Title Page
 Table of Contents
 Materials and methods: Cranial...
 Materials and methods: Dentary...
 Results: Cranial morphometrics
 Results: Dentary morphometrics
 Biographical sketch

Title: Systematics and zoogeography of fossil and recent pocket gophers in Florida / by Kenneth T. Wilkins
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00099230/00001
 Material Information
Title: Systematics and zoogeography of fossil and recent pocket gophers in Florida / by Kenneth T. Wilkins
Physical Description: x, 248 leaves : ill., maps ; 28 cm.
Language: English
Creator: Wilkins, Kenneth T., 1953-
Publication Date: 1982
Copyright Date: 1982
Subject: Pocket gophers   ( lcsh )
Pocket gophers, Fossil   ( lcsh )
Mammals -- Florida   ( lcsh )
Zoology thesis Ph. D
Dissertations, Academic -- Zoology -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis (Ph. D.)--University of Florida, 1982.
Bibliography: Bibliography: leaves 199-207.
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00099230
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000317781
oclc - 08876926
notis - ABU4609


This item has the following downloads:

PDF ( 8 MBs ) ( PDF )

Table of Contents
    Title Page
        Page i
        Page ii
        Page iii
        Page iv
        Page v
        Page vi
    Table of Contents
        Page vii
        Page viii
        Page ix
        Page x
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
    Materials and methods: Cranial morphometrics
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
    Materials and methods: Dentary morphometrics
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
    Results: Cranial morphometrics
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
    Results: Dentary morphometrics
        Page 122
        Page 123
        Page 124
        Page 125
        Page 126
        Page 127
        Page 128
        Page 129
        Page 130
        Page 131
        Page 132
        Page 133
        Page 134
        Page 135
        Page 136
        Page 137
        Page 138
        Page 139
        Page 140
        Page 141
        Page 142
        Page 143
        Page 144
        Page 145
        Page 146
        Page 147
        Page 148
        Page 149
        Page 150
        Page 151
        Page 152
        Page 153
        Page 154
        Page 155
        Page 156
        Page 157
        Page 158
        Page 159
        Page 160
        Page 161
        Page 162
        Page 163
        Page 164
        Page 165
        Page 166
        Page 167
        Page 168
        Page 169
        Page 170
        Page 171
        Page 172
        Page 173
        Page 174
        Page 175
        Page 176
        Page 177
        Page 178
        Page 179
        Page 180
        Page 181
        Page 182
        Page 183
        Page 184
        Page 185
        Page 186
        Page 187
        Page 188
        Page 189
        Page 190
        Page 191
        Page 192
        Page 193
        Page 194
        Page 195
        Page 196
        Page 197
        Page 198
        Page 199
        Page 200
        Page 201
        Page 202
        Page 203
        Page 204
        Page 205
        Page 206
        Page 207
        Page 208
        Page 209
        Page 210
        Page 211
        Page 212
        Page 213
        Page 214
        Page 215
        Page 216
        Page 217
        Page 218
        Page 219
        Page 220
        Page 221
        Page 222
        Page 223
        Page 224
        Page 225
        Page 226
        Page 227
        Page 228
        Page 229
        Page 230
        Page 231
        Page 232
        Page 233
        Page 234
        Page 235
        Page 236
        Page 237
        Page 238
        Page 239
        Page 240
        Page 241
        Page 242
        Page 243
        Page 244
        Page 245
        Page 246
        Page 247
    Biographical sketch
        Page 248
        Page 249
        Page 250
        Page 251
Full Text






Copyright 1982


Kenneth T. Wilkins

Dedicated to

Christine Nasie Wilkins


Roy and Gwendolyn Wilkins


This doctoral program has followed a tortuous course strewn

with numerous obstacles. Many people have aided my progress through

the labyrinth in many ways. To these people I extend my gratitude.

I thank the faculty members serving on my supervisory/examining

committees: Drs. David Webb, Charles Woods, Jon Reiskind, Steve

Humphrey, Ron Wolff, and Jack Ewel. It was through brainstorming

with Drs. Webb and Humphrey that this project was developed. Drs.

Webb, Woods, and Humphrey offered much in the way of guidance and

counseling on matters related and unrelated to dissertation matters

during my tenure as student at the University of Florida. Jack

Ewel and Steve Humphrey started me on the road toward understanding

Florida's ecosystems, without which a project such as this could

not be properly addressed.

The Department of Zoology and the Florida State Museum have

provided many forms of indispensable support including stipends,

office and lab facilities, and research supplies. Howard Converse

has earned for himself the title of "master procurer" for locating

many pieces of essential research equipment not available anywhere

else at the University. The National Geographic Society, Southern

Regional Education Board, and Sigma Xi funded various aspects of

this research endeavor. Others who have aided my work in various

ways (e.g., in field work, discussions, providing information, and

in helping me maintain my mental stability, etc.) include Mary

Ellen Ahearn, James Bain, Annalisa Berta, Dick Franz, John

Hermanson, Pam Johnson, Bruce MacFadden, Gary Morgan, Ann Pratt,

Terry Zinn, and others.

Curators at numerous institutions graciously permitted me to

examine Recent and fossil specimens in their collections: American

Museum of Natural History, Archbold Biological Station, Carnegie

Museum of Natural History, Delaware Museum of Natural History,

University of Florida (Florida State Museum), University of South

Florida, Florida State University, Harvard University (Museum of

Comparative Zoology), University of Kansas (Museum of Natural

History), Michigan State University (The Museum), University of

Michigan (Museum of Zoology), Midwestern State University, Albert

Schwartz private collection, Tall Timbers Research Station, Texas

Memorial Museum, and the United States National Museum of Natural

History. During my travels to various museums, many individuals

generously provided me with lodging; I thank, in particular, Gary

Morgan, John Hermanson, Duke and Barbara Rogers, and Lloyd Logan.

Dr. James N. Layne (Archbold Biological Station) was particu-

larly helpful in discussions of Florida pocket gophers; he gladly

made available to me his field notes, files, and insights. Several

colleagues assisted in the karyological and biochemical aspects of

this study. These include Dr. C. W. (Bill) Kilpatrick, Debra

Boles, Tom Dowhan, and Mark Allard (University of Vermont); Drs.

John Bickham and David J. Schmidly, and Robert C. Dowler, Duke S.

Rogers, and Prilla K. Tucker (Texas A&M University); and Dr. M.

Raymond Lee and William S. Modi (University of Illinois). Northeast

Regional Data Center (University of Florida) provided computer time

for data analyses.

Lastly, and most significantly, I gratefully acknowledge the

constant moral and financial support of family members: my wife,

Christine Nasie Wilkins, my parents, Roy and Gwen Wilkins, and my

grandmother, Mrs. Nell Stafford Browne.



ACKNOWLEDGEMENTS . . . . . . . . ... . . iv

ABSTRACT . . . . . . . . ... . . . . ix

INTRODUCTION . . . . . . . . ... . . . 1
Distribution. . . . . . . . . ... . .. 4
Habitats Preferred. . . . . . . . . . 8
Habitat Mosaic. . . . . . . . . ... ... 15
Habitats of Florida's Past: Palynological Evidence . 20
Correspondence of Geomys Distribution with Geomorphology. 22
Correspondence of Geomys Distribution with Soils. ... . 24
Sealevel Changes. . . . . . . . .... . 25
Analytical Approach . . . . . . . . ... 30

Ageing. . . . . . . . . ... ...... 33
Measurements. . . . . . . . . ... . . 34
Natural Region and Area Definitions . . . . .. 35
Analyses Conducted. . . . . . . . . ... 40
Multivariate Statistics Rationale . . . . ... 44

Measurements. . . . . . . . . ... . . 50
Materials Examined. . . . . . . . . ... 51
Procedural Rationale and Analyses Conducted ...... 54

Non-Geographic Variation. . . . . . . . ... 62
Geographic Variation. . . . . . . . . ... 68
Summary of Results. . . . . . . . . ... 120

Non-Geographic Variation. . . . . . . . ... 122
Overall Relationships between Samples . . . ... .127
Geographic Variation. . . . . . . . . ... 143
Chronological Variation . . . . . . . ... .147
Summary of Results. . . . . . . . . ... 166

DISCUSSION . . . . . . . . ... . . .168
Zoogeography. . . . . . . . . .. . 168
Phylogenetic Trends: Anatomical Considerations ... .189
Summary of Discussion. . . . . . . . .. .195

LITERATURE CITED . . . . . . . . .. . . 199


APPENDICES ................. ...... 209




BIOGRAPHICAL SKETCH. . . . . . . . . ... 248


Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy




May 1982

Chairman: S. David Webb
Major Department: Zoology

Cranial and dental features were examined in Recent and fossil

pocket gophers (Rodentia: Geomyinae) with the objective of docu-

menting their nearly two million year zoogeographic history in

Florida. Movements of these fossorial rodents is inhibited by

unsuitable habitats and by water. Multivariate statistical methods

were used to determine the importance of rivers as barriers between

populations as exhibited by morphological differences between

samples. The Apalachicola and Suwannee Rivers, respectively, were

found to be the two most important barriers to Geomys movement in

Florida. The relative inhibitory influence of these rivers corre-

sponds to features including proximity of suitable habitat to the

river, regional topography and soil types, width of the watercourse,

and volume and rate of water flow.

The distribution of Geomys in Florida has been strongly influ-

enced by repeated Pleistocene and Recent sealevel changes. At low

sea stands Geomvs has been widespread with extensive gene flow

throughout the peninsula. However, opportunity for divergence

of populations has existed during transgressions when populations

have been isolated in refugia of higher elevations. Studies of

geographic variation between Wisconsinan (low sealevel) samples

revealed no significant morphometric differences, whereas variation

between present-day (higher sealevel) samples from the same penin-

sular areas is significant.

Chronological trends in individual characters from the early

Irvingtonian to present are generally towards increasing size. A

phylogenetic trend in increasing strength and efficiency of the

masticatory apparatus is indicated, especially by the enlargement

of the retromolar fossa. Multivariate analyses show a correspon-

dence between geological ages of samples and their phenetic simi-


The preferred habitat of Geomys pinetis in Florida was identi-

fied as the longleaf pine-turkey oak sandhill association. Geomys

less often occurs in xeric hammocks, longleaf pine flatwoods, and

sand pine scrub. Geographic and chronological distributions of

Geomys in Florida are provided.


Isolation is one of several factors leading to genetic

divergence of populations, oftentimes resulting in subspeciation

or speciation. Because of their restrictions to suitable soils,

many pocket gopher species (Rodentia: Geomyidae) are patchily

distributed. Pocket gophers epitomize the effects of isolation

in terms of recognizable geographic forms. The effects of isolation

on pocket gophers are perhaps best seen in the 229 subspecies of

Thomomys umbrinus, the southern pocket gopher, many of which are

restricted to one or a series of mountain peaks (Hall, 1981:469-496).

Although some biologists (see Anderson, 1966:189) berated the offi-

cial recognition of so many forms, phenotypic divergence of popula-

tions is indeed visible. Stephen Durrant (University of Utah) could

name, solely from visual inspection of a specimen, the particular

mountain peak in Utah from which a pocket gopher specimen was


An analogy exists between Thomomys mountaintop isolation and

the mosaic distribution of Geomys pinetis, the southeastern pocket

gopher, in Florida. The Florida "mountaintops" are parcels of

suitable habitat slightly elevated above surrounding wetter habitats

that are impassable to pocket gophers. Relict dunes and other fea-

tures of former shorelines comprise much of the range inhabited by

Geomys pinetis. At least as early as 1939, Hubbell and Goff (1939:

131) recognized the "discontinuous distribution" of pocket gophers

"in the state" as being "the result of the 'patchy' occurrence of

areas having deep, well-drained soils, to which pocket-gophers are

almost restricted in the region." They credit Harley B. Sherman

with having previously recognized this relationship, and further

state that "it would not be surprising . to find that specific

or racial differentiation corresponding to the degree of isolation

had occurred" in Geomys pinetis (Hubbell and Goff, 1939:134).

The main thrust of other recent studies of geographic variation

in the Geomys pinetis species complex (Avise et al., 1979; Williams

and Genoways, 1980) has been towards systematic revision of the group.

The present study, which relies on a similar morphometric data base,

examines zoogeographic aspects of the topic. One objective of this

study is to examine geographic variation in Recent Florida pocket

gophers and to evaluate the effects of isolation on these populations

in terms of morphological divergence. It is hypothesized that the

pattern of geographic variation will correspond to the distribution

of suitable habitats, and that the extent of divergence between

populations will correspond to the magnitude of various physiographic

features (e.g., rivers, unsuitable habitats) separating populations.

The fossil record of pocket gophers in Florida spans nearly two

million years. Beginning with the early Irvingtonian land mammal age,

fossil material is known in at least 37 deposits from more than 27

localities collectively representing the Irvingtonian, Rancholabrean,

and Recent ages. Fossil specimens occur primarily as dentaries and

isolated teeth with fewer fragments of the anterior portion of the

cranium. Isolated postcranial elements also occur, but are not

treated in this study. The material has been attributed to two

species: Geomys pinetis, the only extant geomyid in Florida, and

Thomomys orientalis, an extinct species representing a genus now

restricted to central and western North America (Simpson, 1928).

