
relations with many other variables suggest a positive
interaction between them, although in cultivation adults

of each survived without the other and Cornus seeds ger-

366
minated without Cgptis.

45. The (B) Species, on their native peat, demon-
strate stronger patterning tendencies than are found on
the better-drained segments. This may reflect partly an
amensal reaction between shrub Species, partly the hetero-
geneous hummock-hollow microtOpography feund here. There
appears to be a cyclic pattern of buildup, breakdown and
redevelopment of the bog hummocks in.which all Species
participate. The native bog mosses, deSpite their Spo-
radic occurrence on.AS, seem to occupy the same microsites
here as do the (MD) and (MW), the major components of that
segment's vegetation. This suggests that the low proba-
bility of establishment, not a requirement for rare bog-
like microsites, may cause the rarity of these species

here.

BIBLIOGRAPHY

Aalders, L. 8., and I. V. Hall. 1963. The inheritance and
taxonomic Significance of the "nigrum" factor in
the cannon lowbush blueberry, Vaccinium Eastin-
1ium. Canad. J. Genetics and Cytology 5: 115-118.

Agnew, A. D. Q. 1957. The ecology of British rushes with
Special reference to the invasion of re-seeded
pastures. PhD thesis, University of Wales.

Bemnan, J. H., and J. N. Andresen. 1965. The vegetation,
flora, and phytogeography of the sumnit of Cerro
Potosi, Mexico. Amer Midl. Nat. (in press).

Becking, R. W. 1957. The Zurich-Montpellier School of
Phytosociology. Bot. Rev. 23: 411-488.

Bray, J. R. 1956. A study of the mutual occurrence of
plant Species. Ecology 37(1): 21-28.

, and J. T. Curtis. 1957. An ordination of
the upland forest commities of southern Wisconsin.
Ecol. Monogr. 27(4): 325-349.

 

Buell, M. F., and H. F. Buell. 1941. Surface level fluc-
tuations in Cedar Creek Bog, Minnesota. Ecology
22(3): 317-321.

Byer, M. D. 1960. An analysis of pattern and interSpe-
cific association along a soil moisture gradient
an the Jack pine plains of northern lower Michigan.
M. S. thesis, Dept. of Botany and Plant Pathology,

Michigan State University.

Cam, W. H. 1945. The North American blueberries with
notes on other groups of Vacciniaceae. Brittonia
5: 203-275.
Cantlon, J. E. 1965. Personal comunication.

_ , and E. J. C. Curtis. 1964. Unpuplished data.

.367

368

, , and W. M. Malcom. 1963. Studies
of W lineare. Ecology 44(3): 466-474.

 

, and W. T. Gillis. 1958. Unpublished data.

Caruso, J. L. 1963. Interopecific pattern in an old field
comnunity in southwestern Michigan. M. S. thesis,
Dept. of Botany and Plant Pathology, Michigan State
University.

Clark, P. J. 1961. Mimeographed notes.

Cole, L. C. 1949. The measurement of interSpecific asso-
ciation. Ecology 30: 411-424.

Crow, E. L., F. A. Davis, and M. W. Maxfield. 1960.
Statistics Manual. Dover, N. Y. pp. 176 ff.

Curtis, E. J. C., and J. E. Cantlon. 1963. Germination
of W lineare: interrelated effects of
after-ripening and gibberellic acid. Science 140
(3565): 406-408.

Curtis, J. T., and R. P. McIntosh. 1951. An wland forest
continuum in the prairie-forest border region of
Wisconsin. Ecology 32: 476-496.

_ . 1959. The vegetation of Wisconsin. Univ.
Wisconsin Press. 657 p.

Davis, P. H., and V. H. Heywood. 1963. Principles of
AngiOSperm Taxonomy. Van Nostrand, Princeton N. J.
6: New York. pp. 88-102.

Dice, L. R. 1952. Measure of Spacing between individuals
within a population. Contrib. Lab. Vertebrate
Biol. , Univ. Michigan 55: 1-23.

De Vries, D. M. 1953. Objective Combination of Species.
Acts Botanica Neerlandica 1: 497-499.

 

, J. P. Beretta, and G. Hamning. 1954. Con-
stellations of frequent herbage plants based upon
their correlation in occurrence. Vegetatio 5:

369

Dixon, W. J., and F. J. Massey. 1957. Introduction to
Statistical Analysis (2nd Ed.). McGraw-Hill, New
York, Toronto and London. 488 p.

Ezekiel, M; 1930. Methods of Correlation Analysis.
‘Wiley, New York.

Fernald, H. L. 1950. Gray's‘Manual of Botany (8th. Ed.).
American Book 00., New York.

Fisher, R..A. 1958. Statistical Methods for Research
workers. Hefner, New York. pp. 184, 198-201.

