Journal of Mammalogy, 90(6):1449–1460, 2009
QUANTITATIVE MORPHOLOGICAL PROXIES FOR
FOSSORIALITY IN SMALL MAMMALS
SAMANTHA S. B. HOPKINS*
AND
EDWARD BYRD DAVIS
Clark Honors College and Department of Geological Sciences, 1272 University of Oregon, Eugene, OR 97403,
USA (SSBH)
Museum of Natural and Cultural History and Department of Geological Sciences, 1272 University of Oregon, Eugene,
OR 97403, USA (EBD)
Burrowing behavior is widespread among mammals and has generated a diverse array of adaptive responses to
the physical demands of this lifestyle. While extensive research has been devoted to the morphological,
ecological, and evolutionary implications of burrowing, it remains difficult to compare burrowing adaptations
between mammals of widely divergent ancestry. A reliable quantitative proxy for fossoriality (burrowing) is
necessary for such comparisons as well as for detailed descriptions of ecology from specimens of rare, extinct,
and fossil mammals. This study presents several quantitative indices of the morphology of burrowing mammals
based on 20 measurements of skull and skeletal morphology taken from 123 different mammalian species, both
burrowing and nonburrowing. Discriminant analyses revealed that these quantitative characters successfully
distinguish nonburrowing taxa from those that are adapted to a burrowing lifestyle. Additionally, more subtle
distinctions between subterranean taxa (which rarely emerge above ground) and other burrowers as well as
between mammals using different methods of burrow excavation were identified from these characters. A test
of these indices using 6 extinct species yielded results consistent with more-detailed descriptions of the
functional morphology of these taxa, indicating that our quantitative proxies provide an important basis for
comparisons of fossorial adaptations across divergent mammalian clades.
Key words:
burrowing behavior, discriminant analysis, fossoriality, morphological proxies, skeletal morphology
The ability to reconstruct ecology from morphology has
long been an important tool for evolutionary biologists
studying fossil organisms as well as rare, shy, or recently
extinct animals. Morphological proxies for ecology can
illuminate the biology of such organisms, enabling studies of
otherwise intractable phenomena such as locomotor ecology.
One of the most important ecological attributes of an animal is
how it interacts with its habitat through locomotion and
substrate use. A fundamental ecological distinction among
terrestrial vertebrates is the difference between burrowing and
nonburrowing animals. Burrowing animals have a substantial
influence on the ecosystem and often act as keystone species
(Reichman and Seabloom 2002). Fossorial (burrowing)
animals have evolved numerous times, particularly among
vertebrates, and have a variety of morphological features
related to their energetically demanding mode of locomotion
* Correspondent: [email protected]
E 2009 American Society of Mammalogists
www.mammalogy.org
1449
(Stein 2000). These features make them ideally suited for
drawing ecological inferences from morphology.
Among mammals, fossoriality is widespread and includes
ecologically important burrowers such as pocket gophers
(Geomyidae—Reichman and Seabloom 2002), zokors (Myospalacinae—Zhang 2007), and mole-rats (Bathyergidae—
Reichman and Jarvis 1989). The 4 major digging modes
found in fossorial mammals are scratch digging, chisel-tooth
or incisor digging, humeral-rotation digging, and head-lift
digging (Hildebrand and Goslow 2001; Stein 2000). Scratch
digging is the ancestral mammalian digging pattern, used by a
majority of both fossorial and nonfossorial mammals for
excavation. Scratch diggers use a running motion of the
forelimbs to loosen soil with the claws. Chisel-tooth digging,
used by several different groups of fossorial rodents, involves
gnawing with the incisors to excavate burrows. Humeralrotation digging is found only in moles (and possibly in
modified form in proscalopids—Barnosky 1981, 1982) and
consists of a powerful rotation of the humerus with the
forelimb extended anteriorly to sweep earth behind the digger.
Head-lift digging involves use of the snout and forelimbs to
wedge open tunnels in the substrate. This derived mode of
1450
JOURNAL OF MAMMALOGY
excavation has evolved in widely divergent clades of extant
mammals including golden moles (Chrysochloridae—Gasc et
al. 1986), mole voles (Ellobius, Cricetidae—Stein 2000), and
blind mole-rats (Spalacinae—Nevo 1961), as well as the
extinct clades Mylagaulidae (Hopkins 2005) and Palaeanodonta (Rose and Emry 1983). These digging strategies are not
mutually exclusive; animals frequently use multiple modes of
digging depending on the substrate and the type of burrow
being constructed (Stein 2000; Vassallo 1998).
The variation in body parts used for each mode of digging
makes developing a morphological proxy for digging difficult.
Most prior attempts to describe the morphology of fossorial
mammals (Hildebrand 1985; Shimer 1903; Stein 2000) have
given qualitative, gestalt descriptions of the types of skeletal
modifications observed. Rare attempts to quantify morphological changes in digging mammals have generally used a
single measurement (e.g., Vizcaino and Milne 2002) or dealt
only with very limited taxonomic samples (e.g., Courant et al.
1997; Vassallo 1998). In order to maximize the utility of a
proxy for fossoriality, it must be applicable to a wide variety
of taxa. As part of assessing utility, it is useful to determine
how simple single-measurement and ratio proxies compare
with multimeasurement proxies of varying complexity.
Our study aims to determine the relative performance of
different morphological measures as predictors of fossoriality,
both as individual indices and as part of multivariate
predictive equations. Based on a priori hypotheses regarding
the effects of digging on skeletal structure, we measured a
variety of morphological characteristics; because all 4 modes
of digging influence the morphology of the head and forelimbs
most extensively, we focused our analyses on these regions of
the skeleton. Using these measurements, we generated a series
of size-independent indices that should correlate with
burrowing behavior and compared index values across a wide
range of fossorial and nonfossorial mammals. We expected to
find that these characteristics (or some subset of them) would
enable the diagnosis not only of a fossorial life habit, but also
of the specific digging modes found among fossorial and
subterranean mammals.
MATERIALS AND METHODS
Data collection.—Morphological proxies for fossoriality
were developed based on descriptions of burrowing taxa by
Hildebrand (1985) and Shimer (1903), with information from
Biknevicius (1993) added for analyses of cross-sectional areas
of the humerus and femur. Twenty morphological characters
were measured (Fig. 1); all linear and area measures were
natural-log transformed to account for naturally increasing
variance with increasing values (Zar 1999). Limb characters
examined included humeral length (HL), which described the
shortening of the proximal limb, required to shorten the total
output lever for the digging stroke by posterior rotation of the
shoulder joint. Distal humeral width (DHW) described the
area of attachment and the leverage available for the tendons
that flex the digits of the manus. Length of the deltopectoral
Vol. 90, No. 6
FIG. 1.—Linear skeletal measurements used to describe fossorial
morphology. The specimen shown is University of California
Museum of Paleontology (UCMP) 67110, locality R0 (Vombatus).