This record of Thomomys consists of but a single rostromaxillary

fragment (AMNH 23441) from Sabertooth Cave (Citrus Co.). However,

additional material thought to be Thomomys has been recognized in

this study from the Rock Springs site (Orange Co.) and from Williston

III B (Levy Co.). The present study addresses only Geomys material.

Study of Florida pocket gopher material is intriguing for at

least two reasons. First, geomyid occurrence in Florida is rather

well represented in the fossil record, beginning in the early

Irvingtonian, thereby allowing examination of various phylogenetic

trends characterizing geomyid evolution (Russell, 1968). Particular

emphasis is here accorded to changes in the basitemporal (= retro-

molar) fossa. Second, because Florida has experienced repeated

cycles of sealevel rise and fall (in response to fluctuations of

continental glaciers), the effects (i.e., morphological divergence)

of isolation of central peninsular forms from peripheral peninsular

forms during interglacial periods might possibly be visible in both

the qualitative and the quantitative characters of teeth and den-

taries. The chronological and spatial distribution of Pleistocene

samples comprised of measurable dentary material is such that both

geographic and temporal variation can be separately examined within

the Florida peninsula.


Before a recent taxonomic revision of pocket gophers of the

southeastern United States (Williams and Genoways, 1980), four

species were recognized: the wide-ranging Geomys pinetis and the

narrowly-distributed forms G. colonus, G. cumberlandius, and G.

fontanelus. Williams and Genoways (1980) now consider all forms in

this species complex to be G. pinetis. The general distribution

of G. pinetis in the Atlantic and Gulf coastal plains of Alabama,

Florida, and Georgia. A detailed distribution of the species in

Florida is shown in Fig. 1. Specimens are known from all Florida

panhandle counties, except Gulf Co. where diligent field efforts

would probably reveal a few isolated colonies. The range extends

through approximately the northern two-thirds of the peninsula to

a southern tier of counties (Brevard, Osceola, Highlands, DeSoto,

and Manatee). Thus, the occurrence of G. pinetis in Florida is

much more extensive than depicted in Hall (1981:505) or in Hamilton

and Whitaker (1979:173). I have been unable to substantiate the

"records" of Geomys pinetis in St. Lucie Co. (vicinity of Ft.

Walton), the southeasternmost localities reported in Williams and

Genoways (1980). Although a town named "Walton" does occur in St.

Lucie Co., it is more likely that these specimens were taken in

Okaloosa Co. in the Florida panhandle.

The possibility exists that Geomys pinetis distribution is

(or within the last century might have been) more extensive than

shown on the range map (Fig. 1), which indicates only those locali-

ties documented by voucher specimens. Bangs (1898:176) noted

Fig. l.--Recent distribution of pocket gophers (Geomys
pinetis) in Florida. Dots represent localities for which
museum specimens were examined.



o 10 25 50 100

reports of "hills" of Geomys being seen south of Micco, a mainland

town on the intracoastal waterway in southeasternmost Brevard Co.

Micco is about 32 km south of the Melbourne-Eau Gallie area that

is the type locality of G. p. goffi and comprises the southernmost

documented occurrence of Geomys along the Atlantic coast.

Hubbell and Goff (1939:134) reported that in parts of southern

Florida "an 'archipelago' of turkey oak and sand scrub 'islands'

extends to the vicinity of Palmdale and Punta Gorda" located in

Glades Co. and Charlotte Co., respectively. They further noted

that "pocket gophers are present on some of these as far southwest

as the latter locality" (i.e., Punta Gorda). Apparently no voucher

specimens exist from any localities in Glades Co. or Charlotte Co.

During March 1981, I drove many kilometers of roadway in south-

central Florida in efforts to better delimit the southern extremity

of occurrence of Geomys in the state. Twice we covered the route

from Palmdale to Cleveland (Charlotte Co.) via Highway 74, thence

northward via Highway 17 to Fort Ogden (DeSoto Co.). No evidence

of Geomys in either Charlotte Co. or Glades Co. was found, although

in about a dozen places both east and north of Cleveland and just

south of Fort Ogden we observed lines of sandy mounds in lawns,

pastures, and along roadsides. Excavation of portions of several

mound systems revealed them to be the workings of moles (Scalopus

aquaticus). Tunnels were either immediately beneath or within

10 cm of the surface. Tunnel diameters were less than 5 cm.

Spoil piles were roughly conical and composed of plugs of earth

characteristic of mole diggings. These signs of mole activity

could easily be mistaken for pocket gopher diggings. On this basis

I prefer to discount "records" from these and other counties

located on the periphery of the documented range of G. pinetis

pending procurement of voucher specimens.

Numbered localities in Fig. 2 designate sites from which fossil

geomyid material is available; site numbers correspond to numbers

in Table 1. Past and present ranges of Geomys in Florida overlap

broadly. The occurrence of Geomys material at Vero, Indian River

Co., demonstrates that Geomys previously enjoyed wider distribution

than at present. This is in contrast to Kurten and Anderson's

(1980:229) statement that the Vero site is "within the modern

range of the species." The one non-peninsular Florida locality

yielding fossil geomyid material (a single lower premolar) is the

St. Mark's River, Wakulla Co., site (Gillette, 1976).

Habitats Preferred

Geomys pinetis has been reported from a variety of habitats

differing extensively in terms of characteristic flora but being

similar in regard to the sandy, well-drained nature of the soils.

References detailing such habitat information include Harper (1927:

336-342), Hubbell and Goff (1939:130-133), Quay (1949:67), Golley

(1962:105), and Ehrhart (1978:6-7). My personal observations

identify four primary habitats used by G. pinetis and listed here

in descending order of preference: sandhill association, xeric

hammocks, longleaf pine flatwoods, and sand scrub areas.

Fig. 2.--Pleistocene pocket gopher localities. Legend to
site numbers included in Table 1. Dashed line approximates
southern extent of modern distribution of Geomys pinetis.





01025 50 tIO
Cs 3-


C r :O -t 1 N z--o-r mCm 0 00 D o .00 flOO
0) Z H N M '0

m 0

ca .0
o Ce
0H 1E -a C aaCC a CCaCaC
c- CCH a 14 HaHa-a1a a4
Cm- ) ) CO E U ) ) r 0 0 0 9
OW Wa a mOamm mW C CC CCCCCC
0 r CC C .H H H *H w) 0 000000
So M me a 33333 33 iEE e E e E
u o u u u 0 u w a 0W 0 0 0
0 0 0 w w owwa 00 ru 0 ac to wa aa 00 m
C) M CO 1-4 W 4J 4J _qI B4 m r Cf rm CrC C r C C
- H 0 0 0 c v a o 0 0 m 0H m
S S3; r ilil X cl &m 3 t m m m Em m mU
-H C
H) C
:1 C) w U wa w a o w a)Ca 0o a) w awawawawawa
C0C an .0) p P C) P P P p w p pi-i C pC) p0 w p
Ha C) Ha r-4 a 0 0 c wa )C Cc C ca ca cw C a cc
4 41 -40 00000 0 0 0 0 0 000
C a Z u w) 9 C 0 Ca0 aaa 0Ca C 0 C0 (0 (0 01

0 0
Wa)ca -,4' cc C occcacccc

.C a4- H-

ma a ) p

C )H H H
o > cr t o 00 U
I. 0 -r 1 0 ( C C0 a0 ( C %
0) C CH .C )CC 0u C 00 CCCa) 00

00' 0 0 0) 6 0 0 0 3

C1 > 4r U H) U H "H lO ? H
1- 0 U 0* WHaCC C 0 > 44 f Ma U C)))a>
H m Cma 0 0 o> C Ha la ) H HH 4( a ( a l
Ca C 3 i- C-ia P..i-> <.-4 u IV > )u

00 H a
o C) L 0) -.4 cc h- CO H i O

4 m)H r 0
U wa U CL J H HH HH HH > le U P. 4

C H H 3u"

00 H C>C CEw H I H>0 -4

ro a) > wa UC ca Ca HH *O Q < w0 c c
H C > E Z H x0'0 > > XtC HH ] <

'Ca C iD -)0 ) ri -1 0 'ZH <. C OM
H (U m w a 0 Ca [n (O c X 0 HH C)C < ( Ca

HZ Cn >cn 0 HZocn> CC3 mC


0 ~0 0 L ID 0 1


Q) 1- c c

rn u uu u u
U) I H L C)I~ c C)LIL
c fa w wIcar co wa 0
aWC Q) a) (V )C)C)Wa)

CC~ C4))CCC r-

X = 1-4 -,f 1 1-

ca o o o C 4 4 W - Q

M u v)
cd cd ca ca ca l
X) 0 r)CqC=Ca ) 14X ) C) co U

u H a)u a C -Hca
< 0 0 0 r. w cd u ra

cd H H HHH ca--H :3 14 -H ca U

u 0. '14 A4 'C4 u~ m -:4c
CC 0)
C ) )

erca r rA

u oo (U
a) C C CC) -r4 o Q
oo 0 c to > 0

P m lo r o
0 )u c a 4-4 C:~

"U z ) a) cu Q)CC)E
4.4 u CUC C> C CU aH4 o r
0a-, ca 0H ca a ) C) c 0 ca r_ CU-
W w c :3: U I-4 K
U ~ ~ ~ ~ ~ ~ I H -C Ur)U C C 4 C
H CU H C1 H *- CI
~ CC) ~ U

0-f C4 mHt~1'- la r0C uC

H~ -4:
z CCC.~
~Oarg- )
a,~~ -ZmuU

0 Hr C)gd C UI-C


v, COC CC] H CC)

Prime pocket gopher habitat is the sandhill association of

the co-dominants longleaf pine (Pinus palustris) and turkey oak

(Quercus laevis), and a fire-perpetuated ground cover of grasses

(especially wire grass, Aristida stricta), forbs, and sedges from

which G. pinetis selects its diet. Terrain in the sandhill eco-

system is rolling and the sandy soils are well-drained. Xeric

hammocks are dominated by live oaks (Quercus virginiana) and a

variety of other hardwood species; soils are moister and contain

more organic material than in sandhill or sand scrub situations.

Longleaf pine dominated one of the three types of pine flat-

woods in Florida before intensive land clearing activities (for

cultivation, timber, etc.) removed most virgin stands (Quarterman

and Keever, 1962; Monk, 1968). Although some longleaf pine flat-

woods still remain, the faster-growing slash pine (Pinus elliotii)

preferred by commercial foresters has supplanted longleaf pine in

most flatwoods. Flatwoods are topographically flat areas subjected

to seasonal flooding. Soils are sandy and usually moist enough to

exclude pocket gophers. The characteristic dense understory of

palmetto (Serenoa repens), apparently not suitable for pocket

gopher digging or feeding, further contributes to the undesirable

nature of flatwoods for Geomys habitation. In at least two long-

leaf pine flatwoods situations I have either collected specimens

(2.5 miles E Woodville, Leon Co.) or seen pocket gopher sign

(Morningside Nature Center, Gainesville, Alachua Co.; see also

Ross, 1976).

The sand scrub association is characterized by a very dense

forest dominated by sand pine (Pinus clausa), palmetto (Sabal

etonia), and an assemblage of scrubby oaks and other angiospermous

shrubs and trees. Because ground cover provided by shrubs is so

dense, forbs and grasses suitable for Geomys diet are not abundant.

Hubbell and Goff (1939:133) remarked that the pocket gopher cannot

be regarded as a "characteristic scrub-inhabitant." Generally,

ground not covered by shrubs and trees is heavily exploited by mats

of xerically-adapted fruticose lichens. Soils in the sand scrub

ecosystem contain less organic material and are by far the driest

of the four associations discussed here.

The role of fire in maintaining many ecosystems (e.g., sandhill,

flatwoods, sand scrub) is well-documented (Monk, 1968:444-445; Veno,

1976:498). Man's activities over the past century in Florida have

resulted in control of wildfire. As a result, the role of fire has

been greatly diminished. Removal of fire allows fire-intolerant

hardwoods to gain a foothold in ecosystems dominated by fire-tolerant

species (e.g., many pines), thereby competitively excluding the

pyroclimax communities requisite for Geomys pinetis habitation.

The pyroclimax community includes many herbaceous species

comprising the diet of Geomys pinetis. Pocket gophers may be

locally extirpated as a result of loss of these food sources. The

general trend, then, is a reduction in the amount of land area suitable

for habitation by pocket gophers. An example of such exclusion of

Geomys due to removal of fire may be developed from field studies

in Levy Co., Florida. James N. Layne (pers. comm.) has collected

field data on diversity and density changes in the mammalian fauna

of sand ridges between Rosewood and Cedar Key (Levy Co.) from at

least the early 1960's through his most recent visit to the area in

1979. His field notes for the years 1960-1963 indicate Geomys to

be present in fairly continuous populations in the sand pine habitat

of these relict dunes and along the roadside (Highway 24) between

Rosewood and Cedar Key, a distance of about 11 km. Notes from 1965

to present either do not mention the presence of Geomys in the scrub

areas or explicitly note their absence. My observations in Fall

1980 and Spring 1981 along this transect are that (1) the town of

Rosewood, situated in an area of longleaf pine-turkey oak habitat,

supports a thriving population of Geomys, but that (2) no Geomys

now occur along Highway 24 from outside Rosewood to Cedar Key. I

have not searched the ridge systems adjacent to the highway for

Geomys activity; however, the visible scrub vegetation is so dense

as to seem unfit for pocket gopher occupancy. Layne attributes the

disappearance of Geomys from this habitat to the removal of fire

from the ecosystem in the last two decades. Wildfires during the

last two years might have opened this general area up to re-invasion

by pocket gophers.