Gillis, W. T. 1959. Subterranean Elgphomyces and Rhizo-
pogon in the Michigan Jack pine region. Mycologia
51: 364-367.

Goodall, D. W. 1952. Quantitative SSpects of plant dis-
tribution. Biol. Rev. 27: 194-245.

. 1953. Objective methods for the classifica-
tion of vegetation I. The use of positive inter-
specific correlation. Australian Jour. Bot. 1:

39-63.

 

‘__. 1954. Objective methods for the classifica-
tion of vegetation 111. An essay on factor anal-
ysis. Australian Jour. Bot. 2: 304-324.

 

Graig-Smith, P. 1964. Quantitative Plant Ecology.
Butterworths, Whahington. 256 pp.

Hall, I. V., and L. E. Aalders. 1961. Cytotaxonomy of
lowbush blueberries in Eastern Canada. Amer. Jour.
Botany 48(3): 199-201.

Harper, J. L., and O. R. Sager. 1953. Same SSpects of the
ecology of buttercup in.permanent grassland.
Proc. Brittish Weed Control Conference 316-324.

Hopkins, B. 1957. Pattern in the plant community. Jour.
Ecol. 45(2): 451-463.

Hutchinson, G. E. 1953. The concept of’pattern in ecology.
Proc. Acad. Nat. Sci. Philadelphia 105: l-12.

370

Joffe, J. 1949. Pedology. New Brunswick, N. J.: Pedol-
ogy Publication. p. 368.

Johnson, P. J., and W. D. Billings. 1962. The alpine
vegetation of the Beartooth Plateau in relation to

Cryopedogenic processes and patterns. Ecol. Mono-
graphs 32: 105-135.

Keever, C. 1950. Causes of succession on old-fields of
the Piedmont, North Carolina. Ecol. Monographs
20: 231-250.

Kendall, M. A. 1952. The Advanced Theory of Statistics
(5th. Ed.). Hafner, New York. pp. 368-379.

Kershaw, K. A. 1958. An investigation of the structure
of a grassland community I. The pattern of
égrostis tennis. J. Ecol. 46: 571-591.

. 1961. Association and co-variance analysis
of plant communities. J. Ecol. 49: 643-654.

 

_g . 1963. Pattern in vegetation and its causal-
ity. Ecol. 44(2): 377-388.

_. ._. 1964. Quantitative and Dynamic Ecology.
American Elsevier, New York. 183 p.

Kilburn, P. D. 1960. The effects of logging and fire on
xer0phytic forests in Northern.Michigan. Bull.
Torr. Bot. Club (6)87: 402-405.

Kittrige, J. 1934. Evidence of the rate of forest succes-
sion on Star Island, Minnesota. Ecol. 15: 24-35.

Lindemann, R. L. 1942. The trophic-dynamic aspect of
ecology. Ecol. 23: 399-418.

Lflve, A., and D. Love. 1943. The significance of differ-
ences in ther distribution of diploids and poly-
plois. Hereditas 29: 145-163.

Malcolm, W. M. 1962. The root parasitism of Castilleja
coccineg. PhD thesis, Dept. of Botany and Plant
Pathology, Michigan State Univ.

371

[Michigan Department of Conservation. 1927. Soil and Lay
of the Land Map of Crawford County.

McIntosh, R. P. 1962. Pattern in a forest community.
Ecol. 43(1): 25-33.

McKenzie, L., and E. P. Whiteside. 1956. Soils of the
Tri-Township Area, Kalkaska County, Michigan.
Michigan State Uhiv. Agric. Experiment Station
Publ. 56 p.

Odum, E. P., and H. T. Odum. 1959. Fondamentals of
Ecology (2nd. Ed.). ‘W. B. Saunders Co., Philadel-
phia.

Odum, H. T. 1960. Ecological potential and analogue cir-
cuits for the ecosystem. American Scientist 49(1):
1-8.

Olson, J. S. 1963. Energy storage and the balance of

producers and decomposers in ecological systems.
Ecol. 44(2): 322-331.

Parmelee, G. 1965. Personal communication.

Phillips, M. E. 1953. Studies in the quantitative mor-
phology and ecology'of’ErigphorumIggggstifolium.
I. The rhizome system. J. Ecol. 41: 295-318.

__g . 1954. Studies in the quantitative morphol-
ogy and ecology of Erigphorum anggstifolium Roth.
II - Competition and diSperSion. J. Ecol. 42:
18 7-2100

Piehl, M. A. 1962. The parasitic behavior of W
lineare and a note on seed color. Rhodora 64:
15-230

Shkai,.K. I. 1955. Competition in plants and its relation
to selection. cold Spring Harbor Symposium Quant.
Biol. 20: 137-157. .

snedecor, G. W. 1956. Statistical Methods Applied to
Experiments in Agriculture and Biology (5th. Ed.)
Iowa State Coll. Press, Ames, Iowa. pp. 160-193.

