:10“, L1“);
I_,
(315‘:

1.11111F1
'2;

 
 

    
     
   
    
    

_.-—‘
_4— A

“O-"
u-.-

.
t ' '5 ‘_.--_::.‘='. A
_- ..._. ...
w
.

w

w
.‘o—Ml
..

u

  
 

'—

Dru...
.. .

$3

    

to 0-.

    
      
 
    
      
  
  
   

.s.

‘7...—
- ~v -c

own ,

an
--Aor—

‘0—

  

‘ I 4
v ‘1
. “.57.

 
 
 
 

. o.
I n J 4
l'-'.':h

- Iv
a.“
. v 1

v- A»
' . ."9 .
- A’ n
' 'u:3@‘
“gt—ox -
4 a p)? ‘
raw“

  

. 3::
10.311 "

  
  
     
     
  

‘11

.. . .‘ {111111111111 111“ 711.1.
1113' 111" 111111111111
1‘ :13". 4:31.111“? 3%?

4

   
   

    

.40.
.- ,
.r.'-v¢_-..

‘7...
1;..401.
I4 -.,..

.. ..

‘ a

.9
.1 o‘
..-

55.?

f:“"-. < ‘0’ -04}.
. no . .
.

  
   

-,“. I

...‘
‘»
'.u..
I...
- >01. .1
.

   
  
 
 
  
  
   
 
  
  
 
 

.
9—5:”...
a

 
 
  
 

  
   
     
 
 

   

 

  

‘7. I. : '.

'1’511‘ ”11131 ..
1

51-1, 1.: 1111-111: 1111' if 195;:th :1 I: .1,

1131:4151)? 511 v5
‘1.”

 
    

“1171111191 I

9‘ 11'3“":
~ oo’idzptzz

{1:

I.
.l
‘ #uYfil.
w".-.
M. ,
m
- ..

     
   
   
   

  

"‘«z 131,;
5,1; 11 o
- 13:1 111- I:
11‘1‘11'1111111111111111111

Itz':
1 “udlflfl‘gd’ 1am
I :1.» U- 11
1 «‘1 *3‘ g 1111521
.' 1111111111111-111111111'1
‘L ' - .H‘HI'Iz' [11‘1de 1'13" 3111111111 1111‘
11111 111111“ ‘111‘31'11‘1 1‘11 111111111 11111-11 .
11‘ .5 111.1 ‘1. 11 .1

.5111

0.. 1L
7-1!

  

pm
‘—

g—o-
u. ,
.

53‘ in! 10:1'. I!" ' .
11 11111111111 111111 1 "1141:1111: :1 g2.
11111111- I 11-1111 151 i111}: 1:; I: 51
”41 1 -
.n 1

  
 
    
 

   
   

   
 

‘11

   
    
 
  
   

”-
.o

 

:21";

¢
-
-
o

 
  
 

 

   
    
   

. V

r . . V
. , J

- ' '._'a....._..,

[1:11; {11111 I

  
   

      

11'

IUAI I,
0; 1

 

:
-
n
W “—4—
o.

trim- --

r-w‘

.1
‘7 1:._ 1‘ y
_. a}; f
; f
u I i"- _rl\ p‘fi
,1 . ‘ I'M" 9‘ '
I r- - j'l’t‘ 53' 'iv
Q. ~
.L. ' ”37;
i 3 w -.;.. r: as”
\‘. i
m“~'“~‘-
--—.___.. -~_ “‘ -‘
m

This is to certify that the

thesis entitled
A MORPHOMETRIC ANALYSIS OF THE BAT

SPECIES IN THE GENUS CAROLLIA
(MAMMAL IA : PHYLLOSTOMIDKE )

presented by
Laura John McLellan Moritz
has been accepted towards fulfillment

of the requirements for

M.S. degree in Biological Science

wag

 

 

Major professor

Date June 26, 1983

 

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

 

}V153I_J BEIURNING MATERIALS:I
Place in book drop to
untuuuss remove this checkout from
All-It’ll... ‘your record. FINES will

 

 

be charged if book is
returned after the date
stamped below.

 

 

 

 

 

 

~-..~ (“a ,1- g-v ’7'» ~— 4. i“ _. . ,1

WW" ESE mm

 

 

A MORPHOMETRIC ANALYSIS OF THE BAT SPECIES IN THE
GENUS CAROLLIA (MAMMALIA:PHYLLOSTOMIDAE)

BY

Laura John McLellan Moritz

A THESIS

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

MASTER OF SCIENCE

Department of Biological Science

1983

©1984

LAURA JOHN MC LELLAN MORITZ

All Rights Reserved

[34' 35200

ABSTRACT

A MORPHOMETRIC ANALYSIS OF THE BAT SPECIES IN THE
GENUS CAROLLIA (MAMMALIA:PHYLLOSTOMIDAE)

by

Laura John McLellan Moritz

The species of bats in the genus Carollia present a
complex pattern of morphological variation. This varia-
bility complicates Species identification. I examined sex-
ual and geographic variation in cranial and mandibular mea-
surements of 475 Carollia Specimens selected to represent
the distributional range of each species. The presently

accepted Species include: brevicauda, castanea,

 

perspicillata, and subrufa.

 

The Species were easily separated by canonical var-
iates analysis, with Q. castanea the most distinctive
species. Sexual dimorphism is present in all species and
males are consistently larger than females. Significant
differences between pOpulations are present in all Species.

Q. subrufa and g. brevicauda have morphologically distinct

 

pOpulations in the northern and southern portions of their
range, while different p0pu1ations of Q. castanea Show no
clear geographic pattern. 9. perspicillata pOpulations
form a continuum with nonoverlapping individuals from the

northernmost and southernmost portion of the range.

To my mother and father

ii

 

ACKNOWLEDGMENTS

I am especially grateful to Dr. D. O. Straney for his
patience and guidance on this thesis. I also thank Dr. T.
W. Porter and Dr. C. Cress for their counsel during the
course of this study. Dr. R. H. Pine was most helpful in
reviewing and commenting on this work. Learning to use
Michigan State University's computer facilities and the
statistical packages used in analyses was a pleasure,
thanks to Bruce Johnston. I thank Dr. H. Genoways, Dr. R.
Fisher, Dr. K. Koopman, Ms. Sheila Kortlucke, Dr. D. Patton,
Dr. J. D. Schmidly and Dr. R. Timm for permitting access to
Specimens under their care. Financial support was sup-
plied by The Museum and The Department of Zoology, Michigan

State University.

iii

 

TABLE OF CONTENTS

LIST OF TABLES . . . . . . .
LIST OF FIGURES.

INTRODUCTION .
METHODS AND MATERIALS.
RESULTS.
Variation Between Species.
Patterns of Variation Within Species
Sexual Dimorphism.
Locality Trends.
Phenetic Relationships Between POpulations
DISCUSSION .
CONCLUSIONS.

APPENDIX A .
APPENDIX B .

BIBLIOGRAPHY . . . . . . . . . . . . . .

iv

Page

vi

16
16
24
26
28
63
72
94

99
103

107

LIST OF TABLES

Percent<xfTotalVariation Attributable to
Species Differences. . . .

Mean Character Values (m i 2 Si) for those
Characters Useful in Species Discrimination.

Percent Contribution to Total Variance of
Locality, Sex and Residual for Each Species
of Carollia. . . . . . . . . . . . . .

Percent of the Total Variation Attributuble
to Locality, Sex and Residual from 4 P0pula-
tions of C. castanea . . . . . . .

Percent of the Total Variation Attributable
to Locality, Sex and Residual from 8 P0pula-
tions of g. subrufa.

Percent of the Total Variation Attributable
to Locality, Sex and Residual from 9 Popula-
tions of Q. brevicauda . .

 

Percent of the Total Variation Attributable
to Locality, Sex and Residual from 15 Popula-
tions of g. perspicillata. . . . . . .

 

Characters with High Residual Variance
Components

Page

17

19

25

29

37

45

56

76

10.

11.

12.

LIST OF FIGURES

Mexican and Central American localities from
which populations were sampled.

South American localities from which p0pula-
tion samples were examined.

Views of Carollia Skull and mandibular ramus
showing measurements used in study. .

Canonical variates analysis of Carollia species

Dice-Leras diagrams representing character
trends for basilar and palatar length across
latitudes for Q. castanea . .

Canonical variates analysis of four pOpula-
tions of Q. castanea. . . . . .

Dice-Leras diagrams representing character
trends in basilar and mandibular length across
latitude for Q. subrufa .

Canonical variates analysis of eight popula-
tions of C. subrufa .

Dice-Leras diagrams representing character
trends in basilar and palatar length across
latitudes for Q. brevicauda .

 

Dice-Leras diagrams representing character
trends in maxillary tooth row and dorsal
rostral length across latitude for Q.
brevicauda. -

 

Canonical variates analysis of nine pOpula-
tions of C. brevicauda.

 

Dice-Leras diagrams representing character
trends in mandibular and basilar length across
latitude for Q. perspicillata . . . . . .

 

vi

11
21

32

34

40

42

47

49

S3

S8

13.

14.

15.

16.

17.

vii

Canonical variates analysis of eight pOpula-
tions of Q. perSpicillata . . . . . . . .

 

Canonical variates analysis of C. perspicillata

 

populations sampled, illustrating only Mexican
and Central American populations. . . . . .

Canonical variates analysis of Q. persPicillata
populations sampled, illustrating only South
American p0pulations. . . . . . . . . . .

Dendogram of all 36 Species populations result-
ing from cluster analysis . . . . .

Diagram of Species affinities . . . . . . . .

64

66

69
92

 

INTRODUCTION

Bats of the genus Carollia (family Phyllostomidae)
are among the most common mammals in tropical America.
They range from northern Mexico to Paraguay living in most
habitats, except those at higher elevations and in very
arid regions. DeSpite their abundance, the taxonomy of
Carollia has been confused largely because of difficulty in
accurately delimiting species. The past difficulties in
determining species limits is reflected in the nomencla-
tural history of the genus.

Linnaeus (1758) first described a Carollia under the

name Vespertilio perspicillatus. In part because of the

 

ambiguity of his description, considerable nomenclatural
confusion existed for members of this genus until 1907,
when Hahn (1907) produced the first revision of the genus.
He recognized three species of Carollia under the generic

name Hemiderma: H. perspicillata with subspecies

 

 

perspicillatum and aztecum, H. subrufum Hahn and H.

 

castaneum H. Allan. Unfortunately, at that time only 374

 

Specimens were available for Hahn to study and these poorly
represented the distributional range of the genus. In
1924, Miller resurrected the generic name Carollia (pre-

viously thought to be a junior synonym of the fossil

pelcypod Carolia). The Species subrufum and castaneum were

 

considered conspecific by Felten (1956) who, following
Elliot (1904) presented a key based on measurements to

separate the two species (perspicillata and castanea) that

 

he recognized. Subsequently, there was a fair amount of
confusion over the number of Species of Carollia. It
proved particularly difficult to find distinguising char-
acters that consistently separated species over their

entire range. Hall and Kelson (1956; species perspicillata

 

and castanea recognized) noted that individuals of each
species were larger in the northern part of their geo-
graphic range than those from the southern portion. They
were unable to make a key that could distinguish the
Species they recognized from one another at all locations.
Pine (1972) presented a revision of the genus where

he recognized four species, adding Q. brevicauda to the

 

list of widely accepted species. He used classical methods
in his revision of the genus, examining a large number of
Specimens from throughout the range of each species, in
search of distinguishing characters. He was able to find a
combination of characters to distinguish four species.

These characters include cranial features, tooth morphology,
body Size and pigment distribution in hair shafts. Pine
provided a key to the species that is useful when both

skins and cleaned skulls are available and when the speci-
mens are from the same geographic region. Although Pine's

classical approach has proven useful at the species level,

 

intraSpecific variation between pOpulations has not been
clearly described. Pine noted that intraspecific varia-
tion existed, but was unable to quantify this variation
using a classical approach. Two intraspecific size trends
noted by Pine are: (1) individuals from the northern pOpu-
lations were larger than those from the south in all

species except 9. castanea and (2) male 9. perspicillata

 

tended to be larger than females. He further noted that
sympatric Species populations were easier to distinguish
from one another than were allOpatric populations of the

same species. Pine treated Q. perspicillata as polytypic

 

and suggested that Q. subrufa may also have more than one
form (but was uncertain of how many and what their taxo-
nomic Status should be).

I re-examined the genus Carollia extending Pine's
analysis to describe character variation within the four
species. Morphometric techniques were used on cranial and
mandibular measurements to evaluate individual, intrapopu-
lational and intraspecific variation. I examined the fol-
lowing points:

1. How distinctive, in a quantitative sense, are the

four nominate species rec0gnized by Pine?

2. How is the variation due to individuals, sex and

location apportioned?

3. How do the individual Species vary between sexes

and between populations?

4. Is the variation within each species concordant
across the Species?
5. What are the phenetic relationships of the pOpu-
lations within and between species?
This analysis will aid in identifying the patterns of geo-
graphic variation as well as sexual and individual varia-
tion within Carollia Species. Once the patterns are
described, hypotheses concerning factors responsible for

these patterns of variability can be addressed.