Habitat Mosaic

The distribution of the sandhill association, one of the four

general habitats that Geomys pinetis occupies in Florida, is out-

lined in Fig. 3 (modified from Davis, 1980). Dots represent






ri '




0 c
4 q

0 '
C, '


44 4



44l C

---? z

Y h

localities (or clusters of localities) from which voucher specimens

of pocket gophers are known. Over 125 of the more than 150 locali-

ties plotted coincide with or are at the edges of parcels of the

sandhill association. Most of the remainder occur in xeric ham-

mocks. Only a few records of pocket gophers are available for the

sand scrub regions located sporadically on sand ridges along the

Atlantic and Gulf coasts and in the interior of north-central

peninsular Florida. The pine flatwoods ecosystem is the predominant

habitat in the state (Monk, 1968:444). In most cases, parcels of

the other three associations are bounded on all sides by pine flat-

woods'. Only a few pocket gopher specimens are known from the pine

flatwoods of Florida.

Fig. 3 is a composite of the general distribution of sandhill

habitat overlaid by specimen localities. The coincidence of a

collecting site with a certain habitat need not always correctly

indicate the actual microhabitat inhabited by pocket gophers. As

an example, Ocala National Forest is eastern Marion Co. is depicted

on this map (Fig. 3) as being characterized by single parcels of

two distinct habitat types: sand scrub and sandhill. The mis-

leading nature of this illustration is apparent to those who have

visited Ocala National Forest. Sand scrub is the predominant

association and is probably more extensive than depicted. The

sandhill association is present also, but is not distributed quite

as depicted; rather than comprising the entire northeastern section

of the forest, longleaf pine-turkey oak stands (and some xeric

hammocks) occur as isolated parcels sprinkled through broad areas

of sand scrub. Such parcels of sandhill habitats are referred to

as "islands" in the local vernacular. It should be noted that the

sites in which essentially all Geomys have been collected (or most

sign has been observed) in Ocala National Forest coincide with such

sandhill islands. Additionally, the range of G. 2, goffi (a taxon

no longer valid: Williams and Genoways, 1980) along the Pineda

Ridge in the Melbourne-Eau Gallie area of eastern Brevard Co. is

depicted on the Davis (1980) vegetation map as sand scrub surrounded

by pine flatwoods. Personal visits to this area (October 1978)

revealed the sandhill and xeric hammock associations to be present


From the preceding examples it is evident that microgeographic

distribution of habitats in Florida is a mosaic. On a broader scale

the statewide distributions of these habitats are also in a mosaic

pattern (Fig. 3). Parcels of sandhill, xeric hammock, flatwoods,

and sand scrub are all widely distributed throughout the state,

but in a disjunct manner. Areas of a particular habitat are iso-

lated from other areas of that same habitat by features that may

form partial or complete barriers to dispersal of species that are

adapted and restricted to the isolated habitat types. In the case

of pocket gophers, these intervening barriers may take the form (1)

of rivers or streams (Udvardy, 1969:14), or (2) of substrates

unsuited to digging because of excessive moisture, hardness of soil,

or shallowness or lack of soil, or (3) of communities lacking vegeta-

tion required in the diets of Geomys (Winchester and DeLotelle,

1978). In many species of plants and animals, and notably in other

species of pocket gophers (Wright, 1943; Kennerly, 1954, 1963;

Patton and Dingman, 1970; Patton et al., 1972; Honeycutt and

Schmidly, 1979; Nevo, 1979), such isolation corresponds to patterns

of geographic variation often to the point of defining subspecies

or species. Hence, in Florida the opportunity for allopatric

differentiation exists for pocket gophers in spite of what appears

to be fairly uniform topography.

Habitats of Florida's Past: Palynological Evidence

A direct record of Pleistocene vegetation conditions in Florida

is provided by pollen samples from lake sediments. These pollen

records clearly indicate the waxing and waning of habitats more and

less favorable to pocket gophers. Two of the most informative.pollen

cores have been taken from Lake Annie in south Florida and Sheeler

Lake in north central Florida (Watts, 1980; Watts and Stuiver, 1980).

Over the last 40,000 years dramatic changes have occurred in vege-

tation in the vicinities of these lakes. The general climatic

trend indicated from Wisconsinan times into the Holocene is from

colder, drier conditions to the warmer, moister situations of today;

numerous minor excursions from this general trend are evident as well.

For example, the following vegetation associations are recorded in

sequence from the vicinity of Sheeler Lake:

24,000 18,500 B.P. open pine forests with broadleaf
trees and localized patches of prairie
and sandhill herbs interspersed;
18,500 16,600 B.P. hiatus, probably due to lake's being
empty during an extreme dry period;
16,600 14,000 B.P. dry oak-hickory with localized prairie;
14,000 13,000 B.P. mesic broadleafed trees;
13,000 9,500 B.P. oak and prairie plants;
9,500 7,200 B.P. pine forests mixed with oak plus swamp,
cypress, and bayhead species.

The pollen profile from Lake Annie, extending from 44,000 B.P. to

present, differs greatly from that of Sheeler Lake in many of the

species and vegetation associations documented, but still demonstrates

the general trend from cold, dry to warm, moist conditions. Such

differences are easily reconciled when discrepancies existing in

present-day floras are considered.

Two conclusions may be drawn from the palynological data.

First, changes in rainfall and temperature regimes correspond to

glacial phases, and hence to sealevel changes. In Florida, glacial

periods tend to be cold and dry, whereas interglacials are warmer

and moister; variations on this theme correspond to geographic

location in the peninsula. Second, vegetation associations dominating

localized areas are subject to swift, drastic change in response to

climatic variation. For this reason, the suitability of local areas

for habitation by various species (e.g., pocket gophers) has likewise

fluctuated through time. Because of known habitat preferences of

pocket gophers, it is conceivable that detailed chronological varia-

tion in the range of pocket gophers in Florida could be inferred if

pollen diagrams were available from throughout the state. Even

with the present limited data, pollen cores support the general

view that pocket gopher habitats were more extensive and surface

water less extensive during glacial maxima, and that the reverse

was true during interglacial stages (including the Recent).

Correspondence of Geomys Distribution with Geomorphology

The distribution of pocket gophers tends to follow certain

physiographic features in various parts of the state. Two examples

follow below. In the southwestern peninsula (Manatee, Sarasota,

and DeSoto counties), the range boundary of G. pinetis corresponds

closely with the edge of the DeSoto Plain (White, 1970: map 1-C

and Fig. 43). Personal field trips to the area in May 1980 and

March 1981 revealed such localities in south-central Manatee Co.

(junction of Highways 64 and 675), southern DeSoto Co. (Fort Ogden),

extreme western DeSoto Co. (sporadic occurrence of small colonies

along Highway 72 and in adjacent fields and pastures within 5 km

of the Sarasota Co. line; no vouchers known), and a suspected sight

locality in NNE Sarasota Co. south of Verna. Proceeding south and

west off the DeSoto Plain, the habitat changes from the sandhill

association to lower, wetter pine flatwoods of the Gulf Coastal

Lowlands from which no pocket gophers are known.

The pattern of occurrence of G. pinetis in Osceola Co. and

eastern Polk Co. is another example of correlation of distribution

with physiographic landforms. Most of Osceola Co. and extreme

eastern Polk Co. consists of flat lands (the Osceola Plain)

supporting primarily pine flatwoods and wet prairies, but also

cypress forests, swamp forests, and an occasional isolated sand

scrub ridge (Davis, 1980; pers. obs.). Highway 441 tracks the more

highly elevated portions of eastern Osceola Co. Geomys vouchers

indicate their occurrence in Osceola Co. only at or near Kenansville,

Nittaw, Illahaw, Mahaw, and St. Cloud, all in central or northern

parts of the county. Within the last five years, James N. Layne

(pers. comm.) says he has seen Geomys diggings in the vicinity of

Yeehaw Junction in extreme southern Osceola Co. As determined from

personal observations in March 1981, the habitat in the immediate

vicinity of Yeehaw Junction, now developed into citrus groves,

appears suitable for pocket gopher habitation although we saw none.

This town is located in an area coinciding with a small parcel of

sand scrub on the Davis vegetation map (1980). To my knowledge,

no vouchers are available from Yeehaw Junction. The closest docu-

mented Geomys locality is about 20 km to the north in Kenansville.

The route from Yeehaw Junction westward into Lake Wales (Polk

Co.) along Highway 60 traverses flat terrain and habitats excessively

wet for Geomys until a point near Indian Lake Estates (Polk Co.)

near the junction of Highways 60 and 630. Here, about 12 km west

of the Osceola Co. line where the highway ascends the "Bombing

Range Ridge" (White, 1970: map 1-C), we observed three or four

lines of Geomys mounds along the roadside and in lawns for a 0.8 km

stretch. The sandhill habitat of the ridge levelled and then

dropped into pine flatwoods devoid of Geomys sign for about 13 km

past which we ascended a second scarp, Lake Wales Ridge (White, 1970:

map 1-C) into Hesperides. These two ridges, at about 35 m elevation,

represent the Wicomico shoreline (MacNeil, 1950: plate 19). Although

no Geomys sign was noted for about an additional 11 km, the rolling

sandhill habitat (mostly developed into citrus groves) appeared

suitable. The Geomys activity closest to Hesperides was found 3 km

east of Lake Wales. Upon entering the city of Lake Wales, situated

in the Polk Upland (White, 1970: map 1-C), elevation increased to

about 50 m. This elevation corresponds to the Okefenokee shoreline

(MacNeil, 1950: plate 19).

Correspondence of Geomys Distribution with Soils

The general soils map of Florida (Fl. Agric. Exp. Sta., 1962)

illustrates the distributions of approximately 37 soil associations

in the state. These are classified into five major groups according

to drainage characteristics. The only soil association not repre-

sented within the overall range of G. pinetis in Florida is rockland

(type number 31), which is most common in the Everglades region.

A general correspondence between Florida pocket gopher habitats and

their component soils is evident by comparing distributions of soils

with vegetation (Davis, 1980).

The most highly preferred habitat (sandhill) and xeric hammocks

occur in the same well to moderately well drained soil types. The

names of these associations and their identifying numbers from the

1962 soil map are (3) Lakeland-Eustis-Blanton, (3a) Lakeland-Eustis-

Norfolk, (4) Jonesville-Chiefland-Hernando, (9) Hernando-Chiefland-

Jonesville, (5) Arredondo-Gainesville-Fort Meade, (8) Hague- Zuber-

Fellowship, (12) Blanton-Klej, (12b) Kanapaha-Blanton, (17a) Rex-

Blanton, (17b) Blanton-Bowie-Sequhanna, and (17c) Goldsboro-Lynchburg

associations. Inland and coastal sand scrub ecosystems are found on

two excessively drained soil associations: (1) St. Lucie-Lakewood-

Pomello and (2) Palm Beach-Cocoa. Pine flatwoods occur in areas

dominated by somewhat poorly drained soils: (13a) Leon-Plummer-

Rutledge, (13b) Leon-Immokalee-Pomano, (13c) Leon-Pomello-Plummer,

and (13d) Leon-Blanton-Plummer associations.

Pocket gophers apparently occur or have recently occurred (at

various densities) in all of the above soil types in Florida.

Variation between these soil types occurs in pH, organic content,

moisture content, and size of component sand grains. However, all

of these soils share a feature important in determining suitability

for pocket gopher inhabitation: all are sandy and, therefore, are

suitably friable for pocket gophers. Differences in abundance of

pocket gophers in these soil associations appear not to be due to

soil qualities per se, but to other parameters such as soil moisture

and associated vegetation types.

Sealevel Changes

The primary mechanism governing the degree of isolation of

populations of Geomys pinetis is tied to the role of the oceans as

both source of glacial ice and as sink holding glacial melt waters.

Sealevel changes during the Neogene period (Miocene through present)

resulted from fluctuations in the extent of worldwide glaciation.

The amount of emergent land characterizing Florida during the

Pleistocene varied with the magnitude of continental glaciers and

the extent of their melting. During the high stands of earlier

interglacials (i.e., those corresponding to the Coharie, Okefenokee,

and Wicomico shorelines) seas covered most of the peninsula, leaving

only an archipelago of islands to represent present-day peninsular

Florida. Pocket gophers, if present on such islands at these times,

were isolated from other such populations, and thereby were presented

with opportunities to differentiate from each other. However,

during ensuing glacials, sealevel fell well below the present level

(by 100 m or more), thereby reconnecting islands with each other.

With sea barriers so removed, habitats were once again in contact.

Potentially, and quite probably, pocket gopher populations expanded

and established a peninsula-wide interbreeding population. Such a

population would likely have been phenotypically homogeneous.