372
Nmnerical Taxonomy.

Sokal, R. R., and H. A. Sneath. 1963.
PP. 169-2150

Freeman, San Francisco and London.

Veatch, J. O. 1953. Soils and Land of Michigan.
State Coll. Press, East Lansing.

Mich.

1924.

, L. R. Schoenmann, and J. W. Moon.
U. S.

Soil Survey of Rosconmon County, Michigan.
D. A. Bureau of Chemistry and Soils.

Walker, H. M., and J. Lev. 1953. Statistical Inference.
Henry Holt, New York. 510 p.

Contributions to the ecology of bracken
The structure of the

98-121.

Watt, A. s. 1945.
(Pteridiun aguilinum) IV.

commity. New Phytol. 46:
1947. Pattern and process in the plant com-
J. Ecol. 35: 1-22.

Vegetation of the Great Smoky
1-80.

muni ty.

 

1956.

Whittaker, R. H.
Ecol. Monographs 26:

Mountains.
.. 1965. Dominance and diversity in land plant
communities. Science 147: 250-260.

 

Whiteside, E. P., I. F. Schneider, and C. A. Engberg.
Taxonomic classification of Michigan soils.

1955.
Mimeographed monograph, Department of Soil Science,
Michigan State University, East Lansing, Mich.

1951. The distribution of plants on polyg-
Contrib.

Wiggins, I. L.
onal ground near Point Barrow, Alaska.
Dudley Herbarium 4: 41-52.

1934. The method of path coefficients.
161-2150

1940. An Introduction to
Charles Griffin, London.

Ann.

Wright, S.
Math. Statistics 5:

Yule, U. G., and M. G. Kendall.
the Theory of Statistics.

PP. 214, 261-287.

APPENDIX A
CRITERIA USED TO SEPARATE SOIL TYPES

Soil** Abbreviation
Limits as here defined*

 

 

5122’ symbol
Grayling G A2 thickness 0-14 mm.,‘gg
sand A2 thickness 15-19 mm., B horizon
color 1-2 on an arbitrary
scale, Lr
A2 thickness 20-24 mm., B color 1
"Wet
Grayling" WC A2 thickness 15-19 mm., B color
sand 2.5-3, Lr
A2 thickness —20- 24 mm., B color
2-3, Lr
A2 thickness 25- 49 nm., B color
1-3
Croswell C A0L +-AOM +Ao thickness under
sand 50 mm., ang/or B texture
"unconsolidated"
A2 thickness 25-49 mm., B color
3.5-5,‘g£
A2 thickness 50 um. or more
Au Gres AG B horizon texture "chunky", Lr
sand B texture "crumbly", A8L L+AOM
+'AOH thickness 8 mm.
more
Saugatuck S B texture "cemented",
sand thickness under 90 A88.

*3 horizon color scale from 1 (lightest) to 5

(darkest)

**Soil "types”'used here for convenience, not rec-
ognized by the.M1chigan Dept. of Agriculture, are placed
in quotes.

.373

Soil** Abbreviation
£122, symbol

Kinross R
and
"wet
Sauga-
tuck"
sands

Dawson DC
and
Green-

peat

*See page 373
**See page 373

374

Limits as here defingd*

B color 5, A H thickness under
12 incheg

B texture "crumbly", 25

B texture "unconsolidated" or
"chunky", A1 thickness
50 mm. or more, 25

B texture "cemented", AOH
thickness over 90 mm.

A0H thickness 12 inches or
more

no. in

APPENDIX B
ABBREVIATION SYMBOLS

Variables

 

tables 319201 name

F‘F‘P‘P‘F‘
$‘h’h’h‘CDKDGD\JO‘U13‘UDNDF‘

ripar-
\:o~u:

NNHH
HO‘DW

MNNNNNNNN
OWNVO‘UMFWN

uuwu
#wNH

Amo
Ams

5

Are
Ash
Asl

15:"

(3
Pl:
o.~n

|§|§|§Ié§

some
:3 a 3 r:

tfiFfrj'
H H-
H-H-n

EEISIEIEI?