METHODS AND MATERIALS

I examined a total of 475 male and female Carollia
Skulls from locations selected to represent the distribu-
tional range of the genus (Appendix A). All of the speci-
mens were examined previously by Pine (1972) in his study
of the genus. I used only specimens judged adult by Pine
(1972), based on fusion of the epiphyses of the metacarples
and phalanges. Only populations from Pine's study with
more than two individuals were included in the analysis.
The average number of individuals examined per population
was 13. Specimens of C. castanea (n=64) were examined from
four localities extending from Honduras to Peru. 9.
subrufa (n=73) were examined from eight localities along
the Pacific versant of Central America, from Mexico to

Nicaragua. Nine populations of Q. brevicauda (n=123) are

 

represented, ranging from Mexico to Ecuador. 9.

perspicillata (n=215) is the best represented species with

 

specimens examined from 15 localities ranging from Mexico
to Paraguay (Figures 1 and 2). Pine left four specimens
unclassified to species. These individuals were from
separate localities in Nicaragua, British Guiana, Ecuador

and Peru. They were included in certain of the analyses

Figure 1. Mexican and Central American localities from
which pOpulations were sampled. Numbers corre-
spond to localities listed in Appendix A.
Vertical lists of symbols indicate sympatric
populations of several Species.

Figure 2. South American localities from which pOpulation
samples were examined. Numbers correSpond to
localities listed in Appendix A. Vertical
lists of symbols indicate sympatric pOpula-
tions.

 

 

DZl

  

O brevicauda

O castanea

0 Cl perspicillata
? unidentified

 

Figure 2.

 

10

to more accurately determine their phenetic relationships
with Carollia Species.

A total of 22 cranial and mandibular characters were
measured (Figure 3). Linear measurements were taken to the
nearest 0.1 mm using dial calipers and one angular measure-
ment was taken using a goniometer mounted on a dissecting
microsc0pe. In order to measure the slope of the forehead,
the skull was set on its side in a clay block on top of a
piece of graph paper. The upper edge of the maxillary
tooth row was aligned with a line on the graph paper and
used as a base line. The crosshair of the goniometer was
aligned perpendicular to the tooth row and passing through
the inflection point between the cranium and the rostrum.
The angular measure was taken between this perpendicular
crosshair and the forehead. Linear characters include:
basilar length (BAL), palatar length (PL), postglenoid
width (PGW), breadth of braincase (BB), depth of braincase
(DB), least interorbital breadth (LIB), rostral breadth
(RB), length of maxillary spur (MSL), length of maxillary
tooth row (MTR), width between first molars (MW), width
between second pre-molars (PMW), width between canines
(CW), dorsal rostral length (DRL), ventral rostral length
CVRIJ, palatal width (PW), foramen magnum width (FMW),
lmandibular length (ML), mandibular depth (MD), coronoid-
angular distance (CAD), height of coronoid (CH), and post-

dentary ramus length (PDRL).

Figure 3.

11

Views of Carollia skull and mandibular ramus
Showing measurements used in study: 1. basilar
length (BAL); 2. palatar length (PL); 3. post-
glenoid width (PGW); 4. breadth of brain case
(BB); 5. depth of brain case (DB); 6. least
interorbital breadth (LIB); 7. rostral breadth
(RB); 8. maxillary Spur length (MSL);

9. maxillary tooth row (MTR); 10. width between
first molars (MW); 11. width between second
premolars (PMW); 12. width between canines
(CW); 13. dorsal rostral length (DRL); 14. ven-
tral rostral length (VRL); 15. palatar width
(PW); 16. foramen magnum width (PMW);

17. mandibular length (ML); 18. mandibular
depth (MD); 19. coronoid angular distance
(CAD); 20. coronoid height (CH); 21. post-
dentary ramus length (PDRL); 22. slope of fore-
head (SF).

12

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.

13

Statistical analysis of the data was performed at
Michigan State University using the Statistical Package for
the SocialSciences (SPSS; Nei gt 31., 1975) and Clustan
(Cluster Analysis Package; Wishart, 1978). The mean squares
for differences due to Species and due to populations and
sexes within Species are calculated in two separate nested
one-way analyses of variance using SPSS MANOVA; from these
the magnitude of the respective variance components are
estimated. The variance components themselves are not com-
parable over characters or taxa because of varying sample
sizes and unequal variances, but the magnitude relative to
the total variance within each character can be compared.
The relative magnitude of these variance components was
used to determine the importance of sources of variation
(Straney, 1978; Leamy, 1982).

Characters that vary across localities were examined
for latitudinal trends by correlating p0pulation character
means with geographic coordinates for each Species. Signi-
ficant differences between character means of the pOpula—
tions within each species was determined using a one-way
analysis of variance and Scheffe's multiple range test
(P=.05). A univariate F test was used for testing for
differences between characters for the sexes within each
species.

Discriminant analysis was used to assess the distinc-
tiveness of species, pOpulations within Species and the

sexes within species. Classical discriminant analysis

14

produces a linear function of the characters for each group
producing scores that can be used to assign unknown speci-
mens to a priori groups. Closely related to discriminant
function analysis is canonical variates analysis, both of
which were obtained from the same program, SPSS DISCRIMI-
NANT. Canonical variates analysis maximally separates a
priori groups by maximizing the between group variance
relative to the within group variance of the characters
used in the analysis. This is accomplished by giving char-
acters which vary according to group (species, population
or Sex) high weightings and those that vary independently
low weightings. This procedure corrects for correlated
characters and adjusts for differing covariance between
groups compared. The first canonical variates axis always
explains the greatest amount of the character variance
between groups. The second axis explains the next greatest
amount of variance of an axis orthogonal to the first axis
and so on for each additional axis. On each axis, the
individual characters are given a value (standardized
canonical variates coefficient) which corresponds to their
importance in separating groups; the larger the absolute
value, the more important the character.

Mahalanobis distance was used to place the four
unknown Specimens (left by Pine, 1972) to species.
Mahalanobis distance is the Euclidean distance squared in

the space defined by the canonical variates. Each specimen

15

is assigned to the species sample with the closest group
centroid (mean character vector).

The relationship between populations was summarized
using cluster analysis (unweighted pair group method).
This analysis produces a dendrogram which graphically
represents similarity between pOpulations in a branching
diagram. The squared Euclidean distance between popula-
tion centroids was used to calculate the similarity matrix.

Populations of all species are entered together.

RESULTS

Variation Between_§pecies

 

Table 1 presents the relative proportion of variation
between and within Species for each of the 22 characters
examined. Characters with large variance components attri-
buted to Species differences are potentially useful char-
acters from discriminating between Species. Differences
between species accounts for only 57.5% of the average
character variance leaving on average 42.5% of the variance
attributable to within Species differences. The best char-
acters for distinguishing between species appear to be
basilar length, maxillary tooth row length, ventral rostral
length, coronoid-angular distance, mandibular length, and
coronoid height. Each of these characters differ signi-
ficantly between species, determined by a one-way analysis
of variance and Sheffe's multiple range test. For each
character, the mean measurement is greatest for Q.

perspicillata followed in decreasing order by g.

 

brevicauda, Q. subrufa and g. castanea (Table 2).

 

 

There are six characters that vary more within
species than between species. They are least interorbital
width, maxillary spur length, distance between the first

molars, foramen magnum width, mandibular depth and slope

16

17

TABLE 1

Percent of Total Variation Attributable to Species
Differences. Total Variation is the Character
Variance Over All Species.

 

% Variation

 

 

Total
Variable Variance
Between Within
Species Species

Basilar length 78.4 21.6 150.18
Palatar length 58.8 41.2 42.57
Post-glenoid width 63.2 36.8 21.03
Breadth of brain case 56.7 43.4 12.42
Depth of brain case 60.9 39.1 20.45
Least interorbital
breadth 18.6 81.4 3.85
Rostral breadth 55.5 44.5 7.50
Maxillary spur length 36.5 63.5 7.45
Maxillary tooth row 82.8 17.2 36.73
Width between first
molars 43.4 56.6 6.94
Width between second
premolars 71.1 28.3 17.98
Width between canines 60.9 39.1 7.08
Dorsal rostral length 64.8 35.2 25.84
Ventral rostral
length 79.4 20.6 27.11
Palatal width 57.1 42.9 3.50
Foramen magnum width 15.8 84.2 3.65
Mandibular length 78.3 21.7 95.64

Mandibular depth 40.2 59.8 6.01

18

Table 1 continued

 

 

 

Total
Variable Variance

Between Within

Species Species
Coronoid angular
distance 79.1 20.9 42.29
Coronoid height 78.3 21.7 38.88
Post-dentary ramus
length 72.3 27.7 31.08
SIOpe of forehead 11.4 88.6 11.47

 

Mean 57.48 42.55

 

19

 

mm.
mm.
me.
mm.
ow.
mm.

-H

+I

+i

+|

+l

+l

m.wm
m.mNH
o.mv
m.Hm
w.No
N.mmH

No.
he.
om.
mm.
mm.

+|

+I

+i

+|

+|

+l

m.v¢
m.mmH
w.om
n.0m
m.nc
m.no~

me.
ow.
me.
me.
me.
mm.

+l

+I

+I

+|

+l

+l

o.mv
m.me
m.mm
w.wm
m.mo
m.©na

om.
05.
am.
mm.
Ne.
ow.

+l

+l

+l

+l

+|

+|

m.Hm
H.¢¢H
o.nm
o.mo
m.mm
a.HwH

EU
4:
90¢
qm>
MHZ

A<m

 

wocmwmmo .u

«WSHQSm .o

 

mpswofi>oun .w

 

mpwHHfiomeHom .w

 

:oflpmcflefipomfla mofioomm ca Hamom:
meouumemcu omonp ppmfimm NM :83 mosfim>.uoaomem:u new:

N mqm<9

20

of forehead. These characters vary in accordance with
other factors.

The four species of Carollia described by Pine (1972)
were easily separated in canonical variates analysis (Fig-
ure 4) using all 22 characters. Over all Species, 98.67%
of the Specimens were correctly assigned to Species using
three canonical axes. C. castanea is the most distinctive
species examined, with 100% of the Specimens correctly
classified. 9. subrufa individuals group tightly with 100%
of the Specimens correctly classified. 9. subrufa falls

closer (using Mahalonobis distance) to Q. brevicauda and Q.

 

perSpicillata than to E. castanea. Carollia brevicauda and

 

 

C. perSpicillata overlap slightly with 3.4% of the Q.

 

brevicauda classified as C. perSpicillata and 1% of the C.

 

 

perSpicillata classified as Q. brevicauda.

 

 

Species are maximally separated along the first canon-

ical variates axis with E. subrufa and g. brevicauda fall-

 

ing between C. castanea and Q. perSpicillata. Characters

 

with high loadings on this axis are maxillary tooth row
(-.56) width between first molars (.42) and palatar length

(.49). The second canonical axis separates g. brevicauda

 

from the rest of the Species. Characters important here
are basilar length (1.14), maxillary tooth row (1.77) and
post-dentary ramus length (.49). The third axis separates

g. subrufa, g. castanea and Q. brevicauda. Important in

 

this separation are maxillary tooth row (~1.12) and maxil-

lary spur length (-.66). Maxillary tooth row is

Figure 4.

21

Canonical variates analysis of Carollia species.
Each species is represented by a tracing of the
perimeter of the cluster formed by individual
specimens included in the analysis. Group cen-
troid for each species is represented by a
solid dot. The x and y axes are the first and
second canonical axes respectively; the third
axis value is printed below the group centroid.
Units of these axes represent the canonical
variates scores. Areas of overlap represent
Specimens which fall into species groups other
than the a priori assigned group.

 

 

 

23

consistently important in Species separation. The nature
of some of these variables (palatar length, maxillary tooth
row length and basilar length) suggests skull length is a
major factor in Species separation using canonical variates
analysis.

Size appears to be an important factor in discriminat-

ing between Carollia species. The first canonical axis

 

accounts for 71.59% of the total variation between species.
Low values correspond with the largest individuals in the

study (specimens of E. perspicillata), while the high

 

values are associated with the smallest Specimens (g.

 

castanea). Specimens of g. subrufa and Q. brevicauda are
not separated on the first axis, but are on the second.

The second axis separates Q. brevicauda and g. subrufa from

 

one another and the third axis fully separates Q. subrufa

g. brevicauda, and Q. perspicillata.

 

 

The specimens left unidentified by Pine (1972) were
classified to species using discriminant analysis.
Unknowns were placed to the Species with the nearest group
centroid (least Mahalonobis distance). The specimens from
Yalaguina, Nicaragua and Napo Pastaza, Ecuador were placed

with C. brevicauda. The specimen from Kartobo, British

 

Guiana was placed with g. subrufa and the specimen from

San Juan, Peru was placed with 9. perSpicillata. All four

 

of these Specimens fell close to the region where C.

subrufa, C. brevicauda and g. perSpicillata individuals

 

 

come into close proximity in canonical variates Space.