Angular measurements (not depicted) are described in the text. Area
measurements were calculated from humerus and femur widths
(HAP35, HTR35, FAP35, and FTR35). Abbreviations are as follows:
humeral length (HL), distal humeral width (DHW), length of the
deltopectoral crest (DPC), ulnar length (UL), length of the olecranon
process (LOP), length of metacarpal III (L3MC), length of distal
phalanx III (L3P), width of distal phalanx III (W3P), depth of distal
phalanx III (D3P), length of metatarsal III (L3MT), anteroposterior
(HAP35) and transverse (HTR35) widths of the humeral shaft
(measured 35% of the length from the distal end), anteroposterior
(FAP35) and transverse (FTR35) widths of the femoral shaft
(measured 35% of the length from the distal end), skull length
(SL), skull width (SW), and occipital height (OH).
crest (DPC) quantified the lengthening of the input lever arm
for the deltoid muscle, a critical stabilizer muscle in the
digging stroke of some mammals, and 1 of the muscles that
powers digging in others. Ulnar length (UL) and length of the
olecranon process (LOP) were included as complementary
measurements, because digging force is greatest when the
olecranon process is long, providing a long input lever arm for
forelimb extension by the triceps, whereas the portion of the
ulna distal to the olecranon fossa is relatively short, providing
a short output lever for triceps extension. Length of metacarpal
III (L3MC) described the shortening of the bones of the hand
to increase the power and stability for digital flexion, whereas
length of distal phalanx III (L3P) described the lengthening of
the digging claw, which affects the amount of soil removed
with each digging stroke of the forelimbs. The 3rd phalanx
was chosen as a representative of all digits because it is
December 2009
HOPKINS AND DAVIS—MORPHOLOGICAL PROXIES FOR FOSSORIALITY
relatively easy to identify as an isolated element. The width of
distal phalanx III (W3P) and depth of distal phalanx III (D3P)
described the differences in shape of the digging claws, which
should be wider (for stabilization of the joint, to avoid
dislocation) and deeper (to strengthen the claw against the
dorsoventral forces exerted as the claw moves through the
soil) in fossorial forms. Length of metatarsal III (L3MT)
offered a comparative measure of the size of the metapodials
of the pes, allowing discrimination of an enlarged manus
relative to the overall size of the hands and feet; this is
meaningful because burrowers tend to enlarge the manus,
whereas some other forms of locomotion (e.g., swimming)
enlarge the manus and pes simultaneously.
Several measures of limb width also were examined. These
included the anteroposterior (HAP35) and transverse (HTR35)
widths of the humeral shaft (both measured 35% of the length
from the distal end of the humerus) as well as the
anteroposterior (FAP35) and transverse (FTR35) widths of
the femoral shaft (measured 35% of the length from the distal
end), all of which describe changes in the robustness of the
proximal limb bones. The great torsional and bending forces
that digging exerts on the limbs must be resisted by these
bones, resulting in increases in bone diameter. These 4
dimensions enabled estimation of cross-sectional area of the
limbs by approximating the cross section of each bone as an
oval with axes equal to the anteroposterior and transverse
measurements, yielding the approximate values of the crosssectional area of the humerus (CSAH) and the cross-sectional
area of the femur (CSAF).
Finally, skull characters examined included upper incisor
angle of procumbency (UIAP) and lower incisor angle of
procumbency (LIAP), which described the ability of the
animal to excavate soil with the incisors. Both angles were
measured as the angle between the cheek toothrow and the
tangent to the incisor halfway between the point of the incisor
and the alveolus. The angle of the occiput from horizontal
(OCA) described the change in the shape of the posterior part
of the skull to accommodate increases in the mass of muscle
used to elevate the skull on the neck, a movement used both to
pack the roof of the burrow, to bulldoze excavated soil from
the burrow, and, in head-lift digging burrowers, to excavate
the burrow with the snout. Skull length (SL) described the
length of the output lever for this elevation of the skull,
whereas skull width (SW) described the amount of area
covered by this stroke; a wider skull enables an animal to
contact a wider area of burrow roof with each stroke. Occipital
height (OH) described the length of the input lever in the
elevation of the skull, with increasing height allowing an
increase in the force exerted with each stroke.
The above morphological characters were measured on 338
specimens from 123 species representing 15 of the 29 orders
of extant mammals (data are available online at the University
of Oregon Scholar’s Bank, http://hdl.handle.net/1794/9604).
Most major clades of fossorial mammals were included, as
were closely related nonfossorial taxa. Mammalian orders that
contain no fossorial taxa and are not closely related to
1451
fossorial clades (e.g., Artiodactyla and Perissodactyla) were
not included in the analysis. Measurements were taken using
Mitutoyo 500-171 15.2-cm (6-inch) digital calipers on
specimens housed in the United States National Museum of
Natural History and the University of California’s Museum of
Vertebrate Zoology. One species, Notoryctes typhlops (the
marsupial mole, family Notoryctidae), was measured from
published images of specimens (Stirling 1891, 1894; Warburton 2003) because no skeletal specimens were available and it
provided important phylogenetic diversity within the burrowing taxa sampled. To maximize the taxonomic coverage of our
sampling, we focused on including as many different species
as possible, most of which were represented by a single
individual. To determine the magnitude and potential
significance of within-species variation, 5–10 individuals per
species were examined for a subsample of 21 species in our
data set. For each measurement, the standard deviation of log
values within these species was compared to the standard
deviation among all species to assess whether our singleindividual samples were representative of the morphologies of
these taxa.
Phylogenetic considerations.—Because this study attempted
to recognize fossoriality based on morphological traits and
without information about phylogenetic relationships among
taxa, no explicit controls for phylogeny were included in our
analyses. Although we attempted to include a phylogenetically
diverse sample of mammalian taxa, our findings may differ
from those of other analyses that explicitly include phylogeny.
Phylogenetically controlled studies of changes in diggingrelated aspects of morphology will be the subject of a future
analysis.
Statistical analyses: simple indices.—As an initial test of the
relationship between burrowing and morphology, we examined several simple potential indices of fossoriality. These
simple indices, most of which consisted of ratios of the
characters described above, were not used in the discriminant
analyses that are the focus of this work but, instead, were used
to determine whether characters previously identified (Biknevicius 1993; Hildebrand 1985; Hopkins 2005; Landry 1957;
Lessa and Thaeler 1989; Shimer 1903; Stein 2000; Vizcaino
and Milne 2002) as diagnostic of fossoriality could be
quantified to provide a simple measure of the degree of
fossorial adaptation. DHW/HL quantified the broadening of
the distal end of the humerus described by Shimer (1903) and
Hildebrand (1985) as diagnostic of burrowers. This expansion
of the distal tuberosities of the humerus provides greater area
of attachment for the digital flexor tendons in burrowing
mammals. DPC/HL described the increase in leverage of the
deltoid muscle, an important stabilizer to digging in some
burrowers and a source of digging force in others. LOP/UL
described the index of fossorial ability of Vizcaino and Milne
(2002), a measure of the leverage for forelimb extension
provided by the triceps muscle. CSAH/CSAF was an index
intended to reflect the burrowing adaptation suggested by
Biknevicius (1993) in which the cross-sectional area of the
humerus is increased to resist bending strains encountered
1452
JOURNAL OF MAMMALOGY
during excavation. For these indices, we compared mean
values for burrowing and nonburrowing mammals using
Student’s t-tests. To determine if these indices were sensitive
to the degree of fossoriality displayed by a species, we also
examined differences in the mean values for these indices
among subterranean, fossorial, and nonburrowing species
using Tukey’s honestly significant difference test.
Three additional simple indices of fossoriality were
considered. UIAP and LIAP were measures of the degree to
which the incisors are modified to allow their use in breaking
up soil. Forward-projecting incisors are required for mammals
to be able to use the anterior dentition in this way, and
equivalent measures have been used in the past (i.e., Lessa and
Thaeler 1989) to describe adaptation to incisor digging in
rodents. OCA was intended as a quantification of the
adaptation for head-lift digging. The anterior-leaning occipital
plate is one of the most recognizable and consistent features of
head-lift diggers, and was illustrated but not directly discussed
by Stein (2000). Hopkins (2005) explained this feature as an
adaptation for accommodating enlarged dorsal neck musculature for the powerful stroke elevating the tip of the snout as
soil is swept upward on the anterior wall of the burrow. We
used t-tests to compare the values of these indices for 2
subgroups of burrowing rodents (incisor digging and head-lift
digging) and all other burrowers.