E

2H:
Ash:

3

"U
0
O

55’

SPECIES

Amelanchier oblongifolia
Amelanchier sanguine;
Anemone quinguefolig
Arctostaphylos uva-ursi
Aster "hairyovate"
Aster laevis
gggchythecium salebrosum

Caliergonella schreberi

Qg£g§_pedunculata

Carex pensylvanica
‘ghgmaedaphne calyculata
Comptonia peregring
Cgptis groenlandigg
Cornus canadensis
Cladonia rangiferina
Cladonia cristatella

Cladonia "cup-shaped"
Danthonia spicata

Dicranum undulatum
Epigaea repens
Egiophorum spissum
Gaultheria procumbens
Ilex verticillata
Kalmiagpglifolia
Kalmia angustifolgg
Ledum groenlandicum
Lycopodium obscurum
Maianthemum canadense
Melampyrum lineare
Nomopanthus mucronata

ngzgpsis pungens

Ogyzgpsis asperifolia
Polztrichum strictum

Prunus serotina

375

376

 

 

 

 

 

 

 

no. in
tables symbol name
35 '25; Pteridium aguilinum
36 ‘ghg Rubus hispidus var. obovalis
37 332_ Rubus canadensis
38 Ԥpl Sphaggum plumosum
39 m Sphagnum £22. , other than _S_. Elumosum
4O .Sgl Solidago "lanceolate long petiole
41 ‘22; Trientalis borealis
42 1g Vaccinium gyrtilloides
43 ‘Xiy, Viburnum cassinoides
44 .ggg Vaccinium angustifblium.var. nigrum
45 11.3 Viola admea
46 17935 Vaccinium oxycoccus
47 ‘112 Vaccinium boreale
"ENVIRONMENTAL" VARIABLES
48 £55. per cent tree cover
49 ‘gh; per cent shrub cover
50 ‘ggg per cent herb cover
51 ‘32; per cent moss cover
52 ‘lgg per cent lichen cover
53 .525 total per cent cover, all synusia
54 .;;5 thickness of litter, spring-summer meas.
55 ‘Egg per cent bare ground
56 l$p| per cent litter cover
57 ‘52; thickness AOL horizon (litter, fall meas.)
58 'ggm thickness AOM horizon (duff)
59 ‘ggh thickness AOH horizon (humus)
60 Ԥ;l thickness A1 horizon (mineral-organic)
61 .512 thickness A2 horizon (leaching)
62 ‘bgg. thickness B horizon (deposition)
63 ‘31; depth to gley
ORIGINAL 5011 DP": COMBINED, ssmmms
symbol name symbol name
G Grayling sand G Grayling
WC "wet Grayling" sand WC Wet Grayling-Croswell
C Croswell sand AS Au Gres-Saugatuck
AG Au.Gres sand R Kinross
S Saugatuck sand DC Dawson-Greenwood
R Kinross and "Wet
Saugatuck" sands
DG Dawson and Greenwood

peats

(MW)
(B)

376a

Quadrat Sizes

 

sygbol

53$‘P‘

size

1 m2
1/4 1112 2
1/100 m

[figoupings of Variables

 

 

SPECIES ENVIRONMENTAL
name symbol name

small solitary
clonal xer0phytic
margin dry
margin'wet

bog species

(tot)

(SY) synusia

(S-) synusia, excluding
major components

(LI) litter parameters

(HO) horizon thicknesses

total, all
variables

APPENDIX C

SPECIES INCLUDED IN GROUPINGS

(SS) - Small Solitary

:fé.

EIE’IE’I

w :> >
Islals

U1
0
'.‘

IEIEIEEI

UGO
slalzli

"U
H
H

G
O
G

E

ERIE

O

H

5E1

name

Amelanchier oblongifolia
Amelanchier sanguine;
Anemone quinquefolig
Arctostaphylos uva-ursi
Aster "hairy ovate"
Aster laevis

Rubus canadensis
Solidago "lanceolgtelong petiole"
Oryzopsis pungens
Oryzopsis agperifolia
Messages

Viola adunca

(CX) - CIOQQLXerophytic

 

name

Carex pedunculata

932W
Comptonia peregrine

Danthonia gpicata
WM

(MD) - Margin DEX

nmne

Enigaea repens
Gaultheria procumbens

Maianthemum canadense

Mel lineare

Vaccinium anggstifolium var. nigrum
vaccinium boreale

377

O

O
0C)
0
H

EIEIEIEEIEI

<:
H.
o‘

H

378

éMW) - Margin Wet

name

 

Cgptis groenlandica

Cornus caggdensis

Ilex verticillata
Lycgpodium obscurum
Nemopanthus mucronata

Rubus hispidus var. obovalis
lzgentalis borealis
Vaccinium myrtilloides
Viburnum‘gggsinoides

1B) - Bog Species
name

 

 

Chamaedaphne calyculata

Eriophorum gpissum

Kalmia polifolia

Kalmia angustifolia

Ledum groenlandicum

Polytrichum strictum

Sphagnum plumosum

Sphagnum spp., other thanԤ. plumosum
Vaccinium augustifolium var. nigrum
Vaccinium oxycoccus

Vaccinium boreale

 

_\1

“7'1?@ifl'fli'fl'IlflllmfllflilfiI