Thereby, any differentiation achieved during isolation (if it had

not proceeded to the point of reproductive isolation) would likely

be obliterated. The high sea stand during the middle of the Wisconsin

glaciation (the Pamlico at 9 m) again formed a habitat mosaic probably

inhabitated by demes of pocket gophers. But divergence amongst these

populations was likely swamped during range extension that accom-

panied the expansion of late Wisconsinan glaciers. At the height

of the late Wisconsinan (about 19,000 B.P.), seas fell to about 90

to 130 m below present (Bloom et al., 1974; Harmon et al., 1978).

Since then, the glaciers have been melting and the sea continuously

rising towards its present level. Isolation of populations has

increased with increasing sealevel.

Because Florida is characterized by low and gentle topographic

relief, even minor sealevel changes have caused notable variation

in the relative amounts of emergent land and water on the Florida

Plateau (present-day Florida and surrounding continental shelf to

a depth of about 100 m). Hence, many authors have investigated

sealevel changes in Florida (e.g., Cooke, 1939, 1945; MacNeil, 1950;

Alt and Brooks, 1965; Scholl and Stuiver, 1967; Healy, 1975).

Evidence has been so variously interpreted by different authors,

however, that little agreement exists on (1) the number of glacial-

interglacial cycles whose effects are visible, (2) the correspondence

of relict shoreline features (e.g., dune systems, scarps, marine

terraces) at different statewide localities, or (3) even the ages

of these shoreline features. Perhaps the sole point of agreement

among these studies is that these shorelines "become progressively

younger as they step down toward present sealevel" (Alt and Brooks,

1965:406). Healy (1975) offers a concise summary and comparison of

findings of Florida shoreline investigations including a substantial

list of pertinent references.

A traditional, composite interpretation drawn from various of

these studies suggests that four glacial-interglacial cycles occurred

during the Pleistocene. In order of occurrence, the four shoreline-

terrace complexes (and approximate elevations above present sealevel)

left at these interglacial maxima are (1) the Coharie (70 m) during

the Aftonian interglacial, (2) the Okefenokee (50 m) during the

Yarmouthian, (3) the Wicomico (35 m) during the Sangamonian, and (4)

the Pamlico (9 m) during the mid-Wisconsinan. A fifth shoreline, the

Silver Bluff at 3 m, is regarded by some as a Recent (6,000-4,000

B.P.) high sea stand during a warmer climatic optimum (MacNeil,

1950:104). This interpretation of the Silver Bluff may be discounted

on two counts. From south Florida data, Scholl and Stuiver (1967)

showed that, at 4,400 B.P., sealevel was about 4 m below present

and has been rising since then at varying rates. Furthermore,

sealevel has been rising since the height of the late Wisconsinan

Woodfordian glaciation between 22,000 B.P. and 12,500 B.P. (Watts,

1980). Additionally, numerous, more recent studies in the vicinities

of New Guinea and the Caribbean Sea have shown that the last time

sealevel exceeded its present height was about 125,000 B.P.; the

estimates for this transgression range from 4 to 10 m (Broecker and

Thurber, 1965; Osmond et al., 1965; Broecker et al., 1968; Mesolella

et al., 1969; Cant, 1972; Bloom et al., 1974; Ku et al., 1974;

Neumann and Moore, 1975; Chappell and Polach, 1976; Harmon et al.,

1978; Marshall and Launay, 1978; and Bender et al., 1979). Of the

shorelines visible in Florida, it is the Pamlico rather than the

Silver Bluff which best corresponds with the 125,000 B.P. transgres-

sion. Therefore, a more reasonable explanation is that the Silver

Bluff might represent a pause during recession of this particular

4-10 m high stand.

Other interpretations of the numbers and ages of these relict

shoreline features have been proposed. Shackleton and Opdyke (1973)

conclusively demonstrated (via oxygen isotope and paleomagnetic

studies) the occurrence of approximately 22 cycles of sealevel

change, rather than the traditional four, during the last 800,000

years of the Pleistocene. Many of these transgressions are not

recorded as emergent shoreline features because many of these high

stands were below present sealevel. All interglacial high stand

terraces mapped and dated in New Guinea since the 125,000 B.P.

4 to 10 m sealevel were below present sealevel (Bloom et al., 1974;

Chappell, 1974). The ages and approximate elevations with respect

to present sealevel of these high sea stand terraces are: 30,000

B.P. (-50 m), 40,000 B.P. 50,000 B.P. (-30 m), 60,000 B.P. (-20

m), 80,000 B.P. (-10 m), and 105,000 B.P. (-10 m). Bender et al.

(1979) extended the record even further, documenting about 10 high

stands between 140,000 B.P. and 640,000 B.P. These interglacial

seas stood from near present sealevel to well below that of present.

The general conclusion, then, is that seas have stood higher than

present level only once in the last 700,000 years. It is very

possible, then, that with regard to ages of these terraces Alt and

Brooks (1965) were correct in assigning only the lowest terrace

(Pamlico) to the Pleistocene. The designate Miocene and Pliocene

ages for the Coharie, Okefenokee, and Wicomico shorelines.

From the proceeding dates for high sea stands, it is evident

that the amount of isolation now seen between populations of Florida

Geomys pinetis is the greatest that has existed since the last time

seas stood higher than present--about 125,000 B.P. when sealevel

was about 4 to 10 m. The earlier Wisconsinan, from its beginning

(at about 300,000 B.P.) to about 125,000 B.P., is characterized by

about five more glacial-interglacial cycles, all having high stands

near present sealevel. Following this 125,000 B.P. transgression,

which apparently corresponds to the Pamlico, seas fell to about -60

m at about 115,000 B.P. and then fluctuated through about five more

cycles before attaining the current level. Estimates of the low

sealevel stand at the maximum extent of the final glacial of the

late Wisconsinan (at about 19,000 B.P.) range from around -90 m

to -130 m (Blackwelder et al., 1979).

Analytical Approach

Both Recent and Pleistocene phases of this study are based on

morphometric analyses of cranial, mandibular, and/or dental charac-

ters. The initial analyses in each phase were directed towards

evaluating non-geographic sources of variation (e.g., sex, age,

and individual variation). This step identified subsets of charac-

ters for which variation between samples exceeded that within

samples. Next, samples were defined such that subsequent analyses

would examine geographic variation with regard to features thought

to comprise zoogeographic barriers. Both univariate (analyses of

variance, Duncan's multiple range tests) and multivariate (clustering,

principal components, canonical variates, multivariate analyses of

variance) techniques were employed for study of geographic and

chronological variation. Patterns of variation elicited are then

discussed in terms of habitat preferences, geomorphology, and the

effects of sealevel changes on pocket gopher distribution over the

last 1.8 million years.

The primary hypotheses examined in this study concern the

interaction of sealevel changes, existing dispersal barriers, and

the degree of isolation of pocket gopher populations in determining

the patterns of morphological variation among these populations.

The hypotheses tested and research directions pursued include the


(1) It is hypothesized that the pattern of geographic variation in

Recent Florida Geomys pinetis corresponds to the magnitude and geo-

graphic occurrence of dispersal barriers with rivers defining inter-

breeding populations. Multivariate analyses of variance testing

inter-area and inter-region variation in morphological features

address this hypothesis.

(2) It is assumed that during the low sealevels of glacial intervals

distribution of pocket gophers in the Florida peninsula was relatively

widespread due to abundant and widespread suitable habitats. Such

broad distribution enhanced gene flow, and populations became

genetically more homogeneous. Genotypic constitution is assumed

to be reflected by morphological characters. It is hypothesized,

then, that morphometric differences between glacial stage samples

(e.g., Wisconsinan) from distant peninsular localities are not


(3) Conversely, the high sea stands of interglacial intervals

(e.g., Holocene or Sangamonian) restricted peninsular populations

into isolated refugia. Reduced gene flow presented the opportunity

for genetic divergence, which may be reflected in morphological

traits. Therefore, it is hypothesized that interglacial stage

samples from distant peninsular localities are significantly

different morphometrically.


(4) Variation in each of 10 dentary characters represented in

pocket gopher material spanning the last nearly two million years

was examined for the purpose of describing morphological changes

characterizing phylogeny of Florida geomyines.


Conventional skin and skull preparations of 1,123 specimens

of southeastern pocket gophers from Florida were examined (listed

in Appendix B). Thirty-three of these specimens were collected from

critical areas during the course of the study. Measuring of

characters was restricted to the adult sample of 854 individuals.

Although juveniles were not measured, their collecting localities

were used to document the distribution of pocket gophers in Florida.


Mature adults were divided into two age groups on the basis

of a suite of osteological features developed from the efforts of

other investigators examining other geomyids: Dunnigan (1967),

Thaeler (1968b), Hoffmeister (1969), Williams and Genoways (1977),

and Honeycutt and Schmidly (1979). Younger adults (age class 1)

and old adults (class 2) were distinguished according to (1) the

degree of fusion of the basisphenoid and basioccipital bones, (2)

the development of the temporal ridges, and (3) the degree of poro-

sity in the maxillary process of the zygoma and the palatine. In

younger adults, the basisphenoid-basioccipital suture is closed

with no interlying gap, but yet not obliterated by accumulated bony

material as in old adults. The temporal ridges in old adults

(especially males) are highly rugose and usually contact each

other to form a single sagittal ridge; rarely if ever is such a

sagittal ridge found in females although, as in males, the distance

between temporal ridges decreases and their degree of pronuncia-

tion increases with age. In both sexes, bone porosity in the

palatine and maxilla decreases with age.

Age, determined as described above, and sex were noted for each

specimen. When sex was not indicated or the assigned sex was

doubtful, cranial morphology was used to designate the sex (see

Merriam, 1895:20-21); however, specimens for which sex designation

via cranial morphology remained ambiguous were deleted from considera-



Descriptions and abbreviations of the 20 cranial features

measured to 0.1 mm with Helios dial calipers for the 854 adults

examined are as follows: greatest length of skull (GLS), from

exoccipital to anterior surface of incisors; greatest zygomatic

width (ZYGO); width across mastoid processes (WMAST); depth of

cranium (DCRAN), dorsoventral distance from top of cranium to

ventral surface of auditory bullae; depth of rostrum (DROST), least

dorsoventral distance from dorsal surface of nasals to ventral

surface of premaxillae; width of rostrum (WROST), greatest width

across rostrum usually at level of premaxillary-maxillary suture;

least interorbital constriction (IOC); least distance between

temporal ridges (TEMP), generally at or anterior to the anterior

width of interparietal; anterior width of interparietal (AWINT);

posterior width of interparietal (PWINT); length or interparietal

(LINT), taken along midsagittal axis; least width across nasals

(LWNAS); greatest width across nasals (GWNAS); length of premaxil-

lary extensions (PMEXT), distance that premaxillaries extend

posterior to posterior tip of nasals; length of upper diastema

(LUDIAST); alveolar length of maxillary toothrow (LMXTR); width of

upper incisor (WUINC); length of lower diastema (LLDIAST); alveolar

length of mandibular toothrow (LMNDTR); and width of lower incisor

(WLINC). These abbreviations are used in the remainder of this

paper. Illustrations of Geomys pinetis crania may be found in

Pembleton and Williams (1978), Hall (1981), and under the names

Geomys tuza and G. mobilensis in Merriam (1895; plate 7).

Natural Region and Area Definitions

Geographic variation was examined using a scheme of combining

specimens from adjacent localities into larger, more zoogeographi-

cally meaningful samples. The state was divided into "natural

regions" delimited by physiographic features (e.g., rivers) sus-

pected to define a number of discrete non-interbreeding populations

that might be morphometrically distinguishable (Fig. 4). Evidence

of the effectiveness of rivers in limiting ranges and dispersal of

pocket gophers is readily available. Lowery (1974:47) noted that

G. bursarius does not range east of a series of Louisiana rivers,

including the Mississippi, Atchafalaya, Ouachita, and Little rivers.

Similarly, G. pinetis occurs in a region bounded to the northeast

by the Savannah River and to the west by the Mobile and Tombigbee

rivers (Lowery, 1974:207). The natural region scheme used here

differs slightly, but importantly, from that used by others studying

Fig. 4.--Areas and component natural regions of Florida
(A-M) divided by rivers (heavy black lines). The unshaded
portion of the southern peninsula lies outside the current
range of Geomys pinetis.

Apolachicolo River -.



AREA III 01025 50 1oo

Vi / Oi

this problem. Williams and Genoways (1980) grouped adjacent

samples with the primary purpose of enhancing sample size. In at

least one instance, such grouping included in one sample specimens

from opposite sides of the Suwannee River--a feature determined in

this study to be an important zoogeographic barrier for Geomys


The regions in the Florida panhandle and in the far northern

peninsula (regions A through I) are discretely defined by rivers on

two or three sides, and most are bounded by the coastline of the

Gulf of Mexico to the south; each of these regions opens northward

into the state of Alabama or Georgia (Fig. 4). Region A is comprised

of the westernmost county in Florida, Escambia Co.; the Perdido

River forms the stateline boundary with Alabama, and the Escambia

River separates regions A and B. The northern portions of Santa

Rosa Co. and Okaloosa Co. comprise region B which is delimited from

region C by the Yellow River. Region C includes parts of four

counties: Santa Rosa, Okaloosa, Walton, and Holmes. Region D,

between Choctawhatchee and Apalachicola Rivers, is composed of

portions of Walton Co. and Holmes Co. plus five other entire counties.