24

Patterns of Variation Within Species

The relative importance of the sources of variation
within species was examined in a two level nested analysis
of variance estimating variance components due to (1) dif-
ferences between localities, (2) differences between sexes
within localities and (3) residual variation. Expressed as
a percentage of the total variation within each species and
averaged over 22 characters, most of the variation is resi-
dual (58.6-77.9%) and between localities (l6.0-29.3%) in
all four species (Table 3). The contribution due to sex is
the smallest source of variation examined (4.1-6.9%) in
agreement with the results of Leamy (1983) using laboratory
mice and Straney (1978) using ocelots, skunks and wild mice
to measure variance components.

The pattern of character variation within each
Species varies between Species (Appendix B). All Carollia
have a large locality component in basilar and ventral
rostral lengths (>30%). Both 9. castanea and Q.

perSpicillata have locality effects in all 22 characters.

 

Rostral breadth is sexually dimorphic in all four species.
The distribution of variance components due to sex are

similar in Q. subrufa and g. brevicauda. Both have char-

 

acters with large sex components independent of locality
effects. Three of these characters are common to both
Species including depth of braincase, rostral breadth, and

mandibular depth.

 

25

 

 

 

 

 

 

No.mo am.m Ho.mN new:
mN.HAN em.mm mw.e oo.¢m weaHHLUHamemm .m
mm.moe om.~a ow.e we.o~ meaaoe>men .w
om.ww mm.ea oa.e No.0H «magnum .m
40.0HH we.oo no.4 mN.mN «weapmau .m
Hench Hmsvflmom xom xuflamooq mofiooam

 

.nmeouomhwno NN eo>o wommp0>< mew mosfim>v
mflflaoumu mo mofloomm comm How Hmswfimmm paw xom

.xuwflmooq mo mocmfiem> HmQOH ow cofiusnfieucou ucooeom

m mqm<H

26

Sexual Dimorphism

 

Sexual dimorphism is a relatively minor source of
variation in all Carollia Species, ranging from 4.07% of
the total variation within Species for Q. castanea to 6.8%

for g. perSpicillata. Pine (1972) saw no average differ-

 

 

ence between the sexes except possibly in Q. perSpicillata
where males were larger than females in average measure-
ments. This is considered in more detail in order to
describe which characters differ between sexes. All
Species are considered because variance component analysis
indicates that there are some sex effects in each Species
and these have not been previously described.

Variance component analysis across characters for
each Species reveals differences in the distribution of
dimorphism between the 22 characters. 9. castanea shows
very little sexual dimorphism. Only 8 characters have a
portion of their total variation due to sex; palatar length
(sex component=17.l9%), coronoid angular (23.19%) and
mandibular depth (18.26%) are most dimorphic. In contrast,

Q. perSpicillata has a larger amount of character vari-

 

ance attributable to sex distributed throughout 16 char-
acters; rostral breadth (36.65%), depth of brain case
(19.4%) and mandibular depth (18.48%) Show the greatest

sex effects. g. brevicauda has 17 characters with sex

 

variance components; rostral breadth (24.31%), coronoid
angular distance (20.89%) and mandibular depth (21.26%)

have a relatively large sex component. 9. subrufa has

 

27

only 13 characters with sex effects, but several with

large sex components including: depth of brain case
(35.65%), rostral breadth (21.88%), mandibular depth
(29.93%) and $10pe of forehead (21.68%). All four species
show dimorphism in mandibular depth and all but 9. castanea
have a relatively large effect on rostral width attribut-
able to sexual dimorphism.

Discriminant analysis between males and females,
within Species, over localities reveals a greater distinc-
tion between the sexes than noticed by Pine (1972). The
highest percent of males and females correctly assigned to
sex were specimens of Q. subrufa with 94.37% correctly

classified. 9. brevicauda had 88.14% and Q. castanea had

 

80.33% correctly classified to Species. Lowest is Q.

perSpicillata with 79.4% classified correctly. Although Q.

 

perSpicillata has a relatively large prOportion of vari-

 

ability attributed to sex differences, discrimination is
difficult due to the distribution of this variation among
many characters that also vary with locality. Q. subrufa
has a much more restricted range with few characters with
sex effects, but four have large sex components. This may
in part explain the better discrimination between the sexes
than seen in the other Species.

The standard coefficents for the canonical variates
separating male and female 9. castanea indicates that
variation between sexes relates largely to basilar and

mandibular length, which are larger in males.

28

Coefficients for Q. subrufa identify depth of brain case as
the primary variate for separating sexes. Males have a
significantly deeper brain case than females. 9.

brevicauda has several characters with high loadings includ-

 

ing: rostral breadth, dorsal and ventral rostral lengths.
Only rostral breadth differs significantly between sexes.

Males have wider rostrum than females. Rostral breadth is
also the most important character in distinguising between

male and female 9. perSpicillata.

 

The identification of characters which differ between
Carollia sexes differs depending on the approach. The
variance component analysis is indicating which characters
vary in accordance to sex, given as a proportion of the
total variance. Characters with a small variance that only
vary according to sex will have a large sex component even
though the difference between males and females may not be
great. The canonical variates analysis identifies char-
acters which can be used in discrimination. These char-
acters have a greater magnitude of difference between
sexes, even though the proportion of variation due to sex

may not appear as large.

Locality Trends

The four populations of Q. castanea examined Show a
complex pattern of character variation using variance-com-
ponent analysis. All characters but one (coronoid angular
distance) have some portion of their variability associated

with location (Table 4). An average of 29.25% of all

29

TABLE 4

Percent of the Total Variation Attributable to Locality,
Sex and Residual from 4 Populations of C. castanea.
Total Variation is the Character Variance Over
Individuals, Sex and Locality.

 

 

 

Character Locality Sex Within Total
Basilar length 49.43 8.76 41.81 19.30
Palatar length 19.35 17.19 19.35 11.80
Postglenoid width 45.75 0.09 54.26 4.28
Breadth of brain case 18.18 0.0 81.82 3.23
Depth of brain case 6.79 0.0 93.39 3.48
Least interorbital
breadth 19.06 0.0 80.95 2.60
Rostral breadth 18.79 9.68 71.34 2.24
Maxillary spur length 2.72 0.0 97.22 4.71
Maxillary tooth row 26.81 0.0 73.24 2.99
Width between first
molars 23.76 0.0 76.75 2.28
Width between second
premolars 38.33 0.0 61.64 4.80
Width between canines 43.22 0.0 56.77 2.02
Dorsal rostral length 19.43 0.0 80.53 6.23
Ventral rostral
length 64.21 0.0 35.79 6.76
Palatal width 9.76 0.0 90.24 1.17
Foramen magnum width 17.57 1.41 80.98 3.05
Mandibular length 41.11 0.0 58.89 11.02
Mandibular depth 27.20 18.26 54.55 1.98

Coronoid angular
distance 0.0 23.19 76.80 3.38

Table 4 continued

30

 

 

 

Character Locality Sex Within Total
Coronoid height 35.78 0.0 62.43 5.75
Post-dentary ramus
length 21.14 4.25 74.68 3.95
SlOpe of forehead 3.75 5.93 90.37 9.60

 

4F.

 

31

variability within the 22 characters is attributed to dif-
ferences in collecting locality. The characters with the
largest locality effect (over 40% of the total) are post-
glenoid width, width between canines, ventral rostral
length, basilar and mandible lengths.

Character trends were found to correSpond with lati-
tude. The southernmost population from Peru has the small-
est character values in all but two characters (dorsal
rostral length and foramen magnum width). Scheffe's
multiple range test identifies 5 characters which Show
significant differences between the Peruvian population and
the other more northern p0pulations. These characters are
basilar length, palatar length, coronoid height, post-
dentary ramus length, and lepe of forehead. The formen
magnum width is largest in the Peruvian population sample
composed of the smallest individuals. Figure 5 illustrates
the pattern of geographic variation in basilar and palatar
length, and is representative of variation in the other 15
characters Showing significant differences across pOpula-
tions with the exception of the formen magnum width.

The distinctiveness of populations was examined using
canonical variates analysis. The degree of differentiation
is high between g. castanea populations (Figure 6). Class-
ification results placed specimens into their respective
populations with 98.4% accuracy, the highest of any of the
species p0pulations of Carollia. The first canonical axis

separates the pOpulations from Costa Rica and Peru from the

Figure 5.

32

Dice-Leras diagrams representing character
trends for basilar and palatar length across

latitudes (corresponding to

collecting locali-

ties) for C. castanea. Vertical lines show

observed ranges; rectangles
deviation; horizontal lines
for the population sample.

along the x axis in degrees,

is given for north latitude
south latitude. The y axis
for character values.

mark standard
represent the mean
Latitude is given
a positive value
and a negative for
indicates scale

95

9O

85

80

95

9O

85

80

33

 

 

 

 

 

 

 

 

 

,1;
15.0 l0.0 7.0 5.0 40.0
latitude
Basilar length
«F -T
11
II
15.0 10.0 7.0 5.0 -l0.0
latitude

Palatar length

Figure 5.

34

Figure 6. Canonical variates analysis of four populations
of C. castanea. Populations examined: 1. Rio
Coc3, Honduras; 2. Palamar, Costa Rica;
3. Tucarcuna Village Camp, Panama; 4. Huanuco,
Peru. All three canonical axes are represented,
one and two on the x and y axis reSpectively.
The third axis is diagramed in the insert.
Group centroids are represented by solid dots.

 

 

 

 

 

 

 

 

Ilo

36

population from Panama. The population samples from Costa
Rica and Peru appear more similar to one another than they
are to the p0pulation from Panama, which is curious. Stan-
dardized canonical coefficients for the first canonical
axis indicates that variation in the width between the
second premolars and breadth of brain case are most impor-
tant in this separation. 0n the second canonical axis the
population from Honduras is separated from the other p0pu-
lations due to differences in the rostral breadth and width
between canines. The third axis separates p0pu1ations from
Peru and Costa Rica; post-glenoid width and coronoid height
are important in this separation.

The eight populations of g. subrufa examined had
16.02% of their total variation due to locality differ-
ences. Variance component analysis indicates that all but
four characters vary with locality (Table 5). Characters
which vary most with location include basilar length,
palatar length, ventral rostral length and foramen magnum
width (all greater than 32% of total variation). Multiple
range test results show only four characters which differ
significantly between p0pu1ations. The characters are:
basilar length, palatar length, width between first molars
and foramen magnum width. The two southernmost p0pu1ations
have significantly smaller values for these characters.

The two southernmost populations from Sabana Grande,
Honduras and San Antonio, Nicaragua have the smallest char-

acter means. Geographic trends in basilar and mandibular

37

TABLE 5

Percent of the Total Variation Attributable to Locality,
Sex and Residual from 8 Populations of Q. subrufa.
Total Variation is the Character Variance Over
Individuals, Sex and Locality.

 

 

 

Character Locality Sex Within Total
Basilar length 47.69 0.0 52.31 18.17
Palatar length 42.38 0.0 57.62 9.19
Post-glenoid width 28.65 1.31 70.02 3.17
Breadth of brain case 16.63 9.63 73.75 2.91
Depth of brain case 0.0 35.65 64.43 4.90
Least interorbital
breadth 8.69 2.37 88.93 2.27
Rostral breadth 0.0 21.88 79.11 1.74
Maxillary Spur length 7.43 0.84 91.18 3.10
Maxillary tooth row 24.50 0.0 75.61 2.71
Width between first
molars 14.73 0.0 85.37 1.35
Width between second
premolars 1.62 0.0 98.58 2.10
Width between canines 14.85 1.51 83.49 1.66
Dorsal rostral length 19.23 0.29 80.48 4.85
Ventral rostral
length 32.71 7.25 60.03 2.29
Palatal width 13.63 0.0 86.47 0.82
Foramen magnum width 33.19 0.0 65.54 2.04
Mandibular length 21.54 0.0 78.45 7.24
Mandibular depth 0.0 20.93 79.34 1.90

Coronoid angular
distance 1.61 3.41 94.96 3.66

Table 5 continued

38

 

 

Character Locality Sex Within Total
Coronoid height 1.76 5.95 92.33 2.15
Post-dentary ramus
length 22.42 0.0 77.59 3.03
Slope of forehead 0.0 21.68 78.32 7.12

 

39

length are illustrated (Figure 7). These characters were
chosen to represent Skull and mandibular length across
localities. The difference in size between northern and
southern populations was also noticed by Pine (1972).
These results Show a stepped cline in character values.

P0pulation samples of Q. subrufa form two distinct
subgroups using canonical variates analysis (Figure 8).
This separation occurs along the first canonical axis with
40.19% of the variance explained. The six p0pu1ations
from Valle, Honduras and north into Mexico form one large
overlapping group, while individuals from Sabana Grande,
Honduras and San Antonio, Nicaragua are clearly distinct.
In this separation, basilar length and Slope of forehead
are most important. The second canonical axis further
separates p0pu1ations accounting for 20.52% of the total
variance. On this axis p0pu1ations from San Antonio,
Nicaragua and Sabana Grande, Honduras are separated from
each other and the p0pu1ations from El Salvador, Guatemala
and Nicaragua are separated from the northernmost p0pu1a-
tions from Mexico. Characters with high loadings on this
axis are breadth of brain case, width between canines and
width between second premolars. The third axis separates
the p0pu1ation from Valle, Honduras from the other p0pu1a-
tions with maxillary tooth row contributing most to group
distinction.