Statistical analyses: multivariate indices.—To identify suites
of characters that, together, indicated a burrowing lifestyle,
morphological data were analyzed using discriminant function
analyses as implemented in the software package JMP 7.0.1
(SAS Institute Inc., Cary, North Carolina). We used these
analyses to determine both how accurately our data could
discriminate burrowers from non-burrowers and how precisely
we could define specializations for burrowing and still
produce reasonably accurate predictions regarding morphology. Our 1st analysis compared burrowing and nonburrowing
mammals by dividing species in our sample into 2 groups
(burrowing or nonburrowing [Y/N]). Burrowing animals
included any species that dig branching burrows or burrows
with chambers; under this definition, subterranean species
were included with fossorial taxa. Our 2nd analysis divided
the species sampled into 3 groups: subterranean, fossorial, or
neither (S/F/N). To examine the relative importance of
different skeletal components, we then reran our discriminant
function analyses (both Y/N and S/F/N partitions) with the
morphological data partitioned in the following 4 functional
units: head, limbs, manus plus pes, and head plus humerus. To
distinguish among digging styles for the fossorial species (the
Ys from the Y/N analyses), we ran discriminant function
analyses for the following specific digging types: scratch,
incisor, humeral-rotation, and head-lift diggers. Finally, we
performed an overall discriminant analysis that included all
species partitioned into 1 of the following life-habit categories: semiaquatic, arboreal, terrestrial, fossorial, or subterranean. For several models in our analyses, we performed
stepwise addition or subtraction of characters; because JMP
does not automatically add in the additional, incompletely
Vol. 90, No. 6
sampled data that become available with simpler models, we
repeated all analyses involving addition or subtraction of
characters outside of the JMP stepwise-addition tool to ensure
that all models were informed by data from as many species as
possible.
Almost all of the above procedures used quadratic
discriminant analysis, which performs best when sample sizes
are larger than the number of predictor variables and
covariances among predictors differ between the groups to
be distinguished (Friedman 1989). Assuming that skeletal
characteristics covary differently in mammals that are under
selection for digging behavior and those that are not, both of
these characteristics were met by our overall analyses. For
analyses that attempted to discriminate among different types
of digging behavior (humeral-rotation, head-lift, incisor, and
scratch), our sample sizes fell below the number of predictor
variables. To avoid the problem of an ill-posed quadratic
discriminant analysis (Friedman 1989), for these analyses we
included n 2 1 predictor variables (where n is the number of
species in the behavioral group in question). For humeralrotation diggers, we posited a priori that characters of the
limbs would be most useful for discrimination; we used
stepwise character removal to eliminate the least-informative
limb character and produce our resulting model. For the other
modes of digging, we could not make an a priori assertion that
any functional group of characters should be better at
discrimination, so we used stepwise character addition to
build up a model of suitable complexity. Although regularized
discriminant analysis (Friedman 1989) would have allowed us
to include all of the predictive variables, our reduced-predictor
quadratic discriminant analyses performed exceptionally well
in all cases, providing no reason to increase the complexity of
these analyses by employing regularized discriminant analysis. Because we had only 3 semiaquatic species in our data set,
however, we used regularized discriminant analysis for the
life-habit analysis. Eigenvalues and eigenvectors for all
discriminant analyses are available online at the University
of Oregon Scholar’s Bank (http://hdl.handle.net/1794/9604).
To determine how well our morphological indices identified
fossoriality among species not included in the model
construction, we performed 2 tests. First, we jackknifed the
data, rerunning the quadratic discriminant analyses for
burrowing (Y/N) and degree of fossoriality (S/F/N) for the
overall data set with 1 species per iteration withheld from the
analyses. This is a typical method of testing the predictivity of
a discriminant function analysis and may provide a better idea
of its accuracy than a simple accounting of misidentifications
of included data (DeGusta and Vrba 2003). Second, we used
the discriminant functions obtained to analyze degree of
fossoriality in a series of relatively well-characterized fossil
mammals. One of these species, Mesoscalops montanensis
(family Proscalopidae), has been the subject of a thorough
analysis of its life habit (Barnosky 1981), which indicated that
it had convergently evolved the humeral-rotation digging
behavior seen today only in true moles. All fossil taxa were
chosen based on the availability of detailed functional
December 2009
HOPKINS AND DAVIS—MORPHOLOGICAL PROXIES FOR FOSSORIALITY
morphological analyses that could be used to assess degree of
fossoriality.
Applications of indices to fossil taxa.—Data on the skeletal
morphologies of 6 fossil taxa (5 members of the suborder
Palaeanodonta: Tubulodon pearcei, Metacheiromys tatusia,
Xenocranium pileorivale, Escavadodon zygus, and Palaeanodon ignavus, and 1 proscalopid, M. montanensis) were
collected from published species descriptions (Barnosky
1981; Colbert 1942; Gazin 1952; Matthew and Granger
1918; Rose 1978; Rose and Emry 1983; Rose et al. 1991; Rose
and Lucas 2000; Simpson 1931). Data for these taxa were not
complete, due to both partial specimens and the absence of
necessary published images and measurements. To accommodate these data limitations, 2 new discriminant analyses were
run with reduced character sets. The 1st, which was applied to
the palaeanodonts, analyzed the degree of fossoriality based
on the following characters: OCA, DHW, HL, DPC, UL, LOP,
L3MC, CSAF, and OH. The 2nd, which assessed fossoriality
in Mesoscalops, took advantage of the more complete fossil
specimen of this taxon and used all characters (Fig. 1) except
UIAP and the dimensions of the 3rd phalanx (L3P, W3P, and
D3P). Discriminant analyses also were used to assess digging
mode in these taxa. In particular, because Xenocranium and
Mesoscalops have been suggested on the basis of functional
morphological analyses (Barnosky 1981; Rose and Emry
1983) to engage in some form of head-lift digging, all the
fossil taxa were examined using the head-lift digging
discriminant function developed for extant mammals. The
character set for this analysis differed somewhat from that
used for extant species because of the incompleteness of the
material available for the fossil taxa of interest. The same
group of skull characters (OCA, SL, SW, and OH) was used,
but rather than the limb characters L3MT and L3P, stepwise
character addition resulted in the inclusion of L3MC and UL
in the analysis. Finally, because M. montanensis, a proscalopid
mole that is closely related to talpid moles, has been suggested
to dig via humeral rotation, all fossil taxa also were analyzed
using the humeral-rotation digging discriminant function.
RESULTS
Within- versus between-species variance.—Standard deviations for traits were low within species compared to between
species. Standard deviations for log-transformed linear and
area measurements within most species were ,0.15 and all
were ,0.35, but between-species standard deviations ranged
from 0.6 to 0.9 for linear measurements and were roughly 1.7
for each of the area measurements. Standard deviations of
angular measurements within species were generally ,10, but
angular standard deviations between species ranged from 14 to
19; relative to the larger mean values for angular traits, these
standard deviations are only slightly greater than those for
linear and area measurements. Thus, the between-species
standard deviations were 1.6–9.7 times greater than the
within-species standard deviation for each of the characters,
suggesting that single individuals provided representative data
for the species included in our analyses.
1453
TABLE 1.—Results for quadratic discriminant analyses of
burrowing (burrowing or nonburrowing [Y/N]). The 1st row lists
results for the model with all characters included and following rows
list results for functional groups of characters. For each analysis, the
number of misidentified taxa (No. mis-id) is indicated, as are the
number of included taxa (No. inc), percentage misidentification (%
mis-id), and number of characters included (No. chars).
Analysis
No. mis-id
No. inc
% mis-id
No. chars
All characters
Head only
Limbs only
Feet only
Head + humeral length
Head + humerus
1
21
16
20
16
12
85
94
115
104
92
92
1.2
22.3
13.9
19.2
17.4
13.0
18
6
7
5
7
10
Simple indices of fossoriality.—Each of the 4 univariate or
bivariate indices examined revealed significant differences
between fossorial and nonfossorial species (DHW/HL: t 5
1.76, d.f. 5 93.1, P 5 0.041; DPC/HL: t 5 4.69, d.f. 5 114,
P , 0.0001; LOP/UL: t 5 4.73, d.f. 5 115, P , 0.0001;
CSAH/CSAF: t 5 3.34, d.f. 5 79.9, P 5 0.0006). However,
Tukey’s honestly significant different tests did not provide
evidence of significant differences between varying degrees of
fossoriality (S/F/N). Although all 4 indices could distinguish
subterranean species from other categories, these measures did
not reveal significant differences between fossorial and
nonburrowing species (mean differences, * indicates P , 0.05:
DHW/HL: S–N 5 0.11*, S–F 5 0.097*, F–N 5 0.0085; DPC/
HL: S–N 5 0.11*, S–F 5 0.076*, F–N 5 0.038; LOP/UL: S–N
5 0.11*, S–F 5 0.078*, F–N 5 0.034; CSAH/CSAF: S–N 5
1.2*, S–F 5 0.96*, F–N 5 0.21). As a final test, we compared the
2 measures of incisor angles for incisor-digging versus other
fossorial forms (the Ys from the Y/N analyses) and the occipital
condyle angle for head-lift versus other fossorial forms (the Ys
from the Y/N analyses). Both incisor angle measures revealed
significant differences between digging modes (UIAP: t 5 3.08,
d.f. 5 35.4, P 5 0.002; LIAP: t 5 2.50, d.f. 5 49.5, P 5 0.008), as
did the measure of occipital condyle angle (t 5 2.84, d.f. 5 11.7,
P 5 0.0077).
Discriminant function analyses, Y/N.—The contrast between
burrowing and nonburrowing forms (Y/N) provided the
simplest test of our multivariate morphological proxies.
Because only 2 categories were compared, only a single
canonical axis was required for discrimination; this simplified
determination of the relative weights given to each character
included in the analyses. The Y/N analysis using the full
character set successfully discriminated all but 1 species, or
1.2% of the 85 species examined (Table 1; Fig. 2).
Character partitions varied in their ability to discriminate
between burrowers and nonburrowers (Table 1). Based on
correct identification of taxa, the most-informative partition
consisted of head and humeral characters, which correctly
identified 80 (87%) of the 92 species examined, and the set of
head characters alone was worst, with only 77.7% correct.
The Y/N model placed greatest emphasis on several
characters of the humerus and ulna as well as the width of
1454
Vol. 90, No. 6
JOURNAL OF MAMMALOGY
FIG. 2.—Histogram of the 1st canonical axis for burrowing or
nonburrowing (Y/N) quadratic discriminant analysis with all
characters included. Fossorial taxa are shown in gray and nonfossorial taxa in black. The mean and 95% confidence interval for each
group are shown at the top of the figure.
the skull. In the analyses that included HL, this character was
given the strongest weighting, except for the limbs-only
partition, in which HL was a close second to LOP. The
weights placed upon SW and LOP were similar to one another
in the overall analysis. SL was not very important when all
characters were included but had a high weighting when only
skull characters were considered. UIAP, LIAP, and OCA were
of low weight in all analyses. Collectively this suggests that
linear characters of the forelimb are most influential, along
with the absolute width of the skull.
Positive values on the discriminant axis indicated an
increased tendency toward burrowing in the overall Y/N
analysis. That is, positive weights indicated positive scaling
with increased burrowing, whereas negative weights indicated
the converse. In general, burrowing animals tended to have
wider skulls and shorter humeri than nonburrowers. The lower
weight for DHW compared to HL indicates allometric scaling:
as humeri get shorter, they do not get narrower at the same
rate, producing animals with short but relatively distally broad
humeri. Weights were reversed in sign for some analyses
because the optimization algorithm in JMP 7 switched the
relative orientation of the discriminant axis such that positive
values indicated decreased burrowing behavior. Jackknifing
the Y/N analysis including all characters provided an
additional criterion for success; individually removing taxa
from this model produced a misclassification rate of 22.4%
(19 of the 85 species included; Table 2), much higher than the
result of 1.2% misclassification for the complete data set.
TABLE 2.—Misidentified taxa from jackknifing of quadratic discriminant analyses for burrowing (Y 5 burrowing, N 5 nonburrowing) and
degree of fossoriality (S 5 subterranean, F 5 fossorial, N 5 nonburrowing). Blank values indicate taxa that were correctly identified.
Burrower (Y/N)
Degree of fossoriality (S/F/N)
Family
Taxon
Actual
Predicted
Actual
Predicted
Tenrecidae
Dasyuridae
Phascolarctidae
Vombatidae
Procaviidae
Abrocomidae
Calomyscidae
Castoridae
Cricetidae
Cricetidae
Ctenomyidae
Ctenomyidae
Echimyidae
Erethizontidae
Geomyidae
Geomyidae
Heteromyidae
Heteromyidae
Heteromyidae
Heteromyidae
Heteromyidae
Heteromyidae
Octodontidae
Octodontidae
Sciuridae
Sciuridae
Sciuridae
Sciuridae
Sciuridae
Thryonomyidae
Talpidae
Talpidae
Microgale thomasi
Dasyurus maculatus
Phascolarctos cinereus
Vombatus ursinus
Procavia capensis
Abrocoma cinerea
Calomyscus elburzensis
Castor canadensis
Arborimus pomo
Mesocricetus auratus
Ctenomys magellanicus
Ctenomys haigi
Proechimys simonsi
Erethizon dorsatum
Geomys bursarius
Thomomys bottae
Chaetodipus intermedius
Dipodomys agilis
Liomys salvini
Microdipodops megacephalus
Perognathus amplus
Perognathus longimembris
Octodon degus
Spalacopus cyanus
Ammospermophilus harrisii
Sciurus carolinensis
Spermophilus beecheyi
Spermophilus beldingi
Sciurus deppei
Thryonomys gregorianus
Scalopus aquaticus
Talpa europaea
N
N
N
Y
N
N
N
N
N
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
N
N
F
F
N
N
N
F
F
N
N
F
S
S
F
F
N
F
F
N
N
Y
Y
S
S
F
F
F
F
F
F
F
S
F
F
N
N
N
N
N
N
N
F
F
F
N
N
N
S
S
F
F
F
Y
N
Y
N
Y
N
Y
N
Y
N
N
N
Y
Y
December 2009
HOPKINS AND DAVIS—MORPHOLOGICAL PROXIES FOR FOSSORIALITY
1455
TABLE 3.—Results for quadratic discriminant analyses of degree of
fossoriality (subterranean, fossorial, or nonburrowing [S/F/N]). The
1st row lists results for all characters, and following rows list results
for functional groups of characters. No. mis-id 5 number of
misidentified taxa; No. inc 5 number of included taxa; % mis-id
5 percentage misidentification; No. chars 5 number of
characters included.