Region E is situated between the Apalachicola and Ocklockonee Rivers.

The Aucilla River separates regions F and G; region G includes

extreme northeastern Jefferson Co. and all of four other counties.

The eastern border of region G is jointly comprised of the Suwannee

River from the Gulf of Mexico to its confluence with the Withlacoochee

River which separates Madison Co. (region G) and Hamilton Co. (region

H). Hamilton Co. is divided between two regions, with regions H and

I consisting entirely of the western and eastern portions of this

county, respectively. The Alapaha River segregates regions H and I.

The continuation of the Suwannee River forms the southern and

eastern edges of region I.

Region J in the north-central peninsula is bounded by the

Suwannee River and the Gulf Coast to the west, by the St. Johns

River and Atlantic Ocean to the east, and the St. Marys River and

Okefenokee Swamp region to the north. The Withlacoochee River

forms part of the southern boundary of region J. Farther to the

south in the central highlands of Lake, Orange, Osceola, and Polk

Counties, region J merges with regions K and L. Region K extends

from Citrus Co. bounded by the Withlacoochee River southward to

include Sarasota Co. and western parts of Charlotte, DeSoto, and

Hardee Counties. The boundary between regions K and L in DeSoto

Co. and Hardee Co. is the Peace River. The southern extent of

region L is denoted by Prairie Creek. The southernmost reaches of

region J (Osceola Co.) are isolated from region L by the Kissimmee

River. Region M is well-defined by the St. Johns River, the Atlantic

coastline, and marshy habitat unsuitable to pocket gophers to the

south in Brevard Co. The names of counties comrpising each natural

region (as well as those from which Geomys pinetis is known) may be

obtained from the list of Specimens Examined (Appendix B).

Polk Co. in the central peninsula is potentially split between

three natural regions (J, K, and L). No Polk Co. localities were

included with region J to the north and northeast because of their

isolation from region J (Lake Co. and Osceola Co.) by (1) the Green

Swamp in northern Polk Co. and southern Lake Co., and by (2)

Snell and Lake Marion creeks, Dead River and other creeks, swamps

and areas impassable to pocket gophers in northeastern Polk Co.

Because no good physiographic feature was found to partition

central from southern Polk Co., and thereby to allow assignment

of localities to either region K or L, all Polk Co. localities

were arbitrarily assigned to region K. Specimens that might be

argued as equally likely to belong to region L are from the Frost-

proof and Lake Wales vicinities. This difficulty of discretely

defining regions K and L (and even J) suggests that study of

zoogeographic relationships of Florida Geomys pinetis might be

better accomplished at a grosser level.

Therefore, each region is considered to belong to one of three

multi-region "areas." All regions west of the Apalachicola River

(regions A, B, C, and D) comprise area I. Those between the

Apalachicola and Suwannee Rivers (regions E, F, G, H, and I) fall

into area II, whereas the four peninsular regions (J, K, L, and M)

constitute area III. These area designations emanated from results

of initial exploratory analyses (e.g., clustering and principal

components analyses); hence, assignment of regions to areas was not

an arbitrary, a priori exercise.

Analyses Conducted

Variation observed in a sample from a particular locality may

result from differences between individuals of different sex or

age. Before undertaking an analysis of inter-population variation,

intra-population variation was examined in the pocket gophers from

Gainesville (Alachua Co.), the largest available local sample (n =

209). Several univariate statistics were computed for each character

for each of the four sex/age combinations using the MEANS and TTEST

procedures of the Statistical Analysis System (SAS, Helwig and

Council, 1979).

In studies of geographic variation, trends between samples are

usually evaluated for each character examined. Clinal variation

among samples may correspond to a variety of environmental para-

meters (e.g., temperature, precipitation, elevation, habitats, etc.),

especially if the samples are taken from a large geographic area.

In the present study of Geomys pinetis in Florida, clinal variation

was examined in 16 cranial characters separately for each of the

four sex/age categories. Maximally non-significant subsets of

regional samples were determined using the Duncan's multiple range

test option of PROC GLM of SAS.

Multivariate statistical techniques are of great value in

systematic studies because, unlike the univariate approach of

examining one character at a time, they permit simultaneous evalua-

tion of large numbers of characters, perhaps reflecting the way that

nature interacts simultaneously with an organism's entire phenotype.

Two general categories of multivariate analyses were conducted.

Clustering and the associated principal components analyses explore

phenetic relationships between samples. Multivariate analyses of

variance and the associated canonical variates analyses are included

in a category of techniques that test the significance of differences

between samples.

The first series of clustering and principal components

analyses (PCA) assessed the similarities among the 13 regional

populations (separately for each sex/age category; see section on

non-geographic variation below). In accordance with conventional

morphometrics methodology (e.g., Baumgardner and Schmidly, 1981),

input data used were the mean values for each of the 16 characters

for each regional sample. Regions represented by only one specimen

were omitted from clustering and PCA on the assumption that individ-

ual specimens might not accurately characterize the population.

Data sets standardized to zero mean and unit standard deviation were

submitted to the Ward's method algorithm (Ward, 1963) of the CLUSTAN

multivariate statistical package (Wishart, 1975). Phenograms and

principal components projections overlaid by computer-generated

minimum spanning trees illustrate clustering and PCA results.

A problem inherent with this approach to cluster analyses is

that mean values provide no measure of the variation within a sample.

Use of individual specimens, therefore, is preferable. A cluster

analysis using standardized values for measure values for each of

the 187 age 1 male specimens as input was conducted to check the

feasibility of a "single specimen" approach over the "mean value"

approach. The inter-specimen relationships resulting from this

experimental run did not follow any comprehensible geographic

pattern. The probable causes for this patternless arrangement are

that (1) the amount of ontogenetic size variation occurring within

this age class is greater than the variation that may be attributed

to geography, and (2) overall variation throughout Florida is slight.

Hence, the "mean value" method is used throughout the study.

Several series of multivariate analyses of variance (MANOVA)

were conducted for each sex/age category to test the null hypotheses

of no significant differences (1) between all regional samples, (2)

between regional samples within their respective areas, and (3)

between area samples. MANOVA's, using character measurements for

individual specimens as input data, were computed by PROC GLM of

SAS. The test statistics generated were the Hotelling-Lawley and

Pillai's traces, Wilks' lambda, and Roy's maximum root criterion.

Canonical variates analyses (CVA) provide a means of pictori-

ally representing relationships between samples by two-dimensional

plotting of specimens along pairs of canonical axes. For clarity,

only the points representing centroid means (bounded by 1 standard

deviation bars) are plotted, rather than plots of individual speci-

mens. The score of each specimen on each axis is the sum of all

measured characters multiplied by their respective coefficients

which indicate their contribution to the variation reflected in a

particular axis. The relative importance of a character to a

particular variate may be determined by a procedure modified slightly

from that outlined in Baumgardner and Schmidly (1981:13). For a

given vector, the mean value of each character is multiplied by

the absolute value of its corresponding loading (i.e., coefficient,

eigenvalue); the products so obtained are then summed. The propor-

tionate influence of a character, then, is the product of its mean

and the absolute value of its coefficient divided by the sum of all

such character-coefficient products.

Multivariate Statistics Rationale

Correct interpretation of the results of multivariate statis-

tical methods requires a general acquaintance with the techniques.

Presented in the following section is a general discussion of the

purpose, operation, and applications of the techniques used in this


Cluster Analyses

Clustering is a classificatory procedure in which entities

(operational taxonomic units or OTU's) are placed in a hierarchical

arrangement in multidimensional space on the basis of phenetic simi-

larities of the entities. The degree of similarity between each

pair of OTU's is expressed by a coefficient (of either similarity

or of dissimilarity). For example, the method used in this study

(Ward's method) employs a squared Eucludean distance coefficient

that varies such that an indistinguishable sample pair has a

coefficient value of 0.0; this value increases with increasing

dissimilarity of OTU's. Relationships may be depicted in two-

dimensional phenograms.

Principal Components Analyses

Ordination procedures (such a principal components analyses,

PCA) may also be applied in an effort to express the relationships

between entities in fewer dimensions than originally measured.

In ordination (as in clustering procedures) data are made dimen-

sionless via standardization to zero mean and unit standard devia-

tion. Thereby, new axes are formed with the new origin centroidd)

being a point corresponding to the mean value for each point;

each entity now possesses new coordinates defined in terms of

standard deviation units.

Multidimensional rotation of axes follows such that perpendi-

cular distances of the entities from the first new axis (first

principal component, first eigenvector, or first latent root) are

minimized. Such a rotation serves to maximize the spread of enti-

ties along the length of this axis. So situated, this first princi-

pal component (PC) accounts for the maximum amount of total variance

that can be explained by a single axis through a particular

n-dimensional cloud of observations. Additional axes, also passing

through the centroid, are positioned at right angles to the first,

with each successively generated axis accounting for less variance

(eigenvalues) than the proceeding one.

It should be noted that none of the new axes directly reflects

the measured attributes, but instead each vector represents a

different linear combination to which each measured character

contributes. The relative contribution of each character to any

particular vector is indicated by the respective character loadings.

The actual percent contribution of a character to a particular

component equals the square of the corresponding character loading

(Neff and Marcus, 1980:54-55). The magnitudes of these loadings,

which may range from -1 to +1, indicate the relative importance of

the original variables in the newly derived component. The position

of any entity along any vector is a linear combination of the products

of measured attributes and their corresponding character loading


The prime objective underlying principal components analyses

is to reduce the number of dimensions (original attributes) from

many to a few that are easily visualized but still preserve the

inter-entity relationships described by the original data set.

Conventionally, spatial relationships between entities are

depicted in bivariate plots having two of the first three principal

components as axes. Proximity of entities on these plots generally

reflects their similarities. Using principal components I, II,

and III as axes, three-dimensional plots may be constructed,

especially when two dimensions do not adequately segregate entities.

However, as Clifford and Stephenson state (1975:186), "addition of

further dimensions in this fashion runs counter to the entire

objective of ordination."

As a further aid to visualizing relationships between entities

plotted in principal components projections, minimum spanning trees

can be overlaid. By this method, pairs of points are joined by a

set of lines such that no closed loops occur (Calinski and Harabasz,

1970, unpublished manuscript cited in Wishart, 1975:123). The

result is that all n entities are connected by n-l lines (edges)

of minimal total length. The comparative value of this method is

evident where samples widely separated in multi-dimensional space

may lie close together in two-dimensional principal components

plots (Clifford and Stephenson, 1975:123-124).

Multivariate Analyses of Variance

Multivariate analysis of variance (MANOVA) is a procedure

that tests for differences between three or more samples on the

basis of simultaneous consideration of many characters (i.e.,

dependent variables). Unlike clustering and principal component

methods, MANOVA groups observations into samples on an a priori

basis. MANOVA operates by determining for each individual specimen

a single value that is a simple linear combination of the products

of actual character measurements and corresponding character

loadings. The set of loadings so obtained allows maximal discrimin-

ation of the a priori designated samples (Harris, 1975:101).

MANOVA examines the null hypothesis that all samples have

the same joint mean or centroid. Rejection of the null hypothesis

indicates that some statistically significant differences exist

in the joint means of the populations sampled. It is not the intent

of MANOVA to identify which samples differ significantly from which

other samples nor to indicate the number of significantly different

subsets of samples present in the analysis. Several different test

criteria are available for establishing the significance of differ-

ences observed between samples. The MANOVA subroutine of PROC GLM

of SAS computes and presents the test statistics and associated F,

degrees of freedom (d.f.), and probability values for three such

criteria (Hotelling-Lawley trace, Pillai's trace, and Wilks' criter-

ion) plus the F-value and associated degrees of freedom for Roy's

maximum root criterion.

Some controversy exists over which is (are) the best test(s)

to use (Neff and Marcus, 1980:153-154). Wilk's lambda, which is

a function of the product of the eigenvalues, requires large

samples for correct probability levels (i.e., is an asymptotic

test). The Hotelling-Lawley and Pillai's criteria are sums of

eigenvalues (i.e., traces) of the matrices resulting from different

manipulations of the effect and error matrices (Helwig and Council,

1979:249, 262). Harris (1975:109-110) argues strongly and convinc-

ingly of the merits of greatest characteristic root criteria over

Wilks' lambda; because Roy's maximum root criterion is an exact test

(unlike the proceeding three), smaller samples suffice for correct

probability levels. Although the SAS MANOVA output does not present

probability values for greatest characteristic roots, the latter is

indicated to be statistically significant if the null hypothesis

of centroid equality is rejected by the other criteria mentioned

above (Neff and Marcus, 1980:154).

Canonical Variates Analyses

Canonical variates analysis (CVA) is an extension of the

MANOVA technique discussed above. Placement of individuals into

samples is performed a priori as in MANOVA. CVA maximizes differences

between these samples on the basis of measured attributes and deter-

mines a single score for each specimen. As for PCA and MANOVA,

such a score is the summation of the products of character measure-

ments and appropriate weighting coefficients. Because these

coefficients are chosen such that within-sample variances are

minimized while between-sample variances are maximized, the series


of canonical axes projected through the multidimensional scatter

of points, in contrast to PCA, are usually not positioned at

right angles to each other (Clifford and Stephenson, 1975:183;

Blackith and Reyment, 1971:49). Each of the canonical vectors

describes different portions of the between-sample variance, and

because characters generally contribute differently to the various

vectors, a unique set of character loading coefficients accompanies

each variate.