Classification results between populations placed

100% of the individuals from El Salvador, Nicaragua and

Figure 7.

40

Dice-Leras diagrams representing character
trends in basilar and'mandibular length across
latitude (corresponding to collecting locali-
ties) for C. subrufa. Vertical lines Show
observed ranges; rectangles mark standard
deviation; horizontal lines represent the mean
for the population sample. Latitude is given
along the x axis in degrees, a positive value
is given for north latitude and a negative for
south latitude. The y axis indicates scale for
character values.

41

.n oesmwm

foe: 5.33.1:02
evaz.0a
00.: 00.2 cog...— 00.0—
4. _ on.

 

 

 

E .L-

- E
a
e -

Eh
L44
#1

 

 

 

 

 

 

faced .0225

 

 

 

£553 .
coa— ooa. ooh. oo 2
m2
; _ ._1 of
mo.

 

a

e - E

 

 

 

Figure 8.

42

Canonical variates analysis of eight populations
of Q. subrufa sampled. Populations included
are: l. Tapanetepec, Mexico; 2. Arriaga,
Mexico; 3. Tapachula, Mexico; 4. Chiquimula,
Guatemala; 5. La Libertad, El Salvador;

6. Valle, Honduras; 7. Sabana Grande, Honduras;
8. San Antonio, Nicaragua. The first and second
canonical axes are represented by the x and y
axes respectively. The third axis is in the
insert. Units on these axes represent canonical
variates scores. Group centroids are indicated
by solid dots.

44

Honduras into their reSpective population locality. These
populations are well differentiated from one another. The
p0pu1ations from Mexico and southern Guatemala are not as
distinctive; only 79% of the individuals were placed into
their correct p0pu1ation locality.

The nine populations of g. brevicauda sampled had an

 

average of 20.18% of the total variation due to locality
differences. Variance component analysis of each of the
22 characters reveals that 16 characters have a locality
effect (Table 6). Eight characters have a large (>26%)

contribution to the total variance. These include the

basilar length, palatar length, maxillary tooth row length,

width between first premolars, dorsal rostral length, ven-

tral rostral length, coronoid height and post-dentary ramus

length. Multiple range test results Show that 11 of the
characters differ significantly between localities. Char-
acters that are indicative of skull length (basilar, pala-
tar, maxillary tooth row lengths, etc.) all show signifi-
cantly smaller measurements from individuals taken from
Ecuador than those taken from more northern p0pu1ations.
The population from Panama has character values which are
significantly larger than those in the sample from Ecuador
and the more northern populations.

Geographic trends in character values are illustrated
for basilar, palatar, rostral and maxillary tooth row
lengths (Figures 9 and 10). These characters were chosen

to represent the general patterns seen in the 22 measurements

 

I

4
I

I
I .'..
“k
." .

45

TABLE 6

Percent of the Total Variation Attributable to Locality,

Sex and Residual from 9 Populations of C. brevicauda.

 

Total Variation is the Character VaFiance Oveff
Individuals, Sex and Locality.

 

 

Character Locality Sex Within Total
Basilar length 44.51 2.90 52.79 30.67
Palatar length 42.72 0.0 57.28 18.45
Postglenoid width 11.76 8.29 79.96 6.50
Breadth of brain case 0.0 10.58 89.56 4.66
Depth of brain case 0.0 17.81 82.36 6.44
Least interorbital
breadth 23.28 4.39 72.20 2.87
Rostral breadth 0.0 24.31 75.77 2.25
Maxillary Spur length 0.0 12.30 87.70 3.68
Maxillary tooth row 38.11 0.0 61.88 5.95
Width between first
molars 20.34 1.43 78.23 3.29
Width between second
premolars 40.82 0.74 58.39 4.04
Width between canines 15.73 2.56 81.70 2.07
Dorsal rostral length. 46.69 0.30 52.99 10.09
Ventral rostral
length 37.09 1.27 61.73 5.50
Palatal width 17.18 0.0 82.20 1.56
Foramen magnum width 7.09 4.38 87.72 2.11
Mandibular length 23.54 7.78 68.68 19.03
Mandibular depth 0.53 21.26 78.24 2.10
Coronoid angular
distance 0.0 20.89 79.11 5.70

Table 6 continued

46

 

 

Character Locality Sex Within Total
Coronoid height 44.79 9.66 45.53 8.99
Post-dentary ramus
length 26.64 0.0 73.35 7.31
Slope of forehead 2.09 0.0 97.91 10.30

 

Figure 9.

47

Dice-Leras diagrams representing character
trends in basilar and palatar length across

latitudes (corresponding to

collecting locali-

ties) for C. brevicauda. Vertical lines Show

 

observed ranges; rectangles
deviation; horizontal lines
for the population sample.

along the x axis in degrees,

is given for north latitude
south latitude. The y axis
character values.

mark standard
represent the mean
Latitude is given

a positive value
and a negative for
indicates scale for

miiL

 

 

 

 

face.— .o.o_on_ .m OHDMHR
o .353
n O— n— Op ON
4.
tr
M faced .3300
.1323
o m o. 2 2 cu

 

 

 

F—'—l

 

 

e.

 

00

no

00—

mo—

0:

n:

no—
Ou—
wk—

on—
na—

00—

no.

Figure 10.

49

Dice-Leras diagrams representing character
trends in maxillary tooth row and dorsal ros-
tral length across latitude (corresponding to
collecting localities) for C. brevicauda.
Vertical lines Show observed ranges; rectangles
mark standard deviation; horizontal lines
represent the mean for the p0pu1ation sample.
Dots represent population sample with equal
character values. Latitude is given along the
x axis in degrees, a positive value is given
for north latitude and a negative for south
latitude. The y axis indicates scale for
character values.

 

50

£93.. _oZuo¢ .0300

0122.5
0 v a N- “—

.OH oesmSm

@—

 

 

 

 

 

 

 

 

 

 

 

le: foo» >3...on

6223
o v a a. n.

 

 

 

m—

 

 

 

 

 

-fi
mm— L. A

 

 

L: I.

L .. L

 

 

 

a

 

ON

mm

on

no

no

Oh

mu

on

51

examined. There appears to be a clinal trend in palatar
and dorsal rostral lengths, but a bimodal trend in the
basilar and maxillary tooth row lengths from the same popu-
lations. Palatar and dorsal rostral lengths are decreasing
in length from north to south, while basilar and maxillary
tooth row lengths are small in Nicaragua and increasing in
length in Panama becoming small again in Ecuador. The
population from Panama has shorter palatar and dorsal
rostral lengths and larger basilar and maxillary tooth row

lengths, thus becoming more like 9. perSpicillata. Pine

 

(1972) noted in his treatment of the species that Q.

brevicauda had a relatively long rear extension of the

 

palate. This p0pu1ation from Tacarcuna Village Camp,
Panama appeared odd to Pine (1972). It has pelage charac-
teristics resembling Q. subrufa; short, sparse, coarse,
indistinct, basal banding. He suggested the possibility

of hybridization between 9. subrufa and C. brevicauda.

 

The greater length of the skull and shorter palate seen in

the g. brevicauda sampled in this study suggests possible

 

hybridication with g. perSpicillata.

 

Multiple range testing of characters identifies char-
acters from the Tacarcuna Village Camp, Panama p0pu1ation
which are significantly larger than those of the northern
p0pu1ations; these are width between the second molars,
palatal width and maxillary length. These character dif-
ferences indicate that there are shape differences not only

Size differences between this population from Panama and

52

the other C. brevicauda populations examined. The palatal

 

width, not only its length, is greater.

Distinctiveness of the nine p0pu1ations was examined
using canonical variates analysis (Figure 11). A northern
and southern grouping is formed along the first canonical
axis, accounting for 47.87% of the total variation between
populations. The northern populations include individuals
from Mexico south to Costa Rica. The southern two popula-
tions are from Panama and Ecuador. Characters with high
loadings include width between the second premolars and
basilar length. P0pulations from Panama and Ecuador are
separated from one another by the second canonical axis,
which accountsiku~20.95% of the total variance. Characters
with high loadings are width between second premolars and
width between canines. The third canonical axis further
separates the northern p0pulations of Teapa, Mexico;
Talanga, Honduras and Cariblanco, Panama from the other
populations sampled. This axis explains 14.81% of the
total variation. Characters with high loadings include
basilar length and width between canines.

Classification results place 79.66% of the individuals
in their correct population locality. Individuals from
Ecuador are placed correctly 100% of the time, but only
three individuals are examined. Those from Panama are next
most distinctive with 92.7% correctly placed. The p0pu1a-
tions from Guatemala and Honduras are not as distinctive.

Individuals from Guatemala are difficult to place and only

Figure 11.

53

Canonical variates analysis of nine p0pu1ations
of g. brevicauda sampled. P0pulations include:
1. Rio Quezalapam, Mexico; 2. Teapa, Mexico;
3. Puerto Barrios, Guatemala; 4. Danli,
Honduras; 5. Talanara, Honduras; 6. Yala-
guina, Nicaragua; 7. Alajuela Cariblanco,
Costa Rica; 8. Tacarcuna Village Camp, Panama;
9. Puyo, Ecuador. The x and y axes represent
the first and second discriminant axes. The
third axis is in the insert. Units on these
axes represent canonical variates scores.

Group centroids are represented by solid dots.

 

 

 

 

 

 

 

 

 

 

55

58% fall into the actual p0pu1ation sample. The other 42%
could just as easily be placed with the sample from Teapa,
Mexico or Danli, Honduras. Those from Honduras are placed
correctly 53% of the time. The other 47% are assigned to
populations from the more northern localities.

The fifteen populations of Q. perspicillata sampled

 

had 34.6% of their total variation attributable to local-
ity differences. Variance component analysis discloses
locality effects in all 22 characters examined (Table 7).
Four of the characters have more than 50% of their varia-
tion due to locality effects; these include basilar length,
post-glenoid width, maxillary tooth row, and post-dentary
ramus length. Multiple range testing identifies 16 char-
acters that differ Significantly between localities. Char-
acters which do not differ significantly between localities
include least interorbital width, maxillary spur length,
width between second premolars, palatar width, foramen
magnum width and slope of forehead.

Geographic trends in skull length for g. perSpicillata
is illustrated by correlating basilar and mandibular length
with latitude (Figure 12). The character trends are
bimodal with large measurements from individuals from
Mexico and Honduras as well as those from Bolivia, ruling
out a smooth cline in character values corresponding with
latitude. The populations from Paraguay and Brazil do,
however, have character values Significantly smaller than

most of the more northern populations examined.

56

TABLE 7

Percent of the Total Variation Attributable to Locality,
Sex and Residual from 15 P0pulations of Q. perSpicillata.
The Variation is the Character Variance Over

Individuals, Sex and Locality.

 

 

 

Character Locality Sex Within Total
Basilar length 66.79 0.99 32.22 50.75
Palatar length 46.41 0.0 53.59 24.63
Postglenoid width 55.77 0.0 44.23 11.99
Breadth of brain case 20.35 9.22 70.42 7.75
Depth of brain case 13.92 19.40 66.68 12.00
Least interorbital
breadth 11.76 4.38 83.86 3.97
Rostral breadth 3.56 36.65 64.45 5.79
Maxillary Spur length 23.87 7.17 75.19 6.27
Maxillary tooth row 53.66 4.29 42.05 10.04
Width between first
molars 36.43 0.0 63.60 5.96
Width between second
premolars 22.07 5.24 72.69 7.69
Width between canines 34.88 12.12 52.99 4.47
Dorsal rostral length 46.62 2.70 50.67 12.32
Ventral rostral
length 46.12 1.31 52.52 8.24
Palatal width 22.86 0.0 77.14 1.94
Foramen magnum width 21.64 4.44 73.87 4.16
Mandibular length 48.44 7.39 44.15 32.79
Mandibular depth 38.43 18.48 43.04 6.32

Coronoid angular
distance 39.79 7.58 52.63 14.95

Table 7 continued

57

 

 

Character Locality Sex Within Total
Coronoid height 49.95 7.12 43.01 14.40
Post—dentary ramus
length 50.33 1.83 47.83 13.87
Slope of forehead 18.23 0.0 81.72 11.12

 

Figure 12.

58

Dice-Leras diagrams representing character
trends in mandibular and basilar lengths across
latitude (corresponding to collecting locali-
ties) for C. perspicillata. Vertical lines
Show observed ranges; rectangles mark standard
deviation; horizontal lines represent the mean
for the p0pu1ation sample. Latitude is given
along the x axis in degrees, a positive value
is given for north latitude and a negative for
the south latitude. The y axis indicates scale
for character values.