FIG. 3.—First 2 canonical axes from quadratic discriminant
analysis of degree of fossoriality (subterranean, fossorial, or
nonburrowing [S/F/N]). For each category, inner and outer ovals
represent the 95% confidence intervals for the means and the 50%
prediction intervals, respectively. Subterranean taxa are represented
with gray stars, fossorial taxa with black squares, and nonfossorial
taxa with gray circles.
Discriminant function analyses, S/F/N.—The subterranean,
fossorial, or nonburrowing (S/F/N) analysis provided a test of
how well morphological data can discriminate finer distinctions between degrees of specialization to a burrowing
lifestyle. This analysis included 3 categories, which produced
2 canonical axes for discrimination (Fig. 3); the 1st axis
discriminated subterranean taxa from all other taxa and the
2nd axis discriminated between fossorial and nonburrowing
taxa. The S/F/N analysis based on the full character set
correctly identified all of the 85 included species (Table 3).
Partitioning morphological characters for the S/F/N analysis
produced results similar to those for the Y/N analysis
(Table 3). The head plus humeral characters were most
successful based on the number of species misidentified (6
of 92; 6.5%). The head characters performed worst, with 33%
(31 of 94) species misidentified.
In all of the S/F/N analyses, the 1st discriminant axis served
primarily to distinguish subterranean from other species,
whereas the 2nd axis distinguished fossorial from nonburrowing species (Fig. 2). In contrast to the Y/N analysis, the 1st
discriminant axis of the overall S/F/N model placed greatest
weight on SW, LOP, and DHW. As with the Y/N analysis, all
heavily weighted characters came from the humerus and ulna,
with the addition of SW. The 2nd discriminant axis was driven
primarily by SL (an unimportant character in the Y/N
analysis) and HL (important on both discriminant axes of S/
F/N, but with opposing signs). For the S/F/N analysis,
subterranean species had higher values on the 1st discriminant
Analysis
No. mis-id
No. inc
% mis-id
No. chars
All characters
Head only
Limbs only
Feet only
Head + humerus
0
31
17
27
6
85
94
115
104
92
0.0
33.0
14.8
26.0
6.5
18
6
7
5
10
axis, whereas fossorial species had lower values on the 2nd
discriminant axis. LOP was the most important character on
axis 1 for both S/F/N analyses in which it was included. HL
was most important on the 2nd axis of the overall S/F/N
analysis, but was 2nd in importance on the 2nd axis of the
skull plus humerus S/F/N analysis. As in the Y/N analyses,
UIAP, LIAP, and OCA were of very low weight in all S/F/N
analyses. In all cases, it was apparent that the ranked order of
importance for characters was: limbs, skull, manus and pes.
The directionalities on the canonical axes for subterranean and
fossorial species are indicated in Fig. 3. Jackknifing the S/F/N
analysis revealed that the accuracy of identification was only
slightly lower than the Y/N analysis, with 61 of 85 taxa (72%)
correctly identified (Table 2).
Discriminant function analyses, digging types.—All of the
digging type quadratic discriminant analyses were informative, with no misidentified species (Table 4). Character
weights along the discriminant axes for each of these digging
types were clearly distinct. For example, there was no overlap
between the characters included for humeral-rotation versus
head-lift digging. Because of our assumptions in model
construction, the humeral-rotation digging analysis was
limited to characters of the humerus and ulna; the most
heavily weighted of those characters were CSAH and UL, the
effects of which were in opposite directions. LOP also was
strongly weighted in the same direction as UL. Collectively,
these results indicated that humeral-rotation diggers had
shorter olecranon processes and shorter ulnae (relative to the
cross-sectional area of their humeri) than other digging modes,
TABLE 4.—Results for quadratic discriminant analyses addressing
specific locomotor types (comparing each category only with other
burrowers). No. mis-id 5 number of misidentified taxa; No. inc 5
number of included taxa; % mis-id 5 percentage misidentification;
No. chars 5 number of characters included.
Digging type
No. mis-id
No. inc
% mis-id
No. chars
Humeral-rotation
Head-lift
Incisor
Scratch
0
0
0
0
63
59
51
51
0
0
0
0
5
6
15
15
1456
JOURNAL OF MAMMALOGY
as revealed by the differing signs of the weights of these
variables for humeral-rotation diggers.
In contrast, head-lift digging was diagnosed through a
combination of characters from the skull, manus, and pes; no
limb characters distinguished this digging mode. The most
heavily weighted characters for head-lift diggers were SL,
followed by OH, SW, L3MT, and characters of the 3rd
phalanx of the manus. As with humeral-rotation digging, headlift digging was associated with lower values on the
discriminant axis. Proportionately, head-lift diggers had
shorter, wider skulls with a taller occiput than other diggers.
Somewhat surprisingly, the most heavily weighted character for incisor diggers was DHW. The next most important
characters were L3MT and SL. LOP and D3P were of some
importance, followed by L3MC, DPC, OH, and L3P. Thus,
compared to other digging modes, incisor diggers had longer
skulls, which may have been in part a result of the inclusion of
the forward-projecting part of the incisor in our measure of the
total length of the skull. Incisor diggers also exhibited
narrower distal humeri, shorter metatarsals, and longer
olecranon processes. Counter to our expectation, incisor
angles did not contribute to discriminating incisor diggers
from other animals. Incisor diggers had higher canonical
values than species with other digging modes.
The most heavily weighted characters for scratch diggers
were UL and SL. HL was almost as strongly weighted as the
preceding 2 characters, but all other characters had much
lower weights. Scratch digging was associated with lower
canonical values indicating that, in comparison with other
digging modes, scratch diggers had proportionately longer
ulnae and humeri and shorter skulls. This comparison was
only with other diggers and did not indicate that scratch
diggers have longer ulnae compared to nondigging mammals.
Discriminant function analyses, life habit.—Five different
life-habit categories (semiaquatic, arboreal, terrestrial, fossorial, and subterranean) were compared using all 18 measured
characters and, consequently, this analysis included 4
discriminant axes. Exploration of the parameter space
indicated that the regularized discriminant analysis for life
habit was optimized with a lambda of 0.03 and gamma of 0.
Based on the number of species misidentified (0 of 84), the
regularized discriminant analysis was highly informative. The
number of discriminant axes made interpretation of the results
complex. The axes discriminated groups in the following
order: axis 1, subterranean (higher values); axis 2, semiaquatic
(higher values); axis 3, arboreal (lower values); axis 4,
fossorial (higher) and terrestrial (lower). Interestingly, the 1st
axis pulled out the subterranean species, as did the 1st axis of
every S/F/N analysis. This is not particularly surprising, given
that the character set was chosen to identify burrowers. The
strongest loadings on the 1st axis were for LOP, DHW, SW,
and HL. LOP and SW both increased with increasing
subterranean tendency, whereas the other characters tended
to decrease with increasing specialization for subterranean
life. The 3 most heavily weighted characters for subterranean
life habit (DHW, HL, and LOP) in this analysis were the same
Vol. 90, No. 6
as for subterranean mammals in the S/F/N analysis. The most
important characters for distinguishing semiaquatic species
were HL, L3MT, DHW, and LOP. Although HL had a
negative relationship with specialization for semiaquatic life,
the relationship for the other 3 characters was positive,
indicating that semiaquatic taxa had shorter, wider humeri,
longer olecranon processes, and longer 3rd metatarsals. The
3rd axis emphasized SL, SW, HL, UL, and D3P to distinguish
arboreal species. Although the relationship of arboreal taxa
with both SL and HL was negative, the relationships for the
other 3 characters were positive, suggesting that arboreal taxa
have shorter, wider skulls, shorter humeri, longer ulnae, and
deeper 3rd phalanges. Finally, on the 4th axis, W3P was the
most heavily weighted character for distinguishing fossorial
from terrestrial mammals, although SL and SW also were
important along this axis. Thus, fossorial mammals have wider
3rd phalanges and smaller skulls (probably a result of
generally smaller body size) than their terrestrial counterparts.
SW was not as heavily loaded as SL, suggesting that the skulls
of fossorial taxa are proportionately wider than terrestrial
mammals, even though the former are narrower in absolute
dimensions.
Inferring fossoriality in fossil taxa.—Although the reduced
character set necessitated by incompletely preserved fossil
specimens did not distinguish extant subterranean, fossorial,
and nonfossorial taxa as effectively as analyses using all
characters, the 3 groups were still significantly different
(Fig. 4), with a misidentification rate of 15%. Data points for
all of the palaeanodonts examined fell near the periphery of
the distribution for extant taxa, which somewhat reduced our
confidence in discriminations made by these models. For
example, M. tatusia plotted well outside the range of extant
taxa, giving no basis for inferring the degree of fossoriality. T.
pearcei plotted on the periphery of the range for subterranean
taxa, a finding that appears to be inconsistent with the
suggestion by Rose et al. (1991) that this species was
terrestrial or fossorial. Conversely, X. pileorivale plotted
nearest several fossorial taxa (mostly fossorial talpids), even
though this species has been suggested to be subterranean
(Rose and Emry 1983). E. zygus and P. ignavus are at the
periphery of the distribution of fossorial taxa, although the
sparse distribution of data in this region of the morphospace
provides only weak support for the conclusion that these
species were fossorial. The most convincing results were
obtained for M. montanensis, a proscalopid mole that plotted
well within the 50% prediction interval for subterranean taxa
(Fig. 4). This finding was consistent with an exhaustive study
of the functional morphology of this species that concluded
that M. montanensis was subterranean (Barnosky 1981).
The analyses of digging modes for fossil species produced
somewhat clearer results. The head-lift digging discriminant
function plotted Xenocranium and Mesoscalops solidly within
the range of extant head-lift diggers (Fig. 5); although not
evident within the resolution of this figure, both occurred
below the minimum value of the canonical axis for non–headlift diggers. The other fossil taxa were within the range of
December 2009
HOPKINS AND DAVIS—MORPHOLOGICAL PROXIES FOR FOSSORIALITY
1457
FIG. 5.—Histogram of the 1st canonical axis for the head-lift
digging quadratic discriminant analysis with fossil taxa included.
Arrows indicate positions of fossil species. Head-lift diggers are
shown in gray and non–head-lift diggers in black. Means and 95%
confidence intervals for each group are shown at the top of the figure.
Counts of species for each bin do not include the fossil taxa.
outside the range of values for living humeral-rotation diggers
(Fig. 6). No palaeanodont has been suggested to engage in this
mode of digging, although Barnosky (1981) reconstructed a
digging stroke for Mesoscalops that includes a component of
humeral rotation.
DISCUSSION
Given the results of the discriminant analyses, our
morphological indices appeared to be effective at distinguishing burrowers from nonburrowers and, in some cases,
distinguishing among degrees of fossoriality. Overall, our Y/
N analyses misidentified ,2% of taxa with the full character
set and ,25% with any of the character subsets. When taxa
were removed from the analyses for the jackknife test,
misidentification of burrowers increased to 22.4%. Although
this is still a relatively low rate of misidentification, it suggests
that there is a risk of incorrectly inferring degree of
fossoriality for taxa not used in generating the discriminant
FIG. 4.—First 2 canonical axes from quadratic discriminant
analyses of degree of fossoriality for fossil taxa. For each category
(subterranean, fossorial, or nonburrowing [S/F/N]), inner and outer
ovals represent the 95% confidence intervals for the means and the
50% prediction intervals, respectively. Subterranean taxa are
represented with gray stars, fossorial taxa with black closed squares,
and nonfossorial taxa with gray closed circles. A) Canonical axes 1
and 2 for the analysis including palaeanodonts, which are plotted with
open squares. Taxa are identified as follows: Ez, Escavadodon zygus;
Pi, Palaeanodon ignavus; Xp, Xenocranium pileorivale; Tp, Tubulodon pearcei; Mt, Metacheiromys tatusia. B) Canonical axes 1 and 2
for the analysis including Mesoscalops montanensis, which is plotted
with an open circle.
non–head-lift diggers, with only T. pearcei occurring within
the zone of overlap between the 2 categories. The humeralrotation digging function plotted Mesoscalops within the 95%
confidence interval for the mean for extant humeral-rotation
diggers, whereas all palaeanodont taxa examined plotted
FIG. 6.—Histogram of the 1st canonical axis for the humeralrotation quadratic discriminant analysis with fossil taxa included.
Arrows indicate positions of fossil species. Humeral-rotation diggers
are shown in gray and non–humeral-rotation diggers in black. Means
and 95% confidence intervals for each group are shown at the top of
the figure. Counts of species in each bin do not include the fossil taxa.
1458
JOURNAL OF MAMMALOGY
function. Misidentifications were more common in groups for
which there was poor taxonomic coverage relative to the
group’s diversity (e.g., Afrotheria, Marsupialia, and basal
Hystricognathi) and in groups with diverse locomotor habits
within narrow taxonomic boundaries (e.g., Octodontidae,
Heteromyidae, Sciuridae, and Proechimys). All subterranean
taxa were correctly identified as burrowers, and the majority
of members of each taxonomic group mentioned above also
were correctly identified to digging type, including all
members of some groups (e.g., Talpidae and Dasypodidae)
with substantial locomotor diversity.
The S/F/N analysis yielded more robust results, presumably
because it better captured the biological differences underlying the morphological data. The biggest morphological break
was between subterranean and all other taxa, suggesting that
lumping subterranean with fossorial taxa in the Y/N analysis
made the boundary between burrowers and nonburrowers
more difficult to detect. The S/F/N analysis was only slightly
less robust to the jackknife test than the Y/N analysis; many of
the same taxa were misidentified, suggesting that these taxa
were of intermediate morphology. The misidentifications that
occurred were primarily underpredictions of fossoriality; there
were no errors in which taxa were incorrectly predicted to be
subterranean and most errors of identification between
fossorial and nonfossorial taxa resulted from the former being
identified as nonburrowing. There were no errors that
identified nonfossorial taxa as subterranean or the converse.
The success of these analyses was fundamentally linked to the
roles of the different morphological characters in the biology of
the animals, with difficulties arising when animals combined
characters in unique ways (i.e., ways not observed in other
animals included in the analysis). Despite the morphological
diversity of burrowing forms, it would be useful to possess a
simple index of fossoriality that was applicable across
evolutionarily and morphologically divergent taxa. Thus, it is
important to consider what information individual characters
reveal about fossorial locomotion. All of the proposed simple
indices were significant indicators of fossoriality and thus, a
change in 1 of the underlying characters is likely to indicate
change in the degree of fossorial adaptation. However, all of
these characters display substantial overlap between the
distributions for fossorial and nonfossorial taxa and hence,
individually, they are ill-suited to predict the locomotor habit of
a species for which behavior cannot be observed. Phylogenetic
information that establishes the direction and magnitude of the
evolution of such characters would likely increase the utility of
these simple indices, but such analyses are outside of the scope
of the present study.
Several of the morphological traits included in this study were
consistently important for discriminating among degrees of
specialization for burrowing. For example, most analyses found
SL and SW to be important characters for discriminating among
digging types, which suggests that another potential simple
index of fossoriality could be generated from the proportions of
the skull, SL/SW. This relationship results from use of the head
to pack soil against the roof and walls of the burrow, to remove
Vol. 90, No. 6
soil from the burrow, and, among head-lift diggers, to excavate
burrows (Hildebrand 1985; Nevo 1999; Stein 2000). The finding
that these same characters were important in differentiating
diggers from nondiggers and in separating fossorial from
subterranean taxa suggests that the morphology of subterranean
taxa (as measured by the characters included in this analysis) is
essentially a more extreme version of the morphology of
fossorial mammals. This finding is significant, because
subterranean taxa employ a variety of specialized mechanisms
of digging that are not used by fossorial or nonfossorial taxa.
When we consider the region of the skeleton from which
characters were derived, it becomes apparent that some portions
of the skeleton were more informative than others in predicting
fossoriality. Analyses based on limb measurements were the
most informative, a finding that is not particularly surprising
given that the vast majority of burrowers use their limbs to move
soil. Skull characters were the 2nd most informative, indicating
that use of the head in burrow excavation and soil transport is
relatively widespread among fossorial taxa. The feet provided
the least-informative characters. This was somewhat surprising,
because the feet are the part of a burrowing animal that interacts
most directly with the substrate. The type of substrate, the
method of burrowing, and phylogenetic effects all contribute to
variation in the structure of the manus and pes, and it is possible
that these factors overwhelmed the commonalities in form and
function generated by a burrowing habit. It is encouraging that
even with limited information available from the skull and limbs
(i.e., head and humerus), it was possible to discriminate between
subterranean, fossorial, and nonfossorial mammals. This
suggests that even with relatively incomplete skeletal material
it should be possible to use the discriminant analyses presented
here to make consistent, quantitative inferences about the degree
of fossorial adaptation in rare or extinct taxa.
Discriminant analyses based on mode of digging (scratch,
incisor, head-lift, and humeral-rotation) were informative
regarding both the best characters for distinguishing among
these modes of excavation and the links between evolution
and these differences in ecology. All of these modes of
digging could be accurately distinguished among burrowing
mammals using the characters in this analysis; in some cases,
the same distinctions were possible with a more limited subset
of characters. Because our quadratic discriminant analyses
used only burrowing taxa to generate discriminant functions,
consideration of the nonburrowing taxa in our data set
provides interesting insights into the relationship between
burrowing mode and evolution of fossoriality. The scratchdigging discriminant function identified all nonfossorial taxa
as scratch diggers, implying that the morphology necessary for
scratch digging is generally present in mammals (or at least in
the mammals included in this analysis), perhaps as a primitive
suite of traits. In contrast, the other modes of digging (incisor,
head-lift, and humeral-rotation) were never attributed to
nonburrowers, suggesting that these modes of burrowing
involve greater evolutionary specializations for fossoriality.
Despite the small number of taxa—and, hence, the small
number of characters—available for discriminant analyses of
December 2009
HOPKINS AND DAVIS—MORPHOLOGICAL PROXIES FOR FOSSORIALITY
the more specialized digging modes (humeral-rotation, headlift, and incisor), these mechanisms of excavation were easily
recognized from skeletal morphology. However, the most
useful diagnostic characters differed dramatically from one
mechanism to another. Head-lift digging uses a greater variety
of body parts to excavate soil and impacts characters from the
head, limbs, and feet, although the most-diagnostic characters
were related to the shape of the skull (e.g., high, forwardslanted occiput and short, wide skull). In contrast, humeralrotation digging was best identified using characteristics of the
humerus and ulna; interestingly, ulnar characters were more
heavily weighted, even though alterations of the humerus are
more visually striking. Surprisingly, angle of procumbency
was not a strong predictor of incisor digging, which was best
identified using a variety of characters. The primary
differences between incisor diggers and other burrowing taxa
seemed to reflect reduced specialization of the skull and
postcranial skeleton in incisor diggers, but the high weight on
SL in incisor diggers reflected the known need for a longer
skull to provide the anchorage for their forward-projecting
upper incisors (Lessa 1990). This mosaic of characters may
occur because incisor digging arises primarily in subterranean
taxa and requires less skeletal modification than other modes
of digging that involve more areas of the body.
Predicting the burrowing habits of fossil taxa provided an
independent test of the performance of our discriminant
functions as descriptors of fossorial adaptations. All of the
palaeanodonts examined fell near the periphery of the
distributions for extant species, suggesting that they used
morphospace differently from the living mammals on which
the discriminant function was based. Interestingly, these
analyses suggest fossorial to subterranean life habits for all of
the palaeanodonts examined—results that are generally consistent with previous studies of these taxa—although some taxa
differed from prior reconstructions in the predicted degree of
fossoriality (Rose and Emry 1983; Rose et al. 1991). Analyses
for specific digging types were more effective, with the
quadratic discriminant analyses for head-lift and humeralrotation predicting the same mode of burrowing as more
exhaustive studies of functional morphology. In particular, the
reconstruction of Mesoscalops as a subterranean burrower with a
mix of head-lift and humeral-rotation digging morphologies is
encouraging given that it is a member of an extinct clade of
burrowing insectivores (Barnosky 1981, 1982). Overall, these
findings suggest that our analysis can be applied successfully to
extinct taxa. This does not imply that such proxies should
replace detailed studies of functional morphology but, rather,
that morphological proxies may allow inferences regarding
function when available skeletal material is incomplete.
Although the performance of our discriminant analyses may
be less effective with taxa that differ dramatically in morphology
or evolutionary history from the species on which the
discriminant functions were based, it is encouraging that
consistently successful discrimination can be achieved with
relatively limited groups of morphological characters. It is likely
that common physical demands on fossorial mammals have
1459
generated similar suites of characters in distantly related
organisms. The results of jackknifing the discriminant analyses
suggest that there is some phylogenetic signal in the morphological adaptations of mammals to digging, and hence a
phylogenetically controlled examination of burrowing morphology would be informative. However, it also is apparent that
burrowers of different types can be recognized on the basis of a
very limited set of skeletal characters. As part of future studies, it
would be important to consider the evolution of fossoriality in a
phylogenetic context. The analysis presented here is intended to
provide a method for quantifying the degree of adaptation for
burrowing in the absence of detailed phylogenetic or functional
morphological information. As such, these analyses allow
inferences regarding burrowing behavior in diverse groups of
mammals for which behavior cannot be observed directly.
SUPPLEMENTARY DATA
Online appendices containing the complete morphological
data, set analyzed here, as well as the eigenvalues and
eigenvectors for discriminant analyses discussed here can be
found at the University of Oregon Scholar’s Bank (http://hdl.
handle.net/1794/9604).
ACKNOWLEDGMENTS
We thank L. Gordon (National Museum of Natural History), C.
Conroy (University of California’s Museum of Vertebrate Zoology),
and P. Holroyd (UCMP) for providing access and guidance to
collections used in gathering data for this study. Research was
supported by a postdoctoral fellowship awarded to SSBH at the
National Evolutionary Synthesis Center (National Science Foundation EF0423641). Statistical methods were improved by discussions
with M. Whitlock and S. Otto, as well as other postdoctoral
researchers at the National Evolutionary Synthesis Center. We are
grateful to E. Lacey, E. Heske, and 2 anonymous reviewers for their
constructive criticism of earlier drafts of the manuscript.
LITERATURE CITED
BARNOSKY, A. D. 1981. A skeleton of Mesoscalops (Mammalia:
Insectivora) from the Miocene Deep River Formation, Montana,
and a review of the proscalopid moles: evolutionary, functional,
and stratigraphic relationships. Journal of Vertebrate Paleontology
1:285–339.
BARNOSKY, A. D. 1982. Locomotion in moles (Insectivora, Proscalopidae) from the middle Tertiary of North America. Science
216:183–185.
BIKNEVICIUS, A. R. 1993. Biomechanical scaling of limb bones and
differential limb use in caviomorph rodents. Journal of Mammalogy 74:95–107.
COLBERT, E. H. 1942. An edentate from the Oligocene of Wyoming.
Notulae Naturae 109:1–16.
COURANT , F., B. DAVID, B. L AURIN , AND J. CHALINE . 1997.
Quantification of cranial convergences in arvicolids (Rodentia).
Biological Journal of the Linnean Society 62:505–517.
DEGUSTA, D., AND E. S. VRBA. 2003. A method for inferring
paleohabitats from the functional morphology of bovid astragali.
Journal of Archaeological Science 30:1009–1022.
1460
JOURNAL OF MAMMALOGY
FRIEDMAN, J. H. 1989. Regularized discriminant analysis. Journal of
the American Statistical Association 84:165–175.
GASC, J. P., F. K. JOUFFROY, AND S. RENOUS. 1986. Morphofunctional
study of the digging system of the Namib Desert golden mole
(Eremitalpa granti namibensis): cinefluorographical and anatomical analysis. Journal of Zoology (London) 208:9–35.
GAZIN, C. L. 1952. The lower Eocene Knight Formation of western
Wyoming and its mammalian faunas. Smithsonian Miscellaneous
Collections 117:1–82.
HILDEBRAND, M. 1985. Digging of quadrupeds. Pp. 89–109 in
Functional vertebrate morphology (M. Hildebrand, D. M. Bramble,
K. F. Liem, and D. B. Wake, eds.). Belknap Press, Cambridge,
Massachusetts.
HILDEBRAND, M., AND G. E. GOSLOW, JR. 2001. Analysis of vertebrate
structure. John Wiley and Sons, Inc., New York.
HOPKINS, S. S. B. 2005. The evolution of fossoriality and the adaptive
role of horns in the Mylagaulidae (Mammalia: Rodentia).
Proceedings of the Royal Society of London, B. Biological
Sciences 272:1705–1713.
LANDRY, S. O., JR. 1957. Factors affecting the procumbency of rodent
upper incisors. Journal of Mammalogy 38:223–234.
LESSA, E. P. 1990. Morphological evolution of subterranean
mammals: integrating structural, functional, and ecological perspectives. Pp. 211–230 in Evolution of subterranean mammals at
the organismal and molecular levels (E. Nevo and O. A. Reig,
eds.). Wiley-Liss, New York.
LESSA, E. P., AND C. S. THAELER, JR. 1989. A reassessment of
morphological specializations for digging in pocket gophers.
Journal of Mammalogy 70:689–700.
MATTHEW, W. D., AND W. GRANGER. 1918. A revision of the lower
Eocene Wasatch and Wind River faunas. Part V: Insectivora
(continued), Glires, Edentata. Bulletin of the American Museum of
Natural History 38:565–657.
NEVO, E. 1961. Observations on Israeli populations of the mole-rat
Spalax ehrenbergi Nehring 1898. Mammalia 25:127–44.
NEVO, E. 1999. Mosaic evolution of subterranean mammals:
regression, progression, and global convergence. Oxford University Press, New York.
REICHMAN, O. J., AND J. U. M. JARVIS. 1989. The influence of three
sympatric species of fossorial mole-rats (Bathyergidae) on
vegetation. Journal of Mammalogy 70:763–771.
REICHMAN, O. J., AND E. W. SEABLOOM. 2002. The role of pocket
gophers as subterranean ecosystem engineers. Trends in Ecology &
Evolution 17:44–49.
ROSE, K. D. 1978. A new Paleocene epoicotheriid (Mammalia), with
comments on Palaeanodonta. Journal of Paleontology 52:658–674.
Vol. 90, No. 6
ROSE, K. D., AND R. J. EMRY. 1983. Extraordinary fossorial
adaptations in the Oligocene palaeanodonts Epoicotherium and
Xenocranium (Mammalia). Journal of Morphology 175:33–56.
ROSE, K. D., L. KRISHTALKA, AND R. K. STUCKY. 1991. Revision of the
Wind River Faunas, early Eocene of central Wyoming. Part 11.
Palaeanodonta (Mammalia). Annals of Carnegie Museum 60:
63–82.
ROSE, K. D., AND S. G. LUCAS. 2000. An early Paleocene palaeanodont
(Mammalia, ? Pholidota) from New Mexico, and the origin of
Palaeanodonta. Journal of Vertebrate Paleontology 20:139–156.
SHIMER, H. W. 1903. Adaptations to aquatic, arboreal, fossorial, and
cursorial habits in mammals. III. Fossorial adaptations. American
Naturalist 37:819–825.
SIMPSON, G. G. 1931. Metacheiromys and the Edentata. Bulletin of the
American Museum of Natural History 59:295–381.
STEIN, B. 2000. Morphology of subterranean rodents. Pp. 19–61 in
Life underground: the biology of subterranean rodents (E. A.
Lacey, J. L. Patton, and G. N. Cameron, eds.). University of
Chicago Press, Chicago, Illinois.
STIRLING, E. C. 1891. Description of a new genus and species of
Marsupialia, ‘‘Notoryctes typhlops.’’ Transactions of the Royal
Society of South Australia 14:154–187.
STIRLING, E. C. 1894. Supplemental note on the osteology of
Notoryctes typhlops. Transactions of the Royal Society of South
Australia 18:1–2.
VASSALLO, A. I. 1998. Functional morphology, comparative behaviour, and adaptation in two sympatric subterranean rodents genus
Ctenomys (Caviomorpha: Octodontidae). Journal of Zoology
(London) 244:415–427.
VIZCAINO, S. F., AND N. MILNE. 2002. Structure and function in
armadillo limbs (Mammalia: Xenarthra: Dasypodidae). Journal of
Zoology (London) 257:117–127.
WARBURTON, N. M. 2003. Functional morphology and evolution of
marsupial moles (Marsupialia; Notoryctemorphia). Ph.D. dissertation, University of Western Australia, Perth, Australia.
ZAR, J. 1999. Biostatistical analysis. 4th ed. Prentice Hall, Inc., Upper
Saddle River, New Jersey.
ZHANG, Y. 2007. The biology and ecology of plateau zokors
(Eospalax fontanierii). Pp. 237–250 in Subterranean rodents: news
from underground (S. Begall, H. Burda, and C. E. Schleich, eds.).
Springer, Berlin, Germany.
Submitted 14 August 2008. Accepted 24 March 2009.
Associate Editor was Eileen A. Lacey.