The fossil material examined in this study was restricted to

cranial specimens, although some post-cranial material is available

from several sites. Dentaries were the only type of cranial

material present in relatively large numbers from most sites;

hence, features of only the dentaries and of the mandibular teeth

were measured.


Using both a 10X Gaertner measuring microscope (to 0.01 mm)

and Helios dial calipers (to 0.1 mm), 25 features were measured

wherever possible for each of the 526 fossil dentaries and for a

Recent reference sample of 50 males and 50 females from Gainesville,

Alachua Co., Florida. The following characters are measured

features of the dentary: (1) The dorsoventral distance from the

level of the lingual edge of the alveolar toothrow to the mental

foramen (MFDV); (2) the dorsoventral distance from the level of

the lingual edge of the alveolar toothrow to the ventralmost point

of the masseteric ridge (MRDV); (3) anteroposterior distance from

mental foramen to dorsoventral plane through the anterior edge of

p4 alveolus (MFAP); (4) anteroposterior distance from masseteric

ridge to dorsoventral plane through anterior edge of p4 alveolus

(MRAP); (5) diastema, the alveolar distance from p4 to incisor

(DIAST); (6) mediolateral width of the dentary at highest point of

masseteric ridge (WDENT); (7) height of the condyle from level

of lingual edge of alveolar toothrow (HTCOND); (8) anteroposterior

distance from posterior of m2 to posterior edge of angle (M2ANG);

(9-10) alveolar distances between p4 and m2 (P4M2) and p4 and m3

(P4M3); (11-13) length (LRMF), width (WRMF), and depth (DRMF) of

retromolar (= basitemporal) fossa; (14) distance from m2 alveolus

to deepest point of retromolar fossa (M2RMF); (15) distance from

m2 posterior alveolus to dental (= mandibular) foramen (DFOR). The

remaining 10 characters pertain to individual teeth: (16) width of

lower incisor (WLINC); (17-20) lengths and widths of ml (LM1, WM1),

and m2 (LM2, WM2); (21) length of p4 (LP4); (22) length of anterior

loph of p4 (LANTP4); (23-24) widths of anterior (WANTP4) and poste-

rior (WPOSTP4) lophs of p4; and (25) width of p4 isthmus (WISTHP4).

From the above measurements, values for four other characters

were computed: (1) distance from the mental foramen to the incisor

(MFINC = DIAST MFAP); (2) relative dorsoventral positions of

mental foramen and masseteric ridge (MFMRDV = MFDV MRDV); (3)

distance between the mental foramen and the anteriormost point of

the masseteric ridge (MFMR = MFAP + MRAP); and (4) length of poste-

rior loph of p4 (LPOSTP4 = LP4 LANTP4). Lateral and medial views

of a Geomys dentary with endpoints of various of the measured charac-

ters are illustrated in Fig. 5.

Materials Examined

Table 1 lists 27 Florida sites and their component deposits

that have yielded fossil pocket gopher material; the number of

Fig. 5.--Drawings of lateral and medial views of Geomys
mandible indicating several anatomical features and end-
points of seven measurements: (A) MFAP, (B) MRAP, (C)
Names and descriptions of these measurements may be
found in text. After Akersten (1973).


Mental Foramen-

Masseteric Ridge-


Retromolor Fossa* r

Dental Foramen-

dentaries is indicated for each site. With the exception of two

mandibles from Melbourne borrowed from the United States National

Museum, all fossil material is housed in the Vertebrate Paleontolo-

gy Collection of the Florida State Museum, University of Florida

(UF); Appendix C lists fossil geomyid material examined in this

study. No dentaries were measured from Sabertooth Cave as none are

included in the UF collection, although such material exists else-

where (e.g., 4 lower jaws, 2 partial skulls at the American Museum

of Natural History; listed in Simpson, 1928:2). Dentaries from four

other deposits (Orange Lakes Cave, Haile XIV and XVI, and Williston

III B) were not included in this study because the vertebrates in

these faunas are presently being studied by Mike Frazier (Mississippi

Museum of Natural Sciences) and by Robert A. Martin (Farleigh

Dickinson College). Assignment of samples to land mammal ages and

to glacial/interglacial stages was based on consultation of a

number of references (Auffenburg, 1957, 1958, 1963; Martin, 1969;

Webb, 1974; Gillette, 1976; Kurten and Anderson, 1980) and through

re-examination of associated materials and discussions with Dr. S.

David Webb. Specimens comprising the Recent reference sample are

housed in the Recent Mammal Collection, Florida State Museum.

Procedural Rationale and Analyses Conducted

Non-Geographic Variation

In studies of geographic variation for any taxon, it is

desirable to determine what proportion of the variation between

and within samples is due to factors other than geographic occur-

rence. These factors may include sex, age, individual variation,

and experimental error. Once identified, the effects of these

factors should be eliminated (as in the craniometric segment of this


Ideally, the same procedure would be followed in analyses of

geographic and chronological variation in the fossil pocket gopher

material from Florida. Unfortunately, the nature of this (as with

most) fossil material disallows certain analytical procedures. In

rare instances where a relatively intact cranium may be found,

knowledge of sexual dimorphism in living geomyids permits an

inference of the sex of a fossil specimen; however, the sex of a

fossil pocket gopher specimen can never be known with certainty.

Even in living forms, the only mandibular characters useful in

distinguishing sexes are related to size; no qualitative features

are apparent. The value of relative size or dimensions is limited

because of the great overlap in males and females. Pocket gopher

crania and mandibles generally increase in size (in most characters)

with increasing ontogenetic age, with old males being larger than

old females. Only in the zone of non-overlap at the upper end of

the size range of males can the sex be inferred with greater than

50% confidence.

Because of the marked ontogenetic variation, comparisons

between samples should involve subsamples not only of the same

sex, but also of the same age. Although actual age of fossil speci-

mens can never be determined, width of the lower incisor (WLINC)

can provide a reasonable index of age. However, it is not known

whether the ontogenetic rate of increase of WLINC is the same for

both sexes; to my knowledge, this information is not available for

any living geomyid species. Furthermore, the constancy of these

rates over the phylogenetic history of geomyids in Florida is

unknown. Nevertheless, width of the lower incisor (WLINC) was used

as an index of specimen size. Although such a relationship has not

been empirically demonstrated for any geomyids, it is apparent from

observation of an ontogenetic series of Recent Geomys pinetis.

Fig. 6 illustrates ontogenetic variation in Geomys pinetis of one

of the characters (DRMF, depth of retromolar fossa) examined in

this study relative to WLINC. Incisor width is known to be strongly

correlated with ontogenetic age in other rodents (e.g., in Marmota

monax, Ruckel and Scanlon, 1978), so that incisor width increases

with increasing age of the individual.

One means of dealing with ontogenetic variation is to subdivide

samples by increments (0.1 mm) of WLINC and then to restrict all

statistical comparisons to subsamples of like WLINC. There is a

major problem with this approach. Subdivision of fossil samples

(already of limited size) reduces statistical power and confidence.

It may also reduce the number of comparisons that can be made,

particularly if not all WLINC categories are present in local

fauna samples. These difficulties may be overcome by combined

pairs of subsamples of consecutive WLINC values. For this approach

to be valid, however, the ascending (or descending) arrangement of

character means (for all characters) must correspond to a consecu-

tive arrangement of subsamples according to 0.1 mm increments (or

decrements) of WLINC. Because such a correspondence between means

C -

a cou

o C)
0 0)-4

O4H 0 0

0 0a

CO rd I--
0t C

U) '-4 Q)

S> m
o C (3 4t

00) C0
I um
'D 4. 0 (U

040 0
a)o tc 0
0-. ca m 4W

40 o
a) o O

a) a) a )

0. 0 0) 0-
Cr $ $4

0 4 .1 -

-4 44- 0) 0

I 0 C

t!*'0 $4 0.
*r-l C 0 4
4 *H 14-4 $4



S 0 0 000
6 *
ee me*


0 ds -- e
*e *. * -{r E

4z 4 -

I 1 I I I I I I I I I -
f o

o E e
0 01

0 N
0 0 0

0 E

a 0
S z

and WLINC's was seen for most characters in only a general way,

combination of adjacent WLINC subsamples was deemed unadvisable.

An alternative (and preferable) means of accounting for age

variation without subdividing samples by incisor width is via

analyses of covariance. In such analyses, the total variation in

the dependent variable (i.e., the various measured dentary fea-

tures) is partitioned among three sources: (1) the independent

treatment variable of major interest (i.e., locality, glacial/

interglacial stage, mammal age, or other means of identifying

samples), (2) the covariate (i.e., WLINC which is a measure of

ontogenetic age), and (3) an error term. The importance of the

variation from each of these sources is examined via F-tests.

If the covariate is found to significantly influence overall

variability, then comparisons of individual samples are conducted

via t-tests for differences in corresponding adjusted means.

Steel and Torrie (1960:315-316) outlined formulae and procedures

for adjustment of treatment means.

In an effort to understand variability in fossil pocket

gophers due to sex and age, non-geographic variation was examined

in a Recent reference sample of Geomys pinetis from Gainesville,

Alachua Co., Florida. The sample included 50 males and 50 females

selected from over 200 available specimens such that all widths of

incisor represented in the Florida State Museum mammal collection

were included. All characters measured and computed for the

Pleistocene material were similarly measured in this Recent refer-

ence sample. The occurrence of significant sexual variation was

examined, separately for each of 28 dental and osteological charac-

ters, in the Recent sample using a series of t-tests (PROC TTEST of

SAS). WLINC, the 29th character, was used to group specimens;

comparisons were made only between subsamples of males and females

of like WLINC. Coefficients of variation were computed for each of

these 29 characters by sex for each WLINC in the Recent reference

sample. Evaluation of individual character variation was restricted

to the largest male (WLINC = 2.2 mm; n = 7) and female (WLINC = 1.9

mm; n = 9) subsamples. Variation due to ontogenetic age (as indica-

ted by WLINC) was examined within each sex using Duncan's multiple

range tests (PROC GLM of SAS). The purpose of these analyses of

non-geographic variation was to indicate which characters should be

retained for subsequent analyses.

Geographic and Chronological Variation

Analyses of covariance, using width of lower incisor as the

covariate to indicate ontogenetic age, provided adjusted character

means for each sample and for the represented glacial/interglacial

stages. Multivariate analyses examining chronological and geograph-

ic variation utilized these adjusted values as input. Relationships

between samples were first explored via clustering (Ward's method)

and principal components analyses using the CLUSTAN computing pack-

age (Wishart, 1975). Univariate and multivariate analyses of

variance (SAS) were then used to test the significance of (1)

overall variation among samples, (2) chronological variation

between samples from geographically proximal areas, and (3) geo-

graphic variation between samples of similar geological age.

The significance of variation between samples can be assessed

using several multivariate test statistics (e.g., Hotelling-Lawley

and Pillai's traces, Wilks' criterion, and Roy's maximum root cri-

terion). Calculation of these statistics requires that the error

sum of squares matrix be inverted. Such matrices can be inverted

only if the number of observations or samples included exceeds the

sum of the number of variables examined and the number of groups

recognized. Matrices which for this reason cannot be inverted are

referred to as "singular." In several of the comparisons made in

this study, matrix singularity has precluded calculations of multi-

variate test statistics. For those cases, statistical testing

consists of univariate analyses of variance and Duncan's multiple

range tests. Background information developing the rationale for

use of various multivariate statistical procedures is presented

in an earlier section of this dissertation.


Non-Geographic Variation

Coefficient of variation values (CV) were used to assess

within-sample variation in the 20 cranial characters. Because of

extremely large CV values (Table 2), four characters (TEMP, AWINT,

PWINT, LINT) were omitted from further analyses.

Variation due to sex and age was evaluated using t-tests;

t-values and associated probability levels are presented in Table 3.

For age class 1, males were larger than females in all of the re-

maining 16 cranial characters; LWNAS was the only feature for which

the sexes were not different (P > 0.05). For the older adults (age

class 2), females were slightly, but not significantly, larger than

males in GWNAS. The sexes were similar in size for IOC, LWNAS, and

PMEXT. Males were significantly larger than females in the remaining

measurements (Table 3).

Male Geomys pinetis of both age categories were similar in size

in four characters: IOC, LWNAS, GWNAS, and PMEXT. Old adult males

were significantly larger than younger adult males in all other

features examined (Table 3). For females, mean values for IOC in

both age groups were equal. In the remaining characters, older

adult females were larger than younger adults; these differences

were significant for all features except LWNAS, GWNAS, and PMEXT.