 

5S3

 

 

 

 

 

 

 

 

.NH magena
faced eating-u:
01220..
cm: 2| 2.. 0| v: 0 o 0. v. o—
n... 4
. L .
L s 3 .
50:00 3.300
.353
out .mu m... :I at v: n n o n. n—

 

 

 

I
fl

 

 

'I—._,___J

 

‘1

 

 

 

 

 

 

 

0n—

mn—

oc. :::

m9—

no.

0:
ms—
028:.
no—
00—

no—

60

A complex pattern of relationships is produced when
canonical variates analysis is performed on all fifteen
populations. There is much overlap in the groups formed
making interpretation difficult. Subsets of this one
analysis are examined so that the relationships between
individual populations can be examined. First, p0pu1a-
tions with more than twelve individuals are examined sepa-
rately (Figure 13). Populations from Mexico and Honduras
are fully separated from the South American populations
which extend south and east of Colombia on the first canon-
ical axis. This first axis accounts for 49.09% of the
total variance, with only basilar length loading highly.
The most divergent population examined, from Tacaruna
Village Camp, Panama, is separated from the other popula-
tions on the second canonical axis. Characters with high
loadings on the second axis are basilar length, mandibular
depth and coronoid-angular distance. The population from
Costa Rica and Panama are completely separated on this
axis, which accounts for 11.9% of the total character
variance. Along the third axis, populations from Costa
Rica, Colombia and Peru are separated from each other.
Characters with high loadings include maxillary tooth row
length and coronoid angular distance; 10.04% of the char-
acter variance is accounted for on this axis. Two more
axes, the fourth and fifth, remain significant in this
analysis and further separate the p0pu1ation from Trinidad

from the other populations sampled. Mandibular and palatar

Figure 13.

61

Canonical variates analysis of eight popula-
tions of Q. perSpicillata sampled. All fifteen
populations were included in the analysis; only
those with greater than 12 individuals are
illustrated in this figure. Populations
included are: l. Teapa, Mexico; 2. Rio Coco,
Honduras; 6. Palmar, Costa Rica; 9. Madden
Dam Road, Panama; 10. Valledupas, Columbia;
12. Rosharinho, Brazil; 13. Pucallpa, Peru;
15. Sapucay, Paraguay. The first and second
canonical axes are represented by the x and y
axes respectively. The third axis separation
is presented in the insert. Units along the
axes represent the canonical variates scores.
The group centroids for each population are
represented by solid dots.

 

63

length have high loadings on the fourth and fifth axes
respectively.
Mexican and Central American p0pu1ations of Q.

perSpicillata are illustrated separately from South

 

American p0pu1ations; drawn from the canonical variates
analysis which includes all p0pulations (Figure 14). The
Mexican and Central American p0pu1ation samples are sepa-
rated very little on the first two canonical axes. The
third axis, however, separates all of the p0pu1ations.
The p0pulations most differentiated here are those from
two localities in Costa Rica, Palmar and Cariblanco and
one from 18 m WSW of Chepo, Panama. The South American
populations (Figure 15) group tightly on the first three
canonical axes. On the fourth and fifth, Trinidad is dif-
ferentiated from the mainland populations. There is very
little differentiation between South American 9.

perSpicillata p0pulations.

 

Phenetic Relationships Between Populations

 

Cluster analysis was performed on all species p0pu1a-
tions to examine how p0pu1ations from throughout the range
of each Species group phenetically. P0pulation means of
canonical variates for each of the 22 characters were used
to calculate the squared Euclidean distance between popula-
tions. The average linkage method of Clustan's hierarchic
fusion procedure was used for this analysis (Clustan,

1975). The average linkage method uses the average of all

Figure 14.

64

Canonical variates analysis of g.
perSpicillata populations sampled, illustrating

 

only Mexican and Central American populations.
These include nine p0pu1ations: 1. Teapa,
Mexico; 2. Rio Coco, Honduras; 3. Sabana
Grande, Honduras; 4. La Gatiada, Nicaragua;
5. Yalaguina, Nicaragua; 6. Palmar, Costa
Rica; 7. Alajuela Cariblanco, Costa Rica;

8. R. de Panama; 9. Madden Dam Road, Panama.
First and second canonical axes are repre-
sented by the x and y axes reSpectively. The
third axis is presented in the insert. Units
along axes represent canonical variates scores.
Group centroids are indicated by solid dots.

 

 

 

 

IIIO

 

 

llo

 

 

Figure 15.

66

Canonical variates analysis of C. perSpicillata

 

populations sampled, illustratifig only those
from South America. Populations include:

10. Valledupar, Colombia; 11. San Rafael,
Trinidad; 12. Rosarinho, Brazil;

13. Pucallpa, Peru; 14. Buena Vista, Bolivia;
15. Sapucay, Paraguay. The first and second
canonical axes are represented by the x and y
axes, respectively. The third axis is indi-
cated in the insert. Units along the axes
represent the canonical scores. Group cen-
troids are represented by solid dots.

68

Similarity coefficients for pairs of populations, one from
each cluster (Clustan, 1975). This procedure will find
clusters of populations that are more similar within
groups than between groups. This method for computing
similarity coefficients attempts to take account of group
structure. The clusters of populations produced are
organized into a heirarchy based on similarity, which is
represented in a dendrogram.

The resulting dendrogram (Figure 16) provides a sum-
mary of the phenetic relationships between the species
p0pu1ations sampled. The four populations of g. castanea
cluster alone, in agreement with earlier taxonomic observa-
tions. The Q. castanea p0pu1ation from Costa Rica is more
morph010gically differentiated from the other C. castanea
p0pu1ations sampled than are the other Species from each
other. The p0pu1ations of g. subrufa sampled are divided
into a northern group of six populations ranging from
Mexico to Valle, Honduras and a southern group of two popu-
1ations from Sabana Grande, Honduras and San Antonio,
Nicaragua. Between these two groups lie two populations

of g. brevicauda from Teapa, Mexico and Talanga

 

Honduras. The populations of Q. brevicauda and Q.

 

perSpicillata are ambiguous. As Pine had noticed, south-

 

ern g. perSpicillata overlap northern Q. brevicauda.

 

 

Southern 9. perspicillata appears as a subgroup within Q.

 

brevicauda. The northern populations of Q. perspicillata

 

 

ranging from Mexico to Panama form a distinct group,

 

Figure 16.

69

Dendrogram of all 36 species populations
resulting from cluster analysis. Species
abbreviations used are: c-castanea, s-subrufa,

b-brevicauda, p-perSpicillata. Distance indi-
cated across tOp of diagram.

 

70

.03 oceanm

050000 00 .0 080000
0E300.0> 0300.022

853:8 22.02 .83 530
8330 3 .3089:

0000 0.: 505000:

0002. 62.82

.8330 cam .32....»

.00 600 000002 050000
22> 82.0 0.2.8
.2500 .00.: 0.000

 

02.0030 .200

«00300:; 0353.00

00:20 000000 «050:0...
003000 $0300.00

005.0000 53.0

00:03:00 0.030? .00.”. 0800
0x30 00003.5

0500

000:5 0032000... 060000
05300.; 0302022

0000 0:... 005000:
«2:00 0:030 6.060.030
500.0330 0.: 00:82

 

l

mfiilrmifiirfijfin

00 09 a a an net-0.00 03.0-03.0.03 QQQQQQQQQQQQGQO.

 

 

 

0.00:2 00m 0300.022
0000.0 0:003 00.30001
0000.: 60.3000...

0002. 62.8.2

00:00.. 3 0000200 E
000:3 .0008!

 

 

0.353020 0.060030
0._0> .3520...
0.300000» 00:62
00020000: .0208:

 

 

 

 

 

b 06.00 .00.: 830

Mk —III 83:03... .300

F 0800

r” 000:5 003.000... .0530...

00 0000 0.0 .3520:
u!

2
0

ON. 0n CV 00 ON

71

except for three p0pu1ations within this range (Sabana

Grande, Honduras; Palmar, Costa Rica; Canal Zone, Panama),

which group with the South American p0pu1ations sampled.
There is morphological differentiation between Central

America and South American populations of g. perSpicillata

 

that in part corresponds to the subspecific designation
suggested by Pine (1972). Specimens from north and west of

the Amazon Basin have been assigned to the subSpecies Q.

 

p. azteca and those from the Amazon Basin and Parana drain-

age have been called C. p. perspicillata and g. p. tricolor

 

respectively. Cluster analysis separates the Mexican and
Central American p0pu1ations from South American popula-
tions in correspondence with the distinction of C. p.

azteca from the South American Q. perSpicillata. There is

 

no distinction made between South American p0pu1ations

correSponding to the subSpecies C. p. perSpicillata and g.

 

p. tricolor. The distinction between Central American and
South American p0pu1ations is not free from overlap.
Several Central American populations sampled cluster with

the samples from South America.

DISCUSSION

In past taxonomic treatments of Carollia there has
been little difficulty in distinguishing Q. castanea

while 9. subrufa, C. brevicauda and C. perSpicillata were

 

 

 

not clearly defined until Pine recognized C. brevicauda in

 

1972 (Pine, 1972). The distinCtiveness of the four Species
of Carollia defined by Pine (1972), is more clearly
observed using canonical variates analysis than classical
means. Canonical variates analysis results agree with past
taxonomic observations in identifying C. castanea as the
most distinctive Species and the Species Q. subrufa, Q.

brevicauda and Q. perSpicillata are less clearly defined.

 

 

E. brevicauda appears intermediate to g. subrufa and Q.

 

perSpicillata in character values and shares phenetic simi-

 

larity with both Species. The intermediate nature of Q.

brevicauda, in part, explains the past difficulty in defin-

 

inging species limits. The Species are still difficult to
distinguish when relying on classical means. Morphological

overlap between populations of Q. subrufa and C. brevicauda

 

and also between C. brevicauda and g. perspicillata popula-

 

 

lations may be the reason for this continuing difficulty in

Species distinction.

72

73

The Specimens left unidentified by Pine (1972) were
easily assigned to existing Species using discriminant
function analysis. Unknowns were placed with the species
having the closest group centroid and thus the greatest
similarity in cranial measures examined. This does not
mean that these specimens are actually members of the
species group that they are placed with, but that this is
the most likely group membership. The Specimen from
Kartabo, British Guiana was placed with Q. subrufa in my
analysis. Pine notes external features such as pelage

which resembles that of Q. brevicauda and that size is

 

small as that in g. castanea. The teeth, however, do resem-
ble those of 9. subrufa. It is easy to see here that when
all factors are considered, identification to Species is
complicated, since not all features agree. The Specimen

from Napo Pastaza, Ecuador was placed with Q. brevicauda

 

using skull characters. Pine also wanted to place this

specimen with C. brevicauda, but its small size excluded it

 

from the rest of the series of g. brevicauda taken from the

 

same locality. The specimen from San Juan, Peru was placed

with g. perSpicillata. Pine noted that this Specimen had a

 

skull reminiscent of both C. perSpicillata and Q. brevicauda,

 

 

the size is a good deal larger than any Carollia Species
from that far south, the central lower incisors are small

for g. perSpicillata and the body hair is quite long and

 

tricolored as found in Q. brevicauda. The above mentioned

 

specimens were considered as possible undescribed species

74

by Pine (1972). The last unidentified Specimen was an old
female individual with worn teeth and pelage, that Pine
was sure belonged in the genus Carollia but he could not
place it to species. He thought it was either 9. subrufa

or g. brevicauda. Discriminant analysis placed it with g.

 

brevicauda.

 

The most important source of variation in the cranial
morphology with Carollia species is the individual vari-
ance component which averaged over Species accounts for
69% of the total variation. The potential causes for this
variability include measurement error, response to chang—
ing environment, sampling more than one biological popula-
tion, random changeznuiunexamined environmental effects.

A portion of this residual variation can be explained by

measurement error which is on the average only 5% of the

total variation. Other factors must also be contributing
to this large amount of individual variation.

The variables exhibiting high residual variance com-
ponents (LIW, MSL, MW, FMW, MD, SF) were not examined in
Pine's (1972) treatment of the genus. This coincides with
the approach in character evaluation that is taken by a
taxonomist. These characters are of no use in distinguish-
ing between species because they are more variable within
species than between. A taxonomist is always looking for
characters that can be used consistently to distinguish

between Species, while ruling out those creating

75

confusion. These areas overlooked represent the most
plastic regions of the skull.

Foraging patterns unique to individuals of the same
species may be contributing to this high individual vari-
ation. Heithaus and Fleming (1978) studied the foraging

pattern of g. perSpicillata in Costa Rica during the wet

 

season and found that individuals used the same feeding
areas consistently with very little overlap between
individuals. Feeding on fruits which differ in thickness
of Skin and consistency could cause a change in the magni-
tude of forces acting on the Skull and mandible produced
by the muscles involved in mastication: the masseter and
temporalis. The bones in the mammalian Skull undergo a
constant remodeling by absorption and deposition of bone
tissue, that may be functionally influenced (Straney,
1983; personal communication). If the individuals within
a p0pu1ation have consistent differences in the food
eaten, the areas where the muscles involved in mastication
originate and insert would be the most variable. Compari-
son of the areas of origination and insertion to regions
of the Skull and mandible with high residual variance com-
ponents shows a fair correspondence (Table 8).

Maxillary spur length has a high residual variance

component across Carollia species. The maxillary spur is

 

the anterior portion of the incomplete zygomatic archof
Carollia species where the masseter originates. The

posterior mandibular regions including the

 

m.H0 4m: N.om 30

 

76

 

 

N.mm 4m: n.5w 4m: m.~m :0 «.mm mm
N.Hw mm c.0w mm o.mm 0<u 0.00 mm
m.mw qu o.wm mm o.wm 220 N.nm 4m:
mumaawommmkwm .w 003000>0un .w mmsgnsm .w mocwpmmo .m

 

Amucmwpm> H0009 0:0 mo w 0 m0 wommmpmxm.
mucmcomsoo mocmwum> Hmswflmom cm.: 20.3 mumuumumco

w m4m<H

77

coronoid angular distance and coronoid height have a large
residual component in Q. subrufa. These regions are
insertion sites for the masseter and the temporalis. It
is possible that the effects of resource partitioning,
between individuals of the same Species p0pu1ation, may be
contributing to this high individual variance in character
values.

There are, however, other explanations for the high
variability in a few of these characters. The maxillary
spur is a cranial feature which is prone to damage because
of its Shape (long and narrow) and position (extending
laterally) on the skull. The SlOpe of forehead was one of
the more involved measurements to take and thus may have a
larger error due to measurement technique. There are
aspects of the p0pu1ation structure not yet considered as
potential cause for this large residual component of vari-
ation seen within p0pu1ations.

A biological population can be defined as a group of
interbreeding individuals that breed more among themselves
than with other such populations. In bat Species, a
biological population may consist of only the individuals
occupying one roost site. When specimens are collected
with a mist net, there is no assurance that only the
individuals from one such p0pu1ation are taken. It is pos-
sible that individuals from several biological populations
are represented in each p0pu1ation sample. Individuals

from different biological populations are likely to differ

 

78

due to genetic divergence. This genetic divergence may be
reflected in morphological variation within p0pu1ations.
This is one more level at which Carollia may be organized
and at which variation is occurring. The data in this
analysis is purely morphological, so genetic differentia-
tion cannot be compared to morphological differences
between p0pu1ations.

Factors, such as random genetic variability and
unexplained environmental effects, must also be considered
as potential sources of variation. Genetic variability is
present between Carollia species (Straney, 1980) but gene-
tic variation between populations has not been examined in
detail. The Specimens examined were taken over a span of
100 years; in this time environmental change has most
likely occurred. Changing environmental factors provide
an external influence which may affect morphology.

Secondary sexual dimorphism contributes a small pro-
portion to the total variation within each Species, but is
present to varying degrees within all Carollia species.
Males are larger than females in all Species. This is com-
monly seen in other mammal Species. The adaptive signifi-
cance most often assigned in Darwinian sexual selection,
whereby males must compete for mates and large body size
is selected because it is advantageous (Darwin, 1859).

This is one possible explanation, but little is known about
the breeding behavior of Carollia except that Porter (1975)

noticed harem behavior in a laboratory colony of g.

79

perSpicillata. Other phyllostomid bats in which males are

 

larger than females are Uroderma bilobatum (Baker gt 31.,

 

1972), Phyllostomus discolor (Power and Tamsitt, 1973;

 

Bradbury, 1977) and in Anoura cultrata (Nagorsen and

 

Tamsitt, 1981). In some vespertilionid bats, the Opposite
trend is seen, particularly in wing dimension (Myers,
1979). This trend has been associated with a need to com-
pensate for extra loading of pregnancy and the transport
of young, coined the "Big Mother" hypothesis by Rall
(1976). Female Carollia also must deal with the extra
loading of pregnancy and carrying young. Examination of
Pine's (1972) data indicates that females have forearm
lengths equal to those of the males, who are larger in
weight. Females appear to have longer wing dimensions
relative to body size, which may compensate for the extra
loading of pregnancy.

Sexual dimorphism differs in degree and in the char-
acters that vary between Carollia Species. Carollia

subrufa and Q. brevicauda are the most dimorphic. They

 

share a number of characters that vary most between sexes,
with no locality affects. These include depth of brain
case, rostral breadth and mandibular depth, which are
larger in males. From these measurements, it appears that
males in these two Species have a deeper brain case and a
wider muzzle than do females. The other two Carollia 7
Species are less sexually dimorphic in terms of the ability

to discriminate between males and females. Rostral

80

breadth is the most important character exhibiting sexual

dimorphism in g. perspicillata, with only a small sex

 

effect in a large number of characters. Many of the char-
acters with a small sex effect also have locality effects
which may reduce the effectiveness of discrimination

between male and female 9. perspicillata. Least dimorphic

 

is g. castanea. Very few characters are involved and only
palatar length and coronoid-angular distance have a large

variance component due to sex. Pine (1972) had noticed an
average difference between male and female cranial measure-

ments only for C. perSpicillata. He had not noticed dif-

 

ferences between the sexes for the other Carollia species.
The reason for these differences between species is impos-
sible to interpret from these data. Perhaps a Study of
the differences in social structure and behavior of
Carollia Species might aid in understanding these differ-
ences.

Geographic differentiation of cranial morphology is
great in all four species of Carollia. Past taxonomic
problems can in part be attributed to an attempt to com-
bine Specimens from throughout the range of each species
without taking geographic variation into account. This
variability between collecting localities was first
described in general terms by Pine (1972) who notices
regular differences among p0pu1ations of C. subrufa, Q.

brevicauda and g. perSpicillata. Morphometric examination

 

 

of cranial features reveals a latitudinal trend in skull

81

length in all Carollia Species. Individuals from the
northern portion of the range of each species are consis-
tently larger than those from the southern portion. Simi-
lar patterns have been found in other phyllostomid bats.
Nagorson and Tamsitt (1981), for example, revealed a south

to north increase in Size for Anoura cultrata. Johnston

 

and Selander (1972) found a large size factor associated
with climate in North American house sparrows. Rees
(1976) found that a size factor in measurements on white-
tailed deer correlated positively with latitude. The
biological Significance of these size trends seen in a
variety of animal species is not clear. The usual assump-
tion is this increase in size along a latitudinal gradient
must have some adaptive significance.

Bergman's rule has been used to explain an increase
in body Size from south to north, with animals in cooler
(more northern) environments tending to have larger body
size to provide a smaller relative surface area. The
adaptive advantage gained is an increase in heat retention
efficiency. One problem with this explanation in the case
of Carollia is that size would be expected to increase as
a function of diStance from the equator, but individuals
from south of the equator in Bolivia (25 S latitude) are
smaller than those from closer to the equator. Another
problem is that the northern and southernmost populations
are still in a warm climate; Bergman's rule is usually

associated with p0pu1ations of the same species which are

82

found in both a cold and warm climate. It seems likely
that other principles are operating.

Another hypothesis used to explain an increase in
Size in more northern p0pu1ations, advanced by Grant (1965)
and Heaney (1978), is based on the assumption that inter-
specific competition is greater in Species-rich areas,
like the tropics, and so selection favors smaller sized
individuals that can occupy more specialized feeding
niches. By the same reasoning, areas with fewer species
should have less competition for resources and so selec-
tion may favor larger size so that a wider range of
resources can be utilized. The smallest Carollia do come
from tropical South America and the largest are from
Mexico, but detailed ecological data are lacking for each
of the p0pu1ations sampled. There are, however, fewer
phyllostomid, glossophagine and total bat species in
Central America than South America (Nagorsen and Tamsitt,
1981). With fewer species of bats in Central America, a
greater variety of food may be available to Carollia
Species and so there may be a selected advantage for larger
size. The general trend in bat species diversity does dis-
play differences across the range of Carollia Species that
are concordant with the size change within these species.
McNab (1971) suggested that latitudinal change in body
Size may be the result of the distribution of food species,
or of competitors for the same food resources. An inter-

action between these factors is likely.

83

The p0pu1ations of Q. castanea are very distinct from
one another. Pine had noticed differences in the average
measurements between 9. castanea populations. The differ-
ences between the four p0pu1ations sampled were greater
than those between any of the other Species p0pulations;
the p0pu1ation from Costa Rica is particularly distinctive.
A reduction in gene flow through geographic distance may
account for this large divergence. Q. castanea is known
to be habitat Specific preferring tropical evergreen for-
ests and to have a relatively small home range (Handley,
1966; La Val, 1970). These factors may be enough to hinder
gene flow allowing small p0pu1ations to differentiate more
than other Species. A closer look at Q. castanea popula-
tions would be interesting to see if this apparent diver-
gence is genuine, not just a function of the small number
of p0pu1ations used in this study.

Geographic differentiation is well developed in g.
subrufa. Q. subrufa has the smallest range of all four
Carollia species. It is distributed from Mexico to
Honduras occurring only on the Pacific coast Side of the
Sierra Madre. Q. subrufa is the only Species of Carollia
throughout most of its range. It occurs sympatrically

with g. perspicillata in Chiapas and Nicaragua, and with Q.

 

brevicauda on the gulf coast Side ongonduras. Q. subrufa

 

is primarily an inhabitant of tropical deciduous forests.
Distinct p0pu1ations from Honduras and Nicaragua cluster

separately from the more northern p0pu1ations. The

 

84

southern populations are smaller in skull and mandibular
lengths than the northern populations. One possible explan-
ation for this divergence is character displacement, due to
contact with other Carollia Species in this southern portion
of the range. This has been described in birds and mammals
by Grant (1972, 1975). Selection acts against similarity
between sympatric p0pu1ations of two systematically related
Species (as compared to their allOpatric p0pu1ations) so
that they do not compete for food or mates. Pine (1972)

makes note of collecting C. brevicauda in the same net as

 

C. subrufa in Honduras, 2 miles west of San Pedro Sula. At
this location he noted an exaggerated difference between the
two Species and suggests the possibility of character dis-
placement. The Q. subrufa sampled from Sabana Grande,
Honduras and San Antonio, Nicaragua, the regions where sym-
patry is likely, are very distinct from the more northern

p0pu1ations where C. subrufa occurs allopatrically from Q.

brevicauda. In order to be sure that character diSplacement

 

is occurring, a large sample of Q. subrufa and g. brevicauda

 

caught in the same net is needed for comparison. Such
specimens could be compared to known allopatric populations
of the same species. Pine treats C. subrufa as monotypic,
but notes that individuals from the northern part are larger
than those of the southernmost part, and Specimens taken
west of the Isthmus of Tehauntepec have hairier forearms

than do those from the other sidecfifthe Isthmus. He sug-

gested that with examination of more Specimens, the

85

existence of subSpecieS may become apparent. Both canoniv
cal variates analysis and cluster analysis identify a dis-
tinct southern p0pu1ation of E. subrufa that may represent
a recognizable subspecies.

Carollia brevicauda is widely distributed and comple-

 

mentary to g. subrufa along the wet Gulf-Caribbean coast of
Central America, north of Honduras. It has been collected
sympatrically with Q. castanea, Q. subrufa and Q.

perSpicillata. There may be character diSplacment occurring

 

between C. brevicauda and C. perSpicillata. Pine (1972) has

 

 

mentioned that specimens of these Species taken from the
same collecting locality are more easily distinguished from
those from a110patric populations. The p0pu1ations of Q.

brevicauda from Teapa, Mexico and Talanga, Honduras cluster

 

with g. subrufa. This divergence may represent character

displacement between C. brevicauda and Q. perSpicillata.

 

 

Sympatric p0pu1ations of Q. brevicauda and Q. perSpicillata

 

 

were not examined from Teapa, Mexico or Talanga. Sympatric
p0pu1ations from Rio Coco, Honduras, Yalaguina, Nicaragua
and Alajuela Cariblanco, Costa Rica were examined and do
Show a large morphological distance in the cluster analysis
results. Character diSplacement may be occurring between

C. brevicauda and g. perspicillata; another potential source

 

 

of intraspecific variation that must be considered at the

p0pu1ation level.

Discriminant analysis of g. brevicauda p0pu1ations sepa-

 

rates the southern populations from Panama and Ecuador from

86

the more northern populations. The divergent p0pu1ation
sample from Panama correSponds to an unusual p0pu1ation
noticed by Pine (1972), that he first regarded as a hybrid

swarm of C. subrufa and Q. brevicauda. Externally, they

 

lack the hairiness of the forearm normally seen in Q.

brevicauda. The skull dimensions differ in palatar and dor-

 

sal rostral lengths, decreasing in length while the maxil-
lary tooth row and basilar lengths increase. These differ-
ences in dimension are evident when correlating character
means with latitude. The decrease in palatar length and
increase in maxillary tooth row length are changes toward a

more C. perSpicillata-like skull, which suggests possible

 

hybridization with Q. perSpicillata, not Q. subrufa as sug-

 

gested by Pine (1972).

Carollia perSpicillata is the most widespread and com-

 

mon of the four Species. It is found in both tr0pical
evergreen and deciduous forests. Bloedel (1955) noted that
individuals may be found solitary or clustered in colonies
of as many as 1,000 individuals. Greenhall (1959) wrote
that it "is probably the most abundant fruit eating bat in
the American trOpics." Darwin (1859) observed that the
wide ranging and common Species tend to be the most vari-
able. When the percent contributions to the total vari-
ance due to locality is compared between Carollia species,
variation increases with an increase in the size of the
species range. The highest value of 34.6% is found in g.

perSpicillata, the most wide ranging species. The lowest

 

87

value of 16.0% is that of Q. subrufa, with the most
restricted range. Darwin's prediction is right in this
case. This is not very surprising, as the wide ranging and
common Species must adapt to many local environments and
develOp genetic isolation because of distance.

Geographic differentiation is greater in Q.

perSpicillata than any of the other Carollia species. The

 

northern p0pu1ations from Mexico and Honduras do not over~
lap the southernmost p0pu1ations from Peru, Bolivia and
Paraguay, but the populations which fall in between form a
continuum. There is no clear division between p0pu1ations
north and west of the Amazon Basin that would correspond to
the subspecific designation 9. p. azteca, given by Pine
(1972). Nor is there a clear division between the sub-
Species g. p. perspigillata and C. p. tricolor from the
Amazon Basin and Parana drainage respectively, when consid-
ering cranial morphology alone. Pine distinguished E. p.

azteca from C. p. perspicillata on the basis of its larger

 

Size, and distinguished C. p. tricolor from g. p.

perspicillata on the basis of its small Size, soft tricolor

 

pelage and hairy forearms and toes. The characters used by
Pine (1972) were all external features, mine were cranial
measurements. This indicates that the pattern of variation in
cranial morphology varies independentlyffimmIthe external
features usedlanine (197Z)1x>distinguish between C.

perSpicillata populations{hmmldifferent geographic regions.

 

88

Two of the southernmost Q. perspicillata p0pu1ations

 

(Rosarinho, Brazil; Sapucay, Paraguay) cluster with northv

ern p0pu1ations of g. brevicauda. This agrees with Pine's

 

(1972) observations that the southern Q. perSpicillata are

 

morphologically Similar to northern C. brevicauda. South-

 

ern C. perSpicillata are small (relative to northern p0pu-

 

1ations) and Northern C. brevicauda are large (relative to

 

southern populations), so a size factor may in part be
reSponSible for this similarity. External features are
also shared, including hairy forearms and toes, character-

istic of E. brevicauda.

 

The affinity between the species of Carollia can be
reexamined in light of these morphometric findings. Pine
(1972) has suggested that Central America may be the center
of origin for the species of Carollia, since they all
co-occur there and C. subrufa occurs nowhere else. The
continental divide may have played a major role in forming
a barrier to gene flow leading to the differentiation of

C. subrufa and Q. brevicauda from a common ancestor (Pine,

 

1972). Using classical techniques, Pine (1972) postulated
that the relationship between species went: 9.

perSpicillata - Q. brevicauda - g. subrufa - g. castanea --

 

 

representing a sequence from least modified to most highly
modified from a common ancestor. My morphometric results
provide a distance (Mahalanobis distance) between each pair
of Species. Mahalanobis distance between C. castanea and

the other Species is greater than the distance between any

89

of the other species, indicating that Q. castanea is the
most phenetically divergent, as Pine (1972) has also noted.

The phenetically closest species are C. subrufa and Q.

 

brevicauda, and these are about equidistant from Q.

perSpicillata in canonical variates analysis. Cluster

 

analysis provides slightly different results: 9. brevicauda

 

and C. perspicillata cluster together equidistant from C.

 

subrufa. Canonical variates analysis results are probably

a better representation of phenetic affinity between species
because separation occurs in multidimensional Space rather
than the two dimensions of cluster analysis. Both tech-

niques indicate that C. brevicauda, Q. perSpicillata, and Q.

 

 

subrufa have morphologically similar cranial features.
Another source of data that can be considered is
chromosomal morphology. There is an X-autosomal trans-
location in all species p0pu1ations, except for some of the
Q. castanea from Peru (Patton and Gardner, 1971). Hetero-

chromatin patterns (in C bands) of g. brevicauda and Q.

 

perSpicillata are very similar, but the chromosomes of g.

 

castanea lacked much of the heterochromatin common to the
other two Species (Stock, 1975; g. subrufa was not examined
by Stock). The lack of a chrosomal translocation in some of
the Peruvian g. castanea indicates that either this p0pu1a-
tion has lost this trait or all of the others gained this
translocation at some point during the evolution of Carollia.
Heterochromatin patterns further indicate the uniqueness of

Peruvian Q. castanea and the similarity of C. brevicauda

 

90

and Q. perSpicillata. This chromosomal affinity agrees

 

with the morphological similarities and differences between
these three species.
Isoelectric focusing results group C. castanea and g.

perSpicillata and also 9. brevicauda and g. perspicillata

 

 

 

(Straney, 1980). The distinctiveness of C. castanea is evi-

dent in the morphological and chromosomal data, but the

electrOphoretic data show possible convergence between C.

 

castanea and g. perspicillata.
The most morphologically and chromosomally differen-
tiated Species is C. castanea. The relationships among 9.

subrufa, Q. brevicauda and Q. perSpicillata is more prOblem-

 

 

atic due to conflicting information. An association between

C. brevicauda and Q. perspicillata is found most often;

 

 

morphological, heterochromatin (C banding) and isoelectric
focusing results agree. The position of Q. subrufa is not

clear. Morphologically it is closest to C. brevicauda. The

 

geographic distribution of Q. subrufa and C. brevicauda indi-

 

cates that they may have Split from a common ancestor in
Central America and Spread north on separate Sides of the
Sierra Madres, where they speciated a110patrically. The
possible convergence between E. castanea and g.

perSpicillata, identified by isoelectric focusing data, is

 

not evident in the morphological or chromosomal data.
The relationships that can be constructed are numerous.
C. castanea is the most morphologically and chromosomally

derived Species. The affinity between C. castanea and any

91

of the other Carollia is unclear, morphologically it is
closest to C. subrufa while biochemically it is closest to

E. perSpicillata. The Species Q. subrufa and Q. brevicauda

 

 

are phenetically very Similar and have a geographic distri-
bution which suggests a Split from a common ancestor. g.

perSpicillata and C. brevicauda Show a close affinity at

 

 

all levels considered. There is not one diagram that can
be used to represent species affinity that fully agrees

with all sources of information available. The most gener-
alized diagram that can be used to represent the relation-
ship between Carollia species in agreement with all data
considered is indicated in Figure 17. The direction of
change is difficult to determine, as is the point of the
original divergence of Carollia Species. This diagram indi-

cates that the affinity between C. perSpicillata, Q. subrufa

 

and C. brevicauda is unclear and Q. castanea has diverged

 

first from an ancestor common to the other three species.

92

Figure 17. Diagram of Species affinities.

93

Cpuspicillata Cbrovicaudo C.subruf0

C. castanea

 

Figure 17.

CONCLUSIONS

Bat species in the genus Carollia, defined by Pine's

 

(1972) classification, are easily distinguished using canon-
ical variates analysis. They form four distinct morphologi-

cal species: C. castanea, Q. subrufa, C. brevicauda and Q.

 

 

perSpicillata. The most distinct Species is C. castanea,

 

 

which displays no morphological overlap with any of the

other Carollia species. The species C. subrufa and C.

 

brevicauda overlap slightly, while 9. brevicauda and C.

 

 

perSpicillata overlap to a greater degree. These findings

 

are in agreement with past taxonomic observations. Q.

brevicauda was the last Species recognized, probably because

 

of its morphological Similarity with both 9. subrufa and Q.

perSpicillata. The four specimens which Pine (1972) was

 

unable to identify to Species using classical methods were
easily placed to Species using discriminant analysis. This
ease is due in part to using only cranial and mandibular
features. Pine's (1972) placement based on cranial and den-
tal features alone agreed with my findings. The external
features did not agree with these and so Pine (1972) did not
place these Specimens to Species. Discriminant analysis
provides an answer in terms of the most probable group

membership.

94

95

There are apparent sympatric effects between C.

subrufa and C. brevicauda. g. subrufa inhabits the Pacific

 

coast, while C. brevicauda is primarily a Gulf-Caribbean

 

versant, separated from one another by the Sierra Madre in
the northern portion of their range. There is a narrow
region of contact on the Gulf-Caribbean side of Honduras.
The E. subrufa p0pu1ations sampled from Sabana Grande,
Honduras and San Antonio, Nicaragua are well differentiated
from the Six northern populations. This differentiation may
be the result of character diSplacement due to sympatry

with g. brevicauda in these locations. Pine (1972) noted

 

the possibility when 9. brevicauda and Q. subrufa caught in

 

the same net in Cortes, Honduras Showed exaggerated differ-
ences. Further comparison of sympatric populations of Q.

subrufa and C. brevicauda to a110patric p0pu1ations would

 

be useful in testing for sympatric effects.

Sympatric p0pu1ations of g. brevicauda and 9.

 

perSpicillata are examined from Honduras, Nicargua and

 

Costa Rica and also Show apparent character displacement.
These sympatric p0pu1ations are phenetically distant com-
pared to allopatric p0pu1ations of the same species. These
sympatric effects provide a source of intraSpecific varia-
tion which should be more closely examined.

IntraSpecific variation is great in all Carollia

 

Species. The individual component of variance is the larg-

est source of variation for all four Species of Carollia.

 

This indicates that the characters used (all considered

96

together) vary more between individuals than between sexes
or between localities. This is not true for each character
separately. This indicates that much of the variability
seen in Carollia speciesisscaused by factors other than the
location or sex of the Species. Other factors include mea-
surement error, response to changing environment, sampling
more than one biological population, random change or con-
sistent individual differences in feeding strategy.
Geographic variation is the second largest source of
intraspecific variation in all Carollia Species. g.

perSpicillata has the largest prOportion of the total char-

 

acter variation attributable to location, followed by C.

castanea, g. brevicauda and then C. subrufa. The most

 

important variable used to discriminate between populations
is basilar length (all four species). This may correSpond
to the general north-south trend in skull length noted
across Species of Carollia. Pine (1972) had noticed a size
trend in all Carollia species except 9, castanea. He noted
larger individuals in the north and p0pu1ations of smaller
individuals in the southern portion of their range. In

both E. subrufa and 9. brevicauda p0pu1ation samples, it

 

is possible to distinguish between a northern and southern
group of p0pulations. Q. castanea p0pu1ation samples are
distinctive, but Show no geographic pattern. P0pulations

of C. perSpicillata form a continuum in canonical variates

 

Space. The distinction between northern and southern popu-

lations is not without overlap. The subSpecific distinction

97

between these p0pu1ations may only represent two ends of a
cline. Without examination of a large number of Specimens
representing the range of a Species, it is difficult to see
whether the variation is continuous or discrete. This is a
limitation in all studies of geographic variation. The num-
ber and location of p0pu1ation samples included in a study
is often limited to what is available in museum collections.
Sexual dimorphism is evident in all four species of

Carollia. Pine (1972) had noted that male 9. perSpicillata

 

were larger than females, but had not noticed dimorphism in
the other Species. In all species, males are larger than
females. Most dimorphic is C. subrufa followed by C.

brevicauda, C. castanea and then C. perSpicillata. Rostral

 

 

breadth is wider in males of the Species g. perSpicillata,

 

C. brevicauda and C. subrufa. Q. castanea males have a

 

greater palatar length than females. The variation attri-
butable to sex is a relatively minor source of variation in
all Carollia Species. Although the sex component is small,
future analysis involving p0pu1ations differences may bene-
fit by treating sexes separately, providing sample sizes
are larger than those available in this study.

The results of this study confirms much of the past
observations concerning morphological variation within
and between Carollia Species. Patterns of variation are
more apparent and can better be compared when morpho-

metric techniques are applied to problematic groups. The

98

morphometric techniques used allowed quantification of the
relative amounts of variation due to species, locality,
sex and individual differences which could not be achieved
by classical means. From this information, hypotheses con-
cerning the underlying causes for these observed differ-

ences can be advanced and when possible, tested.

APPENDIX A

SPECIMENS EXAMINED

The collecting localities are grouped below according
to species. Specific localities from which specimens were
examined are grouped by country. Museum acronyms are:

AMNH - American Museum of Natural History; CM - Carnegie
Museum, Pittsburgh; FMNH-—Field Museum of Natural History;
KU - University of Kansas Museum of Natural History; LACM -
Los Angeles County Museum; TCWC - Texas C00perative Wild-

life Collection; USNM - United States National Museum.

9. brevicauda

 

MEXICO -- Veracruz: Rio Quezalapam (l), 2 mi. E Lago
Catemaco, 566, 799 (TCWC 11131-11137, 11140, 11142-11145);
Chiapas, 21 km WSW Teapa (Tabasco) in Chiapas (5), 200 ft,
366, 499 (TCWC 16502, 16504-16509). GUATEMALA: Izabal (6),

 

25 km WSS Puerto Barrios, 300 ft, 1569, 1099 (TCWC 17295-
17319). HONDURAS: Francisco Morazan (9), 10 mi NE Talanga,
3400 ft, 16, 19 (TCWC 10844-10845); Rio Coco (10), 78 mi
ENE Danli, 900 ft, 1 sex?, 366, 899 (TCWC 9893, 9898, 9901,
9907, 9913-9918, 10654-10655). NICARAGUA: Dept. Madriz

 

Yalaguina (14), 10 km E Somoto, 2200 ft, 595, 19 (TCWC 8873,
8888-8889, 8900-8902). COSTA RICA: Alajuela Cariblanco

 

(16), 18 mi NE Naranjo, 3000 ft, 16, 299 (TCWC 9867, 9872,

99

 

100

9876). PANAMA: Tacarcuna Village Camp (20), 3200 ft, 1899,
2499 (USNM 309474-309479, 309481-309497, 309499-309509,
309516-309526). ECUADOR: Napo Pastaza (25), 8 mi WNW Puyo,
3800 ft, 966, 799 (TCWC 12047-12051, 12054, 12056-12057,
12059-12062, 12064-12067).

9. castanea
HONDURAS: Rio Coco (10), 78 mi ENE Danli, 900 ft, 899, 399
(TCWC 9935-9945). COSTA RICA: Puntarenas Prov. (17), 4 mi

 

NE Palmar, 300 ft, 266, 699 (TCWC 9919-9926). PANAMA:
Darien, Tacarcuna Village Camp (20), 1850 ft, 2896, 1399
(USNM 309406-309411, 309413-309420, 309423, 309425-309431,
309433-309438, 309441, 309443-309449, 309453-309456).
PERU: Huanuco (26), 19 mi S Tingo Maria, 2800 ft, 369, 299
(TCWC 11915-11919).

9. perSpicillata

 

MEXICO -- Chiapas: 21 km WSW Teapa (Tabasco) (S), 200 ft,
566, 699 (TCWC 16435-16439, 16441-16446). HONDURAS:
Francisco Morazan, 2 mi SE Sabana Grande (ll), 16, 19 (TCWC
10856-10857); Rio Coco (10), 78 mi ENE Danli, 900 ft, 469,
1499 (TCWC 9889, 9891-9892, 9894-9897, 9899-9900, 9902-9906,
9908-9911). NICARAGUA: Dept. Madriz Yalaguina (13), 10 mi

 

E Somoto, 2200 ft, 16, 19 (TCWC 8874-8875); Dept. Chontales,
1 km NW La Gatiada (14), 1300 ft, 466, 699 (TCWC 8860-8865,
8867-8870). COSTA RICA: Alajuela Cariblanco (16), 18 mi NE

 

Naranjo, 2900-3000 ft, 466, 399 (TCWC 9864, 9866, 9868-9871,
9873); Prov. de Puntarenas (l7), 4 mi NE Palmar, 300 ft,

 

101

699, 1299 (TCWC 9845-9848, 9850, 9852—9863). PANAMA: Bat
Caves, Madden Dam Rd. (19), 1099, 599 (LACM 20075, 20079-
20093); R de Panama (18), 18 km WSW Chepo, 200 ft, 799, 699
(TCWC 11920-11921,11923, 11926-11929, 11937-11940, 11942-
11943). COLUMBIA: Magdalena Sierra Negra (21), Villanueva
Valledupar, 766, 499, 1 sex? (USNM 281106-281115, 281121-
281122). TRINIDAD: San Rafael (22), 299, 299 (FMNH 61926- I
61929). BRAZIL: Rio Madeira Rosarinho (24), 2699,2699 0000: M
92056-92061, 92063-92070, 92072-92078, 92080-92100, 92182-
92187, 92220, 92627, 92629, 92632, 92634). PERU: Loreto

 

(27), 11 mi SE Pucallpa, 300 ft, 500 ft, 599, 1599 (TCWC
11982-11983, 11985-12002). BOLIVIA: Dept. Santa Cruz, Prov.
de Sara, Beunavista (29), 400 m, 450 m, 500 m, 1099, 499
(LACM 8942; FMNH 22442, 22444-22447; CM 2163, 2200; AMNH
61755-61760). PARAGUAY: Sapucay (30), 19, 1399 (FMNH 18208-
18209); USNM ll4005--the holotype of C. p. tricolor, 115003-

115012).

E. subrufa
MEXICO -- Oaxaca: 4 mi E Tapanatepec (2), 800 ft, 466, 899
(TCWC 16449-16450, 16452-16455, 16458-16560). Chiapas: 5
5 mi N Arriaga (3), 800 ft, 766, 1099 (TCWC 16478-16494);
4 km NW Tapachula (4), 450 ft, 766, 1299 (TCWC 14276-14284,
14499-14508). GUATEMALA: Chiquimula (l), 20 km SSE

 

Chiquimula, 550 m, 366, 799 (TCWC 17269-17278). HONDURAS:
Valle (10), 10 km E San Lorenzo, 25 ft, 466, 299 (TCWC
18353-18359); Francisco Morazan, 3 mi S Sabana Grande (11),

102

1500 ft, 15, 19 (TCWC 10846, 10859). EL SALVADOR: La

 

Libertad (8), 20 mi W La Libertad, 369, 19 (TCWC 8903-8905,

9807). NICARAGUA: Chinandega (15), San Antonio, 15 mi,

 

266, 499 (KU 97660-97665).

Specimens Not Assigned to Species

NICARAGUA: 1 km SE Yalaguina 2600+ ft,9 (TCWC 7414).

 

BRITISH GUINA (Guyana): Kartabo, 9 (AMNH 64168). ECUADOR:V

 

Napo Pastaza, 8 mi WNW Puyo, 3800 ft, 9 (TCWC 12068). PERU:
Puno, Sandia Prov. San Juan, TambOpata Valley, 5000 ft, 9

(FMNH 78394).

APPENDIX B

Pie diagrams of each character with percent
contribution to total variance for residual, sex

and locality for each Carollia species.

103

 

I I.“ @- comsoUOd U—oajmnOK I I

ovaoum>oca 02.0..ou

 

93 oz 3.2 2. no an

an .29. :z 3s: . 3.2 a: 30..
..::. E @32 Q @ @ ®

:3 £2 :6 :2

 

104

I new a: .3202: U .033an 00:03.: 02.0.ou
o<u

:3
2’3:

 

duo

 

32..

 

  
 
 

105

Q

 

 

O<U

 

     

I now 3:302: U .033»...

a m 3.... 32 305

 
 

.321

 

 

 

 

up:

 

 

0.0.2303...

 

e
do?

02.0.09

d<u

 

E

 

duo

106

..U

>>m

 

.u(u

duo

I E as 5:33 D .32....

um O¢<

 

3U 35 a:

‘30; a:

 

spec

 

 

|
0.3.Aaa

 

 

0_:o.0U

no

 

>>0m

 

BIBLIOGRAPHY

BIBLIOGRAPHY

Baker, R. J., W. R. Atchley, and V. R. McDaniel. 1972.
Karyology and Morphometrics of Peters' Tent Making Bat
Uroderma bibobatum Peters (Chiroptera: l
Phyllostomatidae). Syst. 2001., 21:414-429. 1

 

Bloedel, P. 1955. Observations on the Life Histories of
Panama Bats. J. Mamm., 36:232-235.

Bradbury, J. W. 1977. Organization and Communication.
W. Wimsatt (ed.). Biology of Bats. Vol. 3. Academic
Press, New York.

 

Darwin, C. 1859. On the Origin of Species by Means of
Natural Selection. J. Murry, London. 502 pp.

Elliot, D. G. 1904. The Land and Sea Mammals of Middle
America and West Indies. Field Columbian Mus. Pub.,
95, 2001. Ser. 4, Part 2:i-xiii, 441-850.

Felten, H. 1965. Fledermause (Mammalia:Chiroptera) Aus
El Salvador. Teil 3. Senckenbergiana Biologica, 37:
179-212.

Grant, P. R. 1965. The Adaptive Significance of Size
Trends in Island Birds. Evolution, 19:355-367.

Grant, P. R. 1972. Convergent and Divergent Character
DiSplacement. Biol. J. Linn. Soc., 4:39-68

Grant, P. R. 1975. The Classical Case of Character Dis-
placement. Evolutionary Biol., 8:237-265.

Greenhall, A. M. 1959. The Ecological Role of Trinidad
Bats, ESpecially the Vampire and Bat Rabies. Veteri-
nary Record, 71:188-190.

Hahn, W. L. 1907. A Review of the Bats of the Genus
Hemiderma. Proc. U. S. Nat. Mus., 32:103-118.

 

Hall, E. R., and K. R. Kelson. 1959. The Mammals of North
America. Vol. 1. Ronald Press Co., New York: i-xxx,
l-546, 1-79.

107

108

Handley, C. 0., Jr. 1966. Checklist of the Mammals of
Panama. In R. L. Wenzal and V. J. Tipton 2£.El-
(eds.). EctOparasites of Panama. Field Museum of
Natural History, Chicago, pp. 753-795.

Heaney, L. R. 1978. Island Area and Body Size of Insular
Mammals: Evidence From the Tricolored Squirrel
(Callosciurus prevosti) of South East Asia. Evolu-
tion, 32:29-44.

 

Heithan, E. R., and T. H. Fleming. 1978. Foraging Move-
ments of a Frugivorous Bat Carollia perSpicillata
(Phyllostomatidae). Ecological Monographs, 485127-
143.

Johnston, R. F., and R. K. Selander. 1971. Evolution in
the House Sparrow. 11. Adaptive Differentiation in
North American P0pulations. Evolution, 25:1-28.

La Val, R. K. 1970. Banding Returns and Activity Periods
of Some Costa Rican Bats. Southwestern Naturalist,
15:1-10.

Leamy, L. 1983. Variance Partitioning and Effects of Sex
and Age on Morphometric Traits in Randombred House
Mice. J. Mamm., 64(1):55-61.

Linnaeus, C. 1758. Systema Naturae Per Regna Tria
Naturae, Secundum Classes, Ordines, Genera, Species,
Cum Characteribus, Differentiis, Synonymis, Locis,
i-824, i-iii (Reprint Ed. of 1894 - "Cura Societatis
Zoologicae Germanicae iterum edita").

McNab, B. K. 1971. On the Ecological Significance of
Bergmann's Rule. Ecology, 52(5):845-855.

Miller, G. 8., Jr. 1924. List of North American Recent
Mammals, 1923. Bull. U. S. Nat. Mus., 128:i-xvi, 1-
673.

Myers, Philip. 1978. Sexual Dimorphism of Vespertilionid
Bats. American Naturalist, 112(986):701-711.

Nagorset, D., and J. R. Tamsitt. 1981. Systematics of
Anoura culturata, A. brevirostrum and A. werckleae.
J. Mamm., 62(1):82-102.

 

 

 

Nei, N. H., C. H. Hull, J. G. Jenkins, K. Steinbrenner, and
D. H. Bent. 1975. SPSS. Statistical Package for the
Social Sciences, 2nd edition. McGraw Hill Book Co.,
New York.

109

Patton, J. L. and A. L. Gardner. 1971. Parallel Evolution
of Multiple Sex-Chromosome Systems in Phyllostomatid
Bats, Carollia and Choeroniscus. Experientia, 27:105.

 

 

Pine, R. H. 1972. The Bats of the Genus Carollia. Tech.
Monogr., Texas A 8 M University, 8:1-125.

Porter, F. L. 1975. Aspects of Social Behavior in
Carollia. Paper presented at Sixth National American
Symp. on Bat Research, Las Vegas, Nevada.

Power, D. M. and J. R. Tamsitt. 1973. Variation in
Phyllostomus discolor (Chiroptera:Phyllostomatidae).
Canadian J. Zool., 51:418-461.

 

Ralls, K. 1976. Mammals in Which Females are Larger Than
Males. Quart. Rev. Biol., 51:245-276.

Rees, J. W. 1970. A Multivariate Morphometric Analysis of
Divergence in Skull Morphology Among Geographically
Contiguous and Isolated Groups of White-Tailed Deer
(Odocoileus virginianus) in Michigan. Evolution, 24:
2201229.

 

Stock, A. D. 1975. Chromosome Banding Pattern Homology
and Its Phylogenetic Implications in the Bat Genera
Carollia and Choeroniscus. Cytogenet. Cell Genet.

14:34-41.

 

Straney, D. O. 1980. Relationships of Phyllostomatinae
Bats: Evaluations of Phylogenetic Hypothesis. Pn.D.
dissertation, University of California, Berkeley, 278
pp. Dissertation Abstr. B. Sci. Eng., 41(7):2519,
January.

Straney, D. 0., M. H. Smith, I. F. Greenbaum, R. J. Baker.
1979. Biology of Bats on New World Family
Phyllostomatidae. Part III. Edited by R. J. Baker,
J. Knox Jones, Jr., and Dilford C. Carter. Special
Publications, The Museum, Texas Tech. Univ.

Straney, D. O. 1978. Variance Partitioning and Nongeo-
graphic Variation. J. Mamm., 59(1):1l.

Wishart, Clustan. 1978. University College London, 19
Gordon St., London, WC1H OAH.