Because the Gainesville sample of Geomys pinetis demonstrated

such marked sex and age variation, subsequent analyses of geographic


a) a cu n 1
4L . 0 ON
"0C m H- i L O 0 -* C0a

40 00 11 LI m z0
So 0 wO Q 0 ON m 0- A
0 4.4 I| r-. C C oN CN O- 0 I
u 4 0 -j4 *l -
C0- 4 C '-' CNL H-4 H(0 OCO 0 O* 'N *00
O *H 01 Q in m C (N H 0* H 0* rH

cUO O O n N m * to N .

a )C Z- c
< 4 4 )n
.'o00 0 n 0 0
H0,-4O 0 H CN N
m u 44 H c-

H Mr w11 H4 'C Cl m

cO4 n ..... .. ..4 L
HU> rO HO H00 C '0 'N
w c (v N 0 00.
Ol H .i -I -I .'j C H N H
O 4 00 ml Cl n r m 44
Sc >iO re o rim o m .

4- m 4 T44 NN NH HH L _00 q i-

40 H
I rl 1 -I -:>I q I -| r-
c co 1m 4 oc
m o t (u . * * \or- I N OQl l CO O

x 0 (mu p Ln m *q oC
M E N r-i Nm .4
U o to . mI
4 4 B o" -4 L t .4

0C'4 .w H V
4 40 0 u n l \o C. H Ln o0 'I C


-H )a Ln r> m 0O
CM1 o 11I j un, cn Cn r n o
HX 4 L Cl u I N H 0 N 0 Hi

H c a L H H NN "I 0 I I ) I
C m aI * * * r i o n o
0 04 4 . . o m ** N-O ** *' O* L


O m LI O N-
m Z r o It LI

u4 0a m T 11 . .. .. .. L -m iC iN
H N-4 Hl N00 OC '0 0 Nw
"o 4 H . .' ON .H cN N
00 'C C Ow o0 C *
4-> o 4 D . .. Ni m I Oi O

d) 6 a) MS 9
4 Lw 4. M 00O- ON'C LIw C 0 c.... .a
42144 -4 NN NN cH to HON N'C, -4nH

4u ( HH 0
4- 'L4
(q cc (1 o 1-s u

H p c-i 1- E E O 4C
40 4 0 .u 4 0 ) CH

.c0 o ca ) mz4 1- 4 m c 4 o 0 0 40 w
a 0-qa) 00 c 0 D m c Q o c H fu El
EI H44 r E U U
C l 4 H 44- 4u 14-4 4. o 4H
N4 O4 4t .0 44 ''-- 0.-N 0.. 0'- 0'H 0 4 (J c
'0400 U 4- HE-C 0 H H 00' 0' C -'
H 0 c o 00 0< 140 440 440 44 a 0' Z

HF' ',-4 0 Cl N-' Z 0- 0' 00 -5 HO OH'-

SIN r CT n 1

0 11 N *N *D *lD l 0
'i m i.- i-N c Io

0 N14 00
Il -I -I -I -I -I ?mI 0 1
C 4 r InC Nr CN cN- C-I ** *m

O \0 N 0 4 C

ct .4 C) C

rl* .N .> m m .N .
M ,r) mm -4 -4 IN r-1 C-4 r

o O * C C N C O
IN 00 1 II N 1 -I

l C ) U I

11 ci cI 'CU a, IN-? crj a-- 0 o ,
cccn r. ; lo ^ -icc c I
mo m N NN N I

11 r- 0 l' 4 O C Or -

4 N rll I
U, 0 U0 m0 *I * *
*N *IN *C *' *-t -c C o

-4 * * c c- c

-I I I 1 -I l-i iN Ci
cc .o .m m c 0C c c 0C 0 o4

0* 7* *

Si 0 -0 0 0 0 0 En -
ClM m r o m ^r F-'c i

Ir oI I c 4 U)

o o01 "KO cc -H -a z- m- 4 C

ii t- F H '4 a 0H N 44 ZH c
i I c i i- l. m i
OO m O\D r- m N *

co 0xr 0 i04 Q r Q)c cc 0 Ct c a w c i a. o
H ui oo rH c c i lH b4 F-4 -I 1.
I S U < U JJ; 3 s 0 ^' X'-

*Q Hv 4J V 4-i :3 t04 T Uu t 0

N ID a


0* 0\ *
o -4 O O *
a) CM 00 -A 0 CM


rz 0) 10 -4 *C C 0

Nr. .N -4

o 0 *
a) 1 zM* O r-* c

0 C 0.

to -. -4

So 0 0.

co Ni a, a-r- NN
o o o

Z I M 1 1 N-

a) 10 00 r-i *o'
-I CN l

wz ~o

o O. O* N

'n m J

N M O C Ol
I1 0 m M o r I. N
S ) *U 0 *!CU

0 4 0)0 0 *

0I D H I r4
u 0

a) 144 0 co

Iq 0 p E CM
0 1 a c o

$ 4 n W HC
a C o u

o 1 o o o c
E g E m E- 3 H
CM -l- 0 l -l E O O

)u >2


c u

ql 0C01

D H,


,-1 0

0 w ON

u cl
aH -

rl 0

z1 .2 c
,-l 0

H 0

S> W
'3 CO 0-4
H1 Hd 02

u0 0


> 00c

E-ia p [

( 2 'c n
i Q g il s
O~ N ll il C

u~\ ON 0"' -s
~ln m' oN 'T
m On Cm 0o
Inf '0o Cl) c

4 I





I0I 0 M < 0 l m
C C 0 0 0 0
A 0 0 0 0 0 0
P41 0 0 0 0 0 0


-1o ra. 0 .-a -4 (M
SI 0 0 C 4 a0
4 C1 0 o CM 0c' 0

cz -A o c o o
> .4 . 0 .C
o 0 0 0 0 c
00 0 0 0O 0 0


S 0 'a 'a0 0 CM

-4-I 0 0 0 0 0 0
o o 0 0 0 0
Cq A 0 0 O 0 0 0


CLr r, 0 0 a co
r- 't en 10 It

I0 m 10 e0 'a C'



w ci A 0 0 0 0 O 0
( 0 41 C 0 0 0 0


C e l 0 C n 0

0 0

U En Cl n

E- U

variation entailed comparisons of subsamples of like sex and age.

The four subsamples recognized are (1) males of age class 1, (2)

males, age 2, (3) females, age 1, and (4) females, age 2.

Geographic Variation

Univariate Clinal Variation

Geographic variation between samples was examined to assess

the importance of rivers comprising region boundaries, particularly

the Apalachicola and Suwannee Rivers. For a character to warrant

further examination, significant differences between regional

samples for that character were required; otherwise, the character

was disregarded. Trends in characters with mean regional values

falling into two or more significant subsets were studied on the

basis of regional "area" membership (i.e., areas I, II, or III;

refer to cluster section). In a descending sequence of regional

means (Table 4), regional samples from the same area (i.e., geo-

graphically contiguous) would be expected to be adjacent to each

other and to be separate from samples comprising other areas.

Subsets for which included regions represented all three areas

or non-adjacent areas (I and III) were considered to indicate no

trend. Furthermore, maximally non-significant subsets whose

memberships corresponded to regional composition of areas such

that any subset contains regions from only one area, were judged

to support the contention that the Apalachicola and Suwannee

Rivers have served as barriers to pocket gopher dispersal.

D a pa a -C -i
M M M MM M U l
00 0 M

(i 3D O ii U
3D 3D3 CD uH 3D3D
0 0! >, ^3> 3D 3D 0
CD CD C3 3D '- 3D

c d
O4 C
a) 32 C)l

4- H a


m) 0
Pi 0-
c Eu

a4 03
n0 co Q)

0 $

a C 01


0 0
0- 0 u
321 = 0w

' 0 '
Frua a
>, 033


*" 00 m0

a .I 0 z

I m a

- 4 0 4

Sa 3 4-1
0a) 0 0

r-I 0

.0I U
32 32 0] '

















g" o

w M 4
00 00 AA
-4 MCC A

0 0 0
0 3 M

00 003 m

'-2'-) U (l

M p zB U U

S 44 00 ) I

00i '- )0 )fl 3D M

a 'a 33
D $Q C D

H H 3D
CDC 3D3 0
00 00 U

0 A 0A
U ( U ()
M ti ti
c l ^

PT. pa (Q

i-4 t-i <^ 4
" w
0 0 bet

.40 -4
r. L) SS
N ub
a pa tow ua

00 be1

be M i 0
M ;

0p u
0 0m

P0 0

N 0

u Mb

o7 U
MM i00

0 0

w U
o o be 1 00
S0 0 U
^! Q- 00

C.) U U b
4M4 ^ Q MM ^

u0 0

m0 0 M

O4 U

0 0u
M u U
W p

b e


S 0
o M
00 be
r-4 r.4 i
.- iI T4
PO pq

0 N H b b
0 bee c H b b H b
S HS 0 be 00 t U 0
0 0QQ 1 <

< < 0 Q H

E- <
M Q-

6 -4

be 0
ia 5H

- 4

Appendix A consists of four tables (one for each sex/age

category) enumerating the results of univariate statistical analyses

of 16 cranial characters on a regional basis. Included for each

regional sample are sample size, mean, standard deviation, extreme

values, and coefficient of variation. Samples for any particular

character are sequenced in order of decreasing mean values. The

information contained in Appendix A has been reduced to a single

table showing regional sample sequencing and significant subset

membership for each character, separately for each sex/age category

(Table 4). The ensuing results emanate from consideration of this


Males, age class 1. Only five characters failed to show sig-

nificant variation between regional samples: WMAST, DCRAN, LWNAS,

PMEXT, and LMNDTR. Two significantly different subsets of regional

samples describe variation in seven additional characters. In six

of those (GLS, ZYGO, IOC, LMXTR, WUINC, WLINC), the larger subset

contains regional samples from all three areas; the larger subset

of LLDIAST embraces samples from only areas I and II. The smaller

subsets of IOC and WLINC are comprised of regions of areas II and

III and areas I and II, respectively; regions from all areas com-

pose the smaller subset of the remaining five two-subset characters.

DROST, WROST, GWNAS, and LUDIAST are characters with three

significantly different sample subsets. Samples from all three

areas fall into the intermediate and smaller subsets of all four

characters, as well as into the larger subset of GWNAS. The

larger subset of WROST and of LUDIAST embodies samples from areas

I and III; region J (area III) alone comprises the larger subset

for DROST.

Only three of these 16 characters appear to be of value in

evaluating clinal trends for age class 1 males. The pattern of

IOC variation suggests the importance of the Apalachicola River as

a discontinuity between area I and areas II and III. Here regions

A, C, and D (area I) are adjacent to each other and have means of

greatest value whereas the rest of the sequence is an irregular

assortment of samples from areas II and III. In an opposite

direction, but in a similar manner, WLINC and WUINC regional means

vary such that regions J, K, L, and M have the four greatest mean

values and the balance of the sequence includes mixed area I and II

samples. The pattern of variation exhibited by WLINC and WUINC

points to the possible influence of the Suwannee River in dissecting

the range of Geomys pinetis in Florida. It should be noted, however,

that regional sample alignments in IOC, WUINC, and WLINC correspond

only to sequences of means, not to breaks between maximally signifi-

cant subsets.

Males, age class 2. GLS, ZYGO, LWNAS, LMXTR, LMNDTR, and

WLINC showed no significant geographic variation in old adult males.

Two significantly different subsets of regional mean values were

discerned in each of eight additional characters: WMAST, DROST,


tion of the smaller subsets of all eight characters and of the

larger subsets of all characters except LUDIAST is such that all

three areas are represented in each. Only areas I and III are

present in the larger subset of LUDIAST. Only DCRAN and WROST

are characterized by three subsets. All three subsets of DCRAN

and the two larger subsets of WROST include samples representing

all three areas; regions of areas II and III comprise the smallest

subset of WROST.

Unlike the other three sex/age categories, none of the charac-

ter sample sequences for age 2 males are ordered with the means of

the four peninsular regions being consecutive and of greatest

value. Samples of two characters, IOC and LLDIAST, are arranged

with area I regional samples (B, C, and D) having the three largest

means in the sequences; an irregular ordering of area II and III

regions completes these sequences. However, significant subset

breaks for IOC and LLDIAST do not coincide with area membership of


Females, age class 1. No significant differences were seen

between samples of young adult females in seven characters: GLS,

DROST, LWNAS, GWNAS, LUDIAST, and LMNDTR. Seven characters were

represented by samples falling into two significant subsets, thereby

indicating no geographic trend. For five of these (ZYGO, WMAST,

WROST, IOC, LMXTR), both the large and small subsets contained

regional samples representing all three areas. No geographic trend

was evident in the other two-subset characters. The larger subset

in both LLDIAST and WLINC included samples from areas I and III;

the smaller subset of WLINC included samples from areas I and II

whereas areas I, II, and III were represented in the smaller subset

of LLDIAST. Three significant regional subsets describe variation

in WUINC; the intermediate subset contains regions from all three

areas, whereas areas I and III and areas I and II are represented

in the largest and smallest subsets, respectively.

Variation in DCRAN is described by four significant subsets

(Table 4) with a general trend of increasing size from area I to

II to III. That the mean cranial depth of the region M sample is

greatest and that M alone comprises this subset may suggest a

significant role of the St. Johns River in isolating populations

east of this river from other Florida populations. This same

effect is shown by variation in ZYGO and WROST where samples with

the largest mean character values represent populations located

east of the St. Johns River. Large character means for region M

may be a part of a more general pattern for which samples from all

four peninsular regions tend to have greater character means than

those from areas I and II. DCRAN, WUINC, and WLINC are characters

for which peninsular samples J, K, L, and M comprise the four

largest mean values in the sample sequence.

Females, age class 2. Significant variation was lacking in

the following eight cranial features in old adult females: GLS,


six additional characters grouped into two significantly different

subsets; the smaller subset for all six contained samples from all

three areas. IOC, PMEXT, and LMNDTR are characters in which all

three areas are represented in the larger subset as well. Samples

from areas I and II occur in the larger subset of LLDIAST, whereas

the larger subset of LMXTR and WUINC include regional samples from

areas II and II and area III, respectively.

Characters with samples falling into three significant subsets

are DCRAN and WLINC. The intermediate and smaller subsets for

both characters include regions from all three areas. Areas II and

III are represented in the largest subset of DCRAN, as is area III

for WLINC.

Region M has the greatest mean value for five characters

exhibiting significant differences: DCRAN, LMXTR, WUINC, LMNDTR,

and WLINC. WLINC is the only character for which the four penin-

sular regions are the samples with the four largest means. The

arrangement of WLINC samples could support the importance of the

Suwannee River as a dispersal barrier, yet area III regions occur

in all three subsets, and therefore are not significantly different

from certain regions of other areas. The remainder of the WLINC

sequence comprises a seemingly random jumbling of area I and II


Summary of univariate clinal variation. The preceding

univariate examination of characters on a regional and area basis

failed to demonstrate any convincing overall clinal trends that

(1) occur in all four sex/age categories or (2) consistently include

a particular subset of characters. The variation in nine characters,

however, hints at the possible importance of three rivers in parti-

tioning Florida pocket gopher populations.

In at least one character in each sex/age category, region M

possesses the greatest mean value. However, only for DCRAN (females,

age 1) is this value for M significantly different from all other

regions. Sequencing of means with the value for M being greatest

suggests that populations east of the St. Johns River in peninsular

Florida may be isolated from other Florida populations. Characters

for which the four greatest mean values include peninsular regions

J, K, L, and M indicate that peninsular forms may be separated from

panhandle Geomys by the Suwannee River. No characters showed this

pattern in age 2 males, although this trend is evident in as many

as three characters (DCRAN, WUINC, WLINC) in the other sex/age

groups. These area III region means were not significantly different

from means of regions of areas I and/or II in any of these compari-


In only three of the possible 64 cases (from 16 characters for

each of the four sex/age classes) do all regional samples from area

I group together at the end of a character sequence. In the LLDIAST

and IOC sequences for old adult males, area I regions have the

three greatest sample means. The area I regions represented in the

young adult male analysis also are the samples with the three greatest

means. Such contiguous alignment of area I regions (concomitant

with patternless mixing of the area II and III samples) tends to

emphasize the role of the Apalachicola River in separating western

panhandle Geomys from those in the remainder of Florida. Again,

however, breaks between significant subsets do not correspond with

area membership of regional samples.

Cluster Analyses

In this study it was expected that phenetic similarities

between samples of like sex and age would provide insight into the

degree of past or present gene flow between populations in adjacent

regions. The manner of cluster formation in the following analyses

addresses the relative importance of the various rivers (inter-

region boundaries) in inhibiting pocket gopher dispersal.

Males, age class 1. The dendrogram depicting similarities

between samples of males of age class 1 bears a strong resemblance

to the arrangement above. The four peninsular regions cluster

tightly with each other, yet distantly from all area I and II

regions (Fig. 7B). The only two area I regions (C and D) represented

in this comparison comprise a discrete cluster. Regions G and F

(area II) form two single-member branches each more closely con-

nected to area I samples than to area III samples.

Males, age class 2. Fig. 7C presents the phenetic relation-

ships of samples of old adult males. The clarity of separation of

samples with respect to the Suwannee and Apalachicola Rivers seen

previously is not as evident in this comparison due to the position-

ing of the area II regions F and G. G bears little resemblance to

any regions, whereas F is included with area III. Still, the two

regions from area I (C and D) form a distinct cluster emphasizing

the importance of the Apalachicola River as a barrier. Addition-

ally, three of the four peninsular regions comprise another group;

the fourth peninsular region (L) occurs in the same major cluster

as regions J, K, and M.

Females, age class 1. In the dendrogram generated for females

of age class 1, two primary clusters are evident (Fig. 7A). One of

these clusters contains all four peninsular regions (regions J, K,

L, and M) that collectively are designated area III. The remaining

Fig. 7.--Results of cluster analyses of the four sex/age
categories of Recent Geoys pinetis in Florida. Dendrograms
computed from distance matrices using Ward's method. Letters
indicate natural regions; associated numbers represent sample

0 1 2 3 4 5 6
B 2
C 15
D 24
F 12
G 5
H 3
J 162
K 40
M 28
L 11

A 3
C s
D 3
F 7
J 94
K 35
M 27
E s
L 3

1 2 3 4

0 1 2 3
C 4
D 7
G 7
F 8
J 106
K 34
L 6
M 18

0 1 2 3
C 6
D 5
F 8
J 95
K 27
M 23
L 6
G 2


4 5

4 5


regions, all occurring west of the Suwannee River, are split into

three groups. One group contains only panhandle forms (B, C, and

D) from west of the Apalachicola River (area I). Area II includes

those regions situated between the Suwannee and Apalachicola rivers.

Of the three area II regions included in this analysis, two (F and

G) are closely united in one cluster. Region H, however, is closely

joined to neither the area I or II clusters, although its affinity

is clearly with the samples west of the Suwannee River.

Females, age class 2. The age class 2 females dendrogram (Fig.

7D) most closely resembles that for old age males. Region A, C,

and D, the only area I samples included in this analysis, form a

discrete cluster containing regions from no other areas. A second

major cluster includes regions J, K, and M (all peninsular) plus

region F (area II); these same four regions formed a group in the

age 2 male comparison. The remaining two regions (E and L) in this

analysis are similar to each other but very different from all other


In each of the four analyses, all of the included area I

regions were consistently and exclusively grouped together. In

two of the four dendrograms (Figs. 7A and 7B) the four peninsular

regions formed exclusive, highly similar clusters, while in Figs.

7C and 7D three peninsular regions grouped with an area II region.

In all four comparisons, area II regions were placed either alone,

or with another area II region, or with peninsular regions. These

data indicate that both the Apalachicola and Suwannee Rivers are

barriers to gene flow in pocket gophers, with the Apalachicola

being the more effective of the two.

Principal Components Analyses

Males, age class 1. PC I and PC II for age 1 males account for

72.39% of the observed variance; PC III represents an additional

15.24%. Character loadings for PC I are positive for all 16 charac-

ters except PMEXT (Table 5), suggesting that differentiation between

samples along this component reflects general cranial size. The

highest coefficient for PC I is 0.340 for length of maxillary

toothrow. Positive correlations for PC II are with various dental

characters maxillaryy and mandibular toothrows, lower incisor width),

certain rostral features (least width of nasals, extension of pre-

maxilla past nasals), cranial breadth and depth, and others. Least

and greatest widths of the nasals were most important characters

for PC III; various other positively correlated characters reflect

rostral and posterior cranial dimensions.

Table 6 enumerates coordinates of regional samples along the

first three PC's for each of the four sex/age analyses. Computer-

generated minimum spanning trees superimposed on these PC plots

connect the most closely related regional samples and indicate

relationships that might not be evident otherwise; Table 7 presents

minimum spanning tree edge lengths. For age class 1 males, inter-

sample relationships are similar in plots of PC I vs. PC II and

PC I vs. PC III, which explain 72.39% and 64.14% of the variance,

respectively (Fig. 8). The peninsular regions (J, K, L, and M of

area III) comprise a single cluster of close-lying points in Fig.

8A and a more scattered group in Fig. 8B. Samples from regions C

and D (area I) are more or less closely associated in Figs. 8A and


r0 a

-H zH O Ot -T r o C r COO 4 r o a
6D 0 w -4 -4 W C 4 r-I m G CN WO r-I C 00 r, -4 t 0
0) w en m 0 m 0 0 en O 0 0 -T o0 r -l:T m 0 r-

1 o oooooooooooo ra
rl O U I I I I I 1 1 1 1 1 1 l

m H
o 0

0 c cN C m cri rl Ti i i Q cu in IN &
E: iO M o Hc N cN O o rl rn c N c c^ i- i m or f
SP ) -H 1C 1C H 1C 1C H H 11

a) 0 $

*-' 1I I I I Im
a P.
o 0)

0 0

o C 4 O 0T \O Tl m 0 C1 c1 rm l m m 14 C

I Il I I I I
00o H oDT0 o rr 4ct( I

03 ;O c

0 0
0)0 J

0 OO
a- u 9) I l I I l C F

*T-I r0 x a

00 m m
0 0

IC. .
10 "

04 0 U 0o
paW u n o

cou C U (t CC 4 0


c4 10 'O m (N r na c co 0 0 N c m
S Ill I hI I



H HOm CC'00004,M@C' H

-') H-oomooz a o I,

o c D o o c 5


m cc r-i ri rui i c 5 o H
El Mc u N Cl a j u^l a 14 : E -

0 u


H-i U U N Sl Q a J(] -. H3

'.00 -T '.0 '.D C ,-i j

1,I0C 00h N -
lnH O O OO I N

S00o '. N 'O 'o i

Crl CrOnOlrO

r- a- if 00ir on N C

ooo0 iNoN -4

d- 0
M 4

2 0)
n o

o o

.a r
0 0.


-3 00

arl Q0

PM o

02 0

0 20

E 4 H
cO H
44 r'
o n


(0 o
a) a

r Eo

0 HI

$ 4 0
C w


*u 0m

Tl E 0-

3-U 0-
n) u

I ^4-

N 0 -0 '.M0 N 0
in -ON- CN 0 C -m
0oo 'Cr in Nr- rN O

. O. .No
Cr -02.1i n rN '.0 0

o -4 r0o o in 0)

N- It m 00 in ri


0 in '0- 0 0 N- '0

-0 in C0 -4 Cr

m C


N- -02 '. N1 -2 'T '0 -IT rN-
'0 i-40-C 0 -1 0-0

ONC00N4 C4 4

CO n omN -m

in cN NH N- 00 0 'H

c.. . .-i -7 C m x Ln
IN r4- Cr i Cr

00 r-4 N- 0 Cr -02 CN C-

Somm-o m lm0r- 0
'.0 in m0r N-N-0 in r i

0n in- 0 N- Crr n cr0

in M0o0 r-'- r rn- M
- r oQ1 ro r- o "
---- 0 -oo4m -4 0


$4 0

M (U


Table 7.--Results of principal components analyses between natural
region samples of Recent Geomys pinetis from Florida: minimum
spanning tree parameters including edge lengths.

Males, Age 1

Edge No. Vertices Edge Length

1 C to G 1.139
2 D to C 0.820
3 F to G 2.131
4 G to K 0.765
5 J to M 0.690
6 K to J 0.456
7 L to K 0.848

Females, Age 1

B to
C to
D to
F to
G to
H to
J to
K to
L to

Males, Age 2

Edge No. Vertices Edge Length

1 C to F 1.659
2 D to C 1.718
3 F to J 0.500
4 G to F 2.397
5 J to M 0.349
6 K to J 0.229
7 L to F 0.837

Females, Age 2



Fig. 8.--Relationships between regional samples of age 1
male Geomys pinetis as shown on plots of principal compo-
nents I vs. II and I vs. III. Percent variance contained
within each component indicated along axes. Most similar
samples connected by a minimum spanning tree (dashed lines).


Pk M


PC I 48.90 %
___________>-,'^^ ^

c -D

PC I 48.90%


8B, respectively. Region F, a distant outlier in both plots, is

most similar to region G, the only other area II sample in this


Males, age class 2. Over 83% of the total variance is

explained by the first three PC's (Table 5). The pattern of

character loadings for PC I here is very similar to that for PC I

in the younger males; the only negatively correlated character is

PMEXT. The greatest coefficients in PC I for old adult males are

for rostral depth, length of mandibular and maxillary toothrows,

and skull length. The loadings for PC II in both old and younger

adult males are similar; the primary differences are sign reversals

for skull length, zygomatic breadth, and rostral depth. Width of

the upper incisor was the greatest contributing factor to PC III;

other important characters were width of the lower incisor and

rostral width.

The plot of PC I vs. PC II depicts a central core of samples

from which two arms radiate (Fig. 9A). One arm is composed of

regions C and D, the only area I samples represented in this

analysis. Region G, also distant from the major cluster, is most

similar to region F, the only other area II sample in the analysis.

The five regions comprising the central cluster are divisible into

two groups. Region L (area III) lies separate from the other

peninsular samples (J, K, and M of area III) that are most similar

to region F. This plot explains 68.76% of the variance in the

data set.

Fig. 9.--Relationships between regional samples of age 2
male (A) and age 2 female (B) Geomys pinetis as shown on
plots of principal components I vs. II. Percent variance
contained within each component indicated along axes.
Most similar samples connected by a minimum spanning tree
(dashed lines).










PC I 37.01 %



J i

PC I 35.99

PC I 35.99 %

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs