Content uploaded by John J. Wiens
Author content
All content in this area was uploaded by John J. Wiens on Dec 05, 2023
Content may be subject to copyright.
vol. 161, no. 1 the american naturalist january 2003
Explaining Species Richness from Continents to Communities:
The Time-for-Speciation Effect in Emydid Turtles
Patrick R. Stephens
1,2,
*
and John J. Wiens
2,†
1. Department of Biological Sciences, University of Pittsburgh,
Pittsburgh, Pennsylvania 15260;
2. Section of Amphibians and Reptiles, Carnegie Museum of
Natural History, Pittsburgh, Pennsylvania 15213-4080
Submitted May 17, 2001; Accepted July 17, 2002;
Electronically published December 30, 2002
abstract: Speciation is the process that ultimately generates species
richness. However, the time required for speciation to build up di-
versity in a region is rarely considered as an explanation for patterns
of species richness. We explored this “time-for-speciation effect” on
patterns of species richness in emydid turtles. Emydids show a strik-
ing pattern of high species richness in eastern North America (es-
pecially the southeast) and low diversity in other regions. At the
continental scale, species richness is positively correlated with the
amount of time emydids have been present and speciating in each
region, with eastern North America being the ancestral region.
Within eastern North America, higher regional species richness in
the southeast is associated with smaller geographic range sizes and
not greater local species richness in southern communities. We sug-
gest that these patterns of geographic range size variation and local
and regional species richness in eastern North America are caused
by glaciation, allopatric speciation, and the time-for-speciation effect.
We propose that allopatric speciation can simultaneously decrease
geographic range size and increase regional diversity without in-
creasing local diversity and that geographic range size can determine
the relationship between a,b, and gdiversity. The time-for-
speciation effect may act through a variety of processes at different
spatial scales to determine diverse patterns of species richness.
Keywords: species richness, macroecology, phylogeny, speciation,
Emydidae.
* Corresponding author. Present address: Department of Ecology and Evo-
lution, Stony Brook University, Stony Brook, New York 11794-5245; e-mail:
pstephen⫹@pitt.edu.
†
Present address: Department of Ecology and Evolution, Stony Brook
University, Stony Brook, New York 11794-5245; e-mail: wiensj@
carnegiemuseums.org.
Am. Nat. 2003. Vol. 161, pp. 112–128. 䉷2003 by The University ofChicago.
0003-0147/2003/16101-010178$15.00. All rights reserved.
Explaining patterns of species richness is a central problem
in ecology and biogeography. Speciation is the process that
ultimately generates species richness, but in the recent lit-
erature speciation is rarely invoked directly when explain-
ing patterns of species diversity. The ability of speciation
to build up species richness in a region may be constrained
by two factors: space and time (Rosenzweig 1995). The
extent to which speciation (and thereby species richness)
is constrained by area has been demonstrated for Anolis
lizards on Caribbean islands by Losos and Schluter (2000).
Their results suggest that speciation can occur within rel-
atively small areas (e.g., islands as small as approximately
3,000 km
2
). Time may be as important as area in limiting
species richness. For example, when members of a lineage
first colonize an island or continental region, it may take
hundreds of thousands or even millions of years for spe-
ciation to occur and build up species richness in situ (given
that modern species are typically 300,000 to 7 million years
old; Avise 2000). However, the constraints of time on spe-
ciation have been largely ignored in recent studies of spe-
cies richness (but see Gaston and Blackburn 1996; Brown
et al. 2000; McPeek and Brown 2000). We refer to disparity
in species richness between regions or habitats that is
caused by limited time for speciation as the time-for-
speciation effect.
Despite its current neglect, the idea that patterns of
species richness are related to how long members of a
group have been present in a region has a long history.
During most of the twentieth century, it was assumed that
a group of organisms will be most diverse in the region
where the group originated (Willis 1922), and this was a
widely accepted explanation for the high species richness
of many groups in tropical regions (e.g., Axelrod 1952;
Darlington 1957; Cronquist 1968; Stebbins 1974). How-
ever, this hypothesis has become relatively neglected in
recent years. For example, it is not mentioned in a recent
community ecology textbook (Morin 1999) and is not
presented as a current hypothesis in a recent biogeography
textbook (Brown and Lomolino 1998). The importance of
time to patterns of species richness has been discussed in
some recent literature as the ecological and evolutionary
Species Richness in Emydid Turtles 113
time hypotheses (Pianka 1966, 2000; Gaston and Black-
burn 2000). These hypotheses state that species richness
in a region may be limited when there is too little time
for either colonization (ecological time) or niche diver-
sification and speciation (evolutionary time). The evolu-
tionary time hypothesis, clearly related to the time-for-
speciation effect, has only been discussed in relatively few
recent studies (e.g., Rohde 1986, 1992; Bush et al. 1990;
Gaston and Blackburn 1996; Andersen 1999). Conversely,
despite the lack of interest in the general relationship be-
tween time and geographic patterns of species richness,
there is intense research interest in groups that seem to
be undergoing unusually rapid speciation (e.g., Baldwin
and Sanderson 1998; Albertson et al. 1999; Lovette and
Bermingham 2000; Schluter 2000; Magallo´n and Sander-
son 2001; Richardson et al. 2001a, 2001b).
In this article we test the importance of the time-for-
speciation effect on regional and local patterns of species
richness in emydid turtles. The Emydidae is an excellent
model system because the overall species richness and geo-
graphic distributions of individual species appear to be
relatively well known (Iverson 1992). The Emydidae con-
tains 40 currently recognized species in 10 genera (Ernst
and Barbour 1989; Ernst et al. 1994; Vanzolini 1995) and
includes many familiar North American turtles such as
the painted turtles (Chrysemys), box turtles (Terrapene),
and sliders (Trachemys). Although a few ambiguities re-
main in the species-level taxonomy of some emydids (Leg-
ler 1990; Seidel 1994; Lenk et al. 1999; P. R. Stephens and
J. J. Wiens, unpublished manuscript), only a handful of
species have been described in the last quarter century
(Ward 1984; Lovich and McCoy 1992; Vanzolini 1995),
which suggests that few species remain undescribed. Emy-
dids are found in Europe, North America, Central Amer-
ica, South America, and the West Indies (Ernst and Bar-
bour 1989). The family is ecologically diverse and includes
herbivores, molluscivores, insectivores, and omnivores;
species are found in brackish water, freshwater, and ter-
restrial habitats (Ernst et al. 1994). Emydids are a common
and ecologically important group in North America and
may be the major vertebrate component (in terms of bio-
mass) of some freshwater communities due to large body
size and high local abundance (Bury 1979).
There are two major patterns of emydid species richness
to explain. First, eastern North America has much higher
species richness than any other continental region, both
at regional and local scales (fig. 1A). Regions of low emydid
species richness are ecologically varied (temperate, tropi-
cal, arid, mesic), suggesting that no single environmental
factor explains their low diversity. Furthermore, many
emydid-poor regions are extremely large (e.g., Europe,
South America), which suggests that speciation is not con-
strained by the size of these regions (Losos and Schluter
2000). Given these observations, the time-for-speciation
effect is an obvious hypothesis to test.
Second, there is higher species richness in southeastern
North America than in northeastern North America (fig.
1B). If the range of the Emydidae in eastern North America
is divided at its latitudinal midpoint (37.5⬚), there are ap-
proximately twice as many species in the southern half as
in the northern half (Iverson 1992; Ernst et al. 1994). Due
to widespread species that occur in both southeastern and
northeastern North America, the relative age of the family
in each region is not easily determined, and the time-for-
speciation effect could not be addressed directly. Patterns
of species richness within eastern North America required
an interconnected series of analyses to explain.
We first tested the hypothesis that regional patterns of
species richness in eastern North America are related to
differences in local species richness. Many studies have
demonstrated that regional and local species richness tend
to be correlated (e.g., Cornell 1985; Hawkins and Comp-
ton 1992; Caley and Schluter 1997; Shurin et al. 2000).
Given that more species of emydids are found in south-
eastern than in northeastern North America (fig. 1B), we
expected that more emydids would coexist locally in south-
eastern communities. This pattern would correspond to
one of the most widespread trends in species richness: a
negative correlation between latitude and species richness
(i.e., increasing species richness from temperate to tropical
regions; Fischer 1960; Pianka 1966; Brown and Lomolino
1998). Numerous hypotheses have been proposed to ex-
plain this pattern (Pianka 1966; Rohde 1992; Rosenzweig
1995; Brown and Lomolino 1998; Morin 1999), and many
invoke local-scale ecological processes (e.g., abiotic “harsh-
ness,” productivity). A negative correlation between local-
scale species richness and latitude in eastern North Amer-
ica would support the hypothesis that local-scale ecological
processes caused the higher emydid species richness of
southeastern North America. Surprisingly, our results re-
jected this simple explanation, showing instead that local
species richness is similar between southeastern and north-
eastern North America (fig. 1B).
We next tested the hypothesis that the higher species
richness of southeastern North America is related to lat-
itudinal variation in average geographic range size. Many
authors have addressed the relationship between species
richness and range size (e.g., Stevens 1989; Letcher and
Harvey 1994; Brown 1995; Stone et al. 1996; Brown and
Lomolino 1998; Gaston et al. 1998; Willig and Lyons 1998;
Colwell and Lees 2000). A common pattern in temperate
regions is for species at higher latitudes to have larger
geographic ranges than those at lower latitudes (i.e., Ra-
poport’s rule; Stevens 1989; Brown and Lomolino 1998;
Gaston et al. 1998). If the average range size of south-
eastern emydid species is smaller (as implied by Hecnar
114 The American Naturalist
Figure 1: Patterns of local and regional species richness in emydid turtles between (A) continental regions and (B) southeastern and northeastern
North America. Subspecies of Trachemys scripta endemic to Middle America are treated as species for this figure (the number of currently recognized
species in Middle America is only four). Values of local species richness were calculated by averaging the maximum number of species estimated
for each community across the communities in each region.
[1999]), it could allow more species to be “packed” into
the region (Stevens 1989) without increasing local
diversity.
We also investigated the mode of speciation in emydids.
Mode of speciation (e.g., allopatric, sympatric) is rarely
addressed in studies of species richness, but it can be im-
portant to the overall picture of species richness because
different modes will affect diversity at different scales (i.e.,
regional vs. local). For example, a species generated by
sympatric speciation will increase both local and regional
species richness because it is added to existing commu-
nities. Conversely, a species generated by allopatric spe-
ciation is not “added” to a local community because it is
a geographic fragment of an ancestral species. Such a spe-
cies will increase regional species richness but will have
no immediate effect on local species richness. Further-
more, species generated by allopatric speciation may often
have smaller geographic ranges (at least initially) than the
Species Richness in Emydid Turtles 115
species they are descended from (Brown and Lomolino
1998). Allopatric speciation would provide a mechanism
that could increase regional species richness and decrease
average geographic range size in southeastern North
America while maintaining similar local species richness
between the southeast and the northeast.
In this article we show how the time-for-speciation ef-
fect has either directly or indirectly caused the two major
trends in emydid species richness (fig. 1): the higher species
richness of eastern North America compared to peripheral
continental regions and the higher regional species rich-
ness of southeastern North America compared to north-
eastern North America. We also discuss the broader im-
plications of our findings for latitudinal trends in species
richness, geographic range size variation, and the rela-
tionship between regional and local species richness.
Material and Methods
For all analyses, data on species distributions were ob-
tained from Smith and Smith (1979; Mexico), Seidel (1988;
West Indies), Legler (1990; Mexico and Central America),
Iverson (1992; Eastern Hemisphere and South America),
and Ernst et al. (1994; North America). The phylogeny
used was based on a combined analysis of morphological
and molecular data that included all but one currently
recognized species of emydid turtle (P. R. Stephens and J.
J. Wiens, unpublished manuscript). Morphological data
consisted of 237 parsimony-informative characters of os-
teology, external morphology, penial morphology, egg shell
morphology, development, behavior, and allozymes. Mo-
lecular data included sequences from the cytochrome b,
control region, ND4, and 16S mitochondrial gene regions
(547 parsimony-informative characters; from the litera-
ture, e.g., Lamb et al. 1994; Lenk et al. 1999; Feldman and
Parham 2002). Although some parts of this tree are weakly
supported by bootstrapping (Felsenstein 1985a), this weak
support was associated with including a few taxa with
highly incomplete data. When these highly incomplete taxa
are removed, the result is a “backbone” tree with generally
high levels of bootstrap support (P. R. Stephens and J. J.
Wiens, unpublished manuscript). This strongly supported
tree is consistent with the comprehensive tree used in this
study.
Patterns of Species Richness between Continental Regions
To address continental patterns of species richness, we
examined the relationship between regional diversity and
the estimated time when emydids colonized a region
(based on levels of molecular and morphological diver-
gence) using least squares linear regression. We first an-
alyzed the relationship between the timing of each colo-
nization event and the number of species of the dispersing
lineage in the colonized region (i.e., we treated multiple
invasions of the same region as separate data points). This
analysis tested whether lineages that have been in a region
for longer amounts of time will have more species in that
region, as would be expected if there is a time-for-
speciation effect in emydid turtles. We then examined the
relationship between the species richness of each region
and the relative age of the first emydid lineage to colonize
that region. This analysis tested whether the time-for-
speciation effect explains the relative species richness of
each continent-scale region.
Dispersal between regions was localized on specific
branches of the phylogeny using parsimony reconstruction
(see below), and the relative timing of these events was
estimated based on the depth of those branches in the tree
(i.e., summing branch lengths). Our approach assumes
that the relative timing of a colonization event approxi-
mates the relative age (divergence time) of the branch on
which the colonization event is inferred (i.e., the branch
on which a change between continental regions is recon-
structed). Relative divergence times were estimated in-
dependently based on morphological and molecular
branch lengths.
Molecular branch lengths were estimated using maxi-
mum likelihood analyses of mitochondrial cytochrome b
sequences, the molecular data set that was the most thor-
oughly sampled for emydid taxa (Lamb et al. 1994; Lenk
et al. 1999; Feldman and Parham 2002). Likelihood pa-
rameters were estimated using the best-fitting model of
sequence evolution (GTR ⫹G; Rodriguez et al. 1990),
using version 4.0b8a (Swofford 2002). MODEL-
∗
PAU P
TEST version 3.06 (Posada and Crandall 1998) was used
to select the model of sequence evolution that best fit the
cytochrome bdata via a hierarchical likelihood ratio test.
Note that the topology used to estimate branch lengths
was based on all the molecular and morphological data,
not on a separate analysis of the cytochrome bdata alone.
Estimating divergence times directly from branch lengths
assumes a constant rate of evolution between lineages over
time (i.e., clocklike behavior; Hillis et al. 1996). To test
this assumption with cytochrome bsequences, a likeli-
hood-ratio test was performed comparing the likelihood
of our best-fitting model to that of the same model with
a molecular clock enforced (Goldman 1993; Huelsenbeck
and Crandall 1997). This test rejected strict rate constancy
between lineages (likelihood-ratio test statistic ,p33.3
, ). We therefore used Sanderson’s non-df p31 P1.20
parametric rate smoothing method (NPRS; Sanderson
1997), which is robust to violations of rate constancy, to
determine relative divergence times. The NPRS method
was performed using TreeEdit version 1.0a8 (Rambaut and
Charleston 2001).
116 The American Naturalist
No sequence data were available for Neotropical Trach-
emys. However, based on the topology of the combined-
data tree, invasions of Mexico and the West Indies by
Trachemys seemingly occurred after the origin of the North
American clade of Trachemys scripta and before the origin
of the Graptemys ⫹Malaclemys clade (P. R. Stephens and
J. J. Wiens, unpublished manuscript). These “bracketing”
divergences were used to obtain molecular estimates of
the minimum and maximum ages for the colonization of
Mexico and the West Indies by Trachemys, and analyses
were repeated using both estimates for the age of each
invasion. Results were similar using both estimates, and
only those using maximum ages are reported.
Due to the lack of molecular data for some taxa, the
relative ages of several colonization events could not be
estimated from molecular data. We therefore repeated
these analyses using branch lengths based on parsimony
analysis of the external morphological data (109 charac-
ters, available for all but two of 64 taxa) using ACCTRAN
character optimization (DELTRAN gave similar results in
other analyses; see “Results”). The NPRS method, which
can be applied to any set of nonzero branch lengths, was
used to estimate relative divergence times from morpho-
logical branch lengths.
To account for ambiguities in the species-level taxon-
omy of emydids, the preceding analyses were repeated us-
ing both minimum and maximum estimates of emydid
species richness in each region. The minimum estimate
used only currently recognized species whereas the max-
imum estimate counted geographically isolated subspecies
as distinct species. Regression analyses were performed
using Statview version 4.51 (Roth et al. 1995). Relative age
and species richness were significantly correlated (P!
) in all analyses, with one exception. An analysis that.05
compared the minimum number of emydid species in each
region to the age of the oldest emydid lineage in each
region based on molecular data, and using the minimum
age estimate for West Indian Trachemys, did not show a
significant correlation ( , ). However,
2
rp0.711 Pp.073
there is abundant evidence that many currently recognized
emydid species are polytypic (Legler 1990; Lenk et al. 1999;
P. R. Stephens and J. J. Wiens, unpublished manuscript),
and all results using these higher estimates of regional
species richness were significant. Only results using the
maximum number of species and estimated lineage age in
each region are reported. All analyses of age and species
richness were repeated using natural log transformed es-
timates of species richness, because many models of di-
versification predict an exponential relationship between
lineage age and species richness (Sanderson and Donoghue
1996; Schluter 2000). Results were similar to those using
untransformed estimates of species richness and are not
reported.
Colonization events were inferred by coding the regional
distribution of each species as a character state and re-
constructing changes on the phylogeny with parsimony.
We used the following character states: (0) eastern North
America, (1) western North America, (2) Middle America,
(3) South America, (4) the West Indies, and (5) Europe.
These regions are of sufficient size that each contains a
largely unique emydid fauna. Species distributions were
first mapped onto the tree as unordered characters and
coding polymorphisms in distribution (i.e., a single taxa
that occurred in more than one region) using the “poly-
morphic” method (sensu Wiens 1995, 1999). Thus, species
found in two regions were treated as occurring in only
one during reconstructions (with the state chosen to max-
imize global parsimony). Although this coding method is
widely used, it is potentially problematic for ancestral state
reconstructions (Wiens 1999), and so an analysis was also
performed in which polymorphisms were coded with step
matrices using the scaled method (see Mabee and Hum-
phries 1993 for a detailed description and Wiens 1999 for
justification of the use of this method in ancestral state
reconstructions). A third analysis was performed in which
polymorphisms were coded using step matrices and in
which transitions between states were also weighted based
on the simplified proximity of the six areas (e.g., a change
from Middle America to South America is easier than a
change from South America to North America). Analyses
were performed with MacClade version 3.04 (Maddison
and Maddison 1992).
Relationship between Regional and Local Species Richness
Based on regional trends in emydid species richness (i.e.,
higher species richness in eastern North America relative
to other regions and higher species richness in the south-
east than in the northeast), we expected local species rich-
ness to be positively correlated with latitude in the Western
Hemisphere and negatively correlated with latitude in east-
ern North America. We tested these hypotheses by re-
gressing community species richness against community
latitude in two separate analyses. One analysis used com-
munities from throughout the range of the Emydidae in
the Western Hemisphere ( ), whereas the other usednp65
only communities from eastern North America ( ).np46
River drainages and islands were used as units to estimate
local species richness (fig. 2). For each drainage or island,
local species composition was tabulated where the maxi-
mum number of species overlapped, based on the geo-
graphic references listed above. For most of the range of
the Emydidae, individual drainages and islands were cho-
sen to represent unique assemblages of species (areas of
geographic-range overlap between species). In cases where
a particular assemblage covered a wide geographic area, a
Species Richness in Emydid Turtles 117
Figure 2: Drainages, islands, and other areas used to estimate the maximum number of species in local communities. Numbers indicate the following
areas (all names indicate river drainages unless stated otherwise): central and eastern North America: (1) Nova Scotia between Bay of Fundy and
Northumberland Strait; (2) Sakonnet; (3) Oswego and other drainages connected to southeast Lake Ontario; (4) Delaware; (5) Susquehanna; (6)
Roanoke; (7) western state of Michigan; (8) Muskegon; (9) central Nebraskan drainages including Platte, Loupe, North Loupe, South Loupe, Cedar,
Elkhorn; (10) upper Missouri; (11) upper Mississippi; (12) upper Arkansas; (13) middle Arkansas (in eastern Kansas and Oklahoma); (14) Osage,
Gasconade, and Meremec; (15) White, Eleven Point, Current, Black; (16) Salt; (17) Illinois; (18) Wabash; (19) Kaskaskia, Big Muddy; (20) Saint
Francis; (21) Miami; (22) Scioto; (23) upper Ohio; (24) Green; (25) Salt (in state of Kentucky); (26) Kentucky; (27) Licking; (28) Big Sandy; (29)
lower Kanawha; (30) upper Kanawha; (31) lower Tennessee; (32) Duck; (33) Cumberland; (34) upper Tennessee; (35) upper Grande; (36) Pecos;
(37) San Antonio, Guadalupe; (38) Colorado; (39) Brazos; (40) Trinity; (41) Neches, Sabine; (42) lower Mississippi; (43) Ouachita; (44) Pearl; (45)
Pascagoula; (46) Alabama, Tombigbee; (47) Mobile Bay; (48) Escambia; (49) Apilachicola; (50) Suwannee; (51) Withlacoochee, Peace, Caloosahatche;
(52) St. John’s; (53) southeast North American Atlantic coastal drainages (from Edisto south to Altamaha). West Indies: (54) island of Cuba; (55)
Great Inagua island; (56) island of Hispaniola; (57) island of Jamaica; (58) island of Puerto Rico (not depicted). Middle America: (59) southern
peninsula of Baja California; (60) regions of Culiaca´n and Sinaloa (to Cabo Corientes); (61) Nazas; (62) basin of Cuatro Cienegas in Coahuila; (63)
San Fernando; (64) southern state of Veracruz; (65) coast of Oaxaca; (66) Yucatan peninsula; (67) Republic of Panama. South America: (68) northern
Republic of Venezuela (not depicted); (69) Uruguay (not depicted). Western North America: (70) Columbia; (71) central state of California; (72)
southeastern Arizona.
drainage was chosen to represent the assemblage from near
the latitudinal center. In much of central eastern North
America, patterns of range overlap were so complex that
picking out all unique assemblages by eye became prob-
lematic. Therefore, within this area, the drainages from
Mayden’s (1988) study of the historical biogeography of
the fish fauna were used. Drainages with identical emydid
species composition and similar vicariant histories were
collapsed into a single unit. A few groups of drainages had
dissimilar vicariant histories but identical species com-
positions. From each of these groups of drainages, only
one (chosen at random) was used in any analysis. In sum-
mary, each data point represented a different combination
of species, and each unique assemblage of emydids in east-
ern North America was represented by a single data point.
The latitudinal midpoints of all drainages and islands were
118 The American Naturalist
determined by reference to a world atlas (Christie et al.
1991).
Drainages and islands represent a scale larger than what
some might consider to be good units of local species
richness. The number of species found in each drainage
or on each island should indicate the maximum number
of species that may occur sympatrically in any community
but may not reflect the minimum (i.e., a given pond, river,
or meadow could have fewer species of emydids than the
number we infer for a given “community” but could not
have more). However, because each drainage or island
(hereafter referred to as a community) was chosen to rep-
resent a unique assemblage, subdividing them would pre-
sumably have produced similar results only with more
nonindependent data points. Raw data on the species com-
position and latitude of these communities are available
from the authors.
In order to more directly investigate the relationship
between local and regional species richness, the number
of species in each community was regressed against the
number of species in the region in which the community
occurs (using the continent-scale regions defined above).
The communities were the same as in the previous analysis,
although three communities were added to represent Eu-
rope and northern Africa ( ). A similar analysis wasnp68
performed using only communities in eastern North
America ( ) and dividing eastern North Americanp46
into southern and northern regions at 37.5⬚north latitude
(the latitudinal midpoint of the range of the Emydidae in
North America). The preceding analyses violated an as-
sumption of regression analysis in that many data points
could be sampled from a given region, and thus the values
of the independent variable (regional species richness)
were not independent for all data points. Sampling mul-
tiple communities from the same region is also problem-
atic because it inflates the number of regional-local data
points (Caley and Schluter 1997), which leads to a Pvalue
that is not sufficiently conservative. We therefore repeated
these analyses using regions as data points and averaging
local species richness across the communities in each re-
gion. We also repeated all analyses using minimum and
maximum estimates of the number of species in each re-
gion as described above. Results were similar using both
estimates, and only those using the maximum number of
species for each region are reported.
Geographic Range Size Variation
To investigate latitudinal trends in geographic range sizes,
range maps of species that occur in eastern North America
were digitized and fitted onto an Alber’s equal area pro-
jection map of the continental United States. The latitu-
dinal midpoint and the area of each species’ geographic
range were then calculated using the public domain NIH
Image program (developed at the National Institutes of
Health and available on the Internet at http://
rsb.info.nih.gov/nih-image/). Because characteristics that
determine geographic range size may be inherited phy-
logenetically (e.g., Jablonski 1987; Price et al. 1997), we
tested the relationship between range size and latitude in
a phylogenetic framework using Felsenstein’s (1985b)in-
dependent contrasts method. Analyses used a pruned ver-
sion of the combined-data tree used elsewhere but in-
cluded only the 30 taxa that occur in eastern North
America.
The use of independent contrasts requires specification
of branch lengths, and this was dealt with in two ways.
First, branch lengths were estimated from the external
morphological data using parsimony (PAUP
∗
4.0b8a).
These data were available for the largest number of relevant
taxa. This analysis assumes that the extent of morpholog-
ical divergence generally reflects the relative amounts of
time between speciation events (gradual model, sensu
Martins and Garland 1991). Because branch lengths es-
timated using parsimony may vary depending on character
optimization, we used both ACCTRAN and DELTRAN
optimizations (Swofford and Maddison 1987) to assess the
sensitivity of the results. Second, all branch lengths were
assumed to be equal and were arbitrarily set at 1 (the
punctuational model of Martins and Garland [1991]).
To verify that independent contrasts were adequately
standardized by their estimated branch lengths, the ab-
solute values of each independent contrast for each node
were regressed on their standard deviations (the square
root of the sum of the branch lengths for that contrast),
following Garland et al. (1992). Most contrasts were ad-
equately standardized ( ), and no further transfor-
P1.100
mation of the data or branch lengths was considered nec-
essary. However, contrasts in geographic range area were
positively correlated with their standard deviation when
using ACCTRAN optimization to estimate branch lengths.
Contrasts were recalculated in this case using squared
branch lengths (following Garland et al. 1992). This trans-
formation adequately standardized the ACCTRAN-based
contrasts in geographic range area ( ,
2
rp0.027 Pp
).
.391
Independent contrasts and standard deviations for each
branch and character were obtained using COMPARE ver-
sion 4.4 (Martins 2001). The relationship between con-
trasts for each pair of variables was examined using least
squares linear regression, which forced the model through
the origin (as recommended by Garland et al. 1992), with
significance levels based on the regression coefficients.
Species Richness in Emydid Turtles 119
Mode of Speciation
We used Lynch’s (1989) method to determine which geo-
graphic mode of speciation predominates in emydids. Us-
ing this method, sister taxa generated by allopatric spe-
ciation are assumed to have little overlap in their
geographic ranges, whereas sister taxa generated by sym-
patric speciation are expected to have extensive overlap.
Range maps for sister species (based on P. R. Stephens and
J. J. Wiens, unpublished manuscript) were superimposed
graphically, and the area of overlap in geographic ranges
was calculated using NIH Image. Because dispersal is ex-
pected to obscure patterns of distribution indicative of
different modes of speciation given enough time (Lynch
1989; Chesser and Zink 1994; Barraclough and Vogler
2000), only the most recent speciation events were con-
sidered (i.e., only sister species). This restriction is a con-
servative modification of Lynch’s method and is intended
to minimize the effects of dispersal. All subspecies of
Trachemys scripta endemic to Middle America were treated
as separate species in this analysis because of evidence that
at least some are distinct species (Smith and Smith 1979;
Ernst and Barbour 1989). Areas of overlap were trans-
formed into percentage overlap by dividing the area of
overlap in geographic ranges by the geographic range area
of the species having the smaller range (Lynch 1989). Ac-
cording to Lynch’s (1989) criterion, species with less than
10% overlap are “obvious” cases of allopatric speciation,
and we used this as our cutoff. Although this cutoff is
controversial (Chesser and Zink 1994), our results are un-
ambiguous (i.e., no overlap in almost all cases; see
“Results”).
Results
Mapping regional distributions onto the phylogeny sug-
gests that species richness is lower in regions peripheral
to eastern North America because emydids have not been
present in these regions for as long. Regression of the
relative age of each colonizing lineage against the number
of species of that lineage in the colonized region showed
a positive correlation using both molecular (fig. 3A;
, ) and morphological (fig. 3B;
2 2
rp0.665 Pp.014 rp
, ) data. Regression analysis also showed a0.599 Pp.002
correlation between the age of the oldest emydid lineage
in each region and the overall number of emydid species
in each region regardless of whether relative age was es-
timated based on molecular (fig. 3C;,
2
rp0.988 Pp
) or morphological (fig. 3D;,)
2
.001 rp0.895 Pp.004
branch lengths.
Communities in eastern North America showed no re-
lationship between local species richness and latitude (fig.
4A; , ). However, in the Western Hem-
2
rp0.005 Pp.635
isphere overall, there is a strong positive relationship be-
tween latitude and species richness (fig. 4B;,
2
rp0.232
), with species richness of local communities de-P!.001
creasing from temperate to tropical regions (mirroring re-
gional patterns of species richness).
Among continental regions, local species richness is cor-
related with regional species richness ( ,
2
rp0.573 P!
). Average local species richness of each region was.001
also strongly correlated with regional species richness
( , ). Thus, low local diversity in
2
rp0.909 Pp.003
regions peripheral to eastern North America is seemingly
related to the small size of the regional species pool. How-
ever, local and regional species richness are not correlated
when eastern North America is divided into northern and
southern regions and communities of southeastern North
America are compared to those of the northeast (
2
rp
, ; fig. 1B). Communities in the northeast0.005 Pp.603
have local species richness similar to communities in the
southeast despite higher regional species richness in the
southeast ( for unpaired two-sample t-test com-Pp.808
paring local species richness in the southeast to the
northeast).
Within eastern North America, range size is positively
correlated with latitude (fig. 5A;,)
2
rp0.200 Pp.013
such that species in the northeast tend to have larger geo-
graphic ranges. Phylogenetically independent contrasts in
geographic range size and latitude show the same trend
(fig. 5B), using both the gradual model (ACCTRAN:
, ; DELTRAN: , )
22
rp0.205 Pp.012 rp0.194 Pp.015
and the punctuational model ( , ) to
2
rp0.164 Pp.026
estimate branch lengths.
Allopatric speciation has predominated in the Emydi-
dae, at least recently. Nine of 10 pairs of sister taxa show
no overlap in their geographic ranges (table 1). The only
sister species with any overlap in their geographic ranges
are Clemmys insculpta and Clemmys muhlenbergii, with
60.7% overlap. Regardless of whether this is considered
sufficient evidence for sympatric speciation or not, allo-
patric speciation accounts for at least 90% of recent spe-
ciation events in the Emydidae.
Discussion
In this article, we argue that the time-for-speciation effect
explains two major patterns of species richness in emydid
turtles: (1) the high regional and local species richness of
eastern North America compared to other continental
regions and (2) the greater regional species richness of
southeastern North America compared to northeastern
North America (despite the lack of differences in local
species richness). We suggest that the time-for-speciation
effect has the potential to explain patterns of richness in
a variety of systems. Our results also have implications for
120 The American Naturalist
Figure 3: The time-for-speciation effect explains continent-scale patterns of species richness in emydid turtles. The top two graphs (A,B) show
regressions of the timing of each colonization event versus the number of species in the dispersing lineage in the colonized region, with relative
ages of lineages estimated from cytochrome bsequences using maximum likelihood (A) or from morphological data using parsimony (B). The
bottom graphs (C,D) show regressions of emydid species richness in a region versus the relative age of emydids in that region (based on the first
emydid lineage to colonize the region), with relative age estimated from cytochrome bsequences using maximum likelihood (C) or from morphological
data using parsimony (D). Units of relative age are based on summed branch lengths (as described in the text) using either substitutions per site
(#100) for Aand Cor estimated evolutionary change (number and weight of character state changes) for Band D.
the relationship between geographic range size and dif-
ferent measures of diversity as well as for studies of “sat-
uration” in communities.
Continent-Level Patterns of Emydid Species Richness
The time-for-speciation effect seems to provide the sim-
plest explanation for patterns of emydid species richness
at the regional, continent-level scale. No single environ-
mental factor seems to explain these patterns, especially
given the extreme environmental heterogeneity of emydid-
poor regions (e.g., arid, mesic, tropical, temperate) and
the ecological similarity of some of these regions (e.g.,
temperate Europe, northwestern North America) to east-
ern North America, where emydids are most speciose. The
possibility that there was limited open “niche space” in
regions of low emydid diversity also seems unlikely. If
emydid species richness had been limited by competition,
we would expect the number of emydid species to be
inversely correlated with the number of non-emydid turtle
species in each continental region. Regression analysis of
the number of emydid versus non-emydid turtle species
in each continental region confirms that there is no such
pattern ( , ; non-emydid distribution
2
rp0.063 Pp.631
Species Richness in Emydid Turtles 121
Figure 4: Regression of community latitude versus community species richness in (A) eastern North America and (B) the Western Hemisphere
data from Iverson 1992). Most regions outside of eastern
North America contain few non-emydid turtle species:
there is only one each in western North America and the
West Indies and only six in Europe. The number of non-
emydid species is similar in both emydid-poor Middle
America (22 species overlap with emydids) and emydid-
rich eastern North America (18 overlapping species).
Among emydid-poor regions, only South America has
many more non-emydid turtle species (38 species) than
eastern North America.
In contrast to many groups of organisms (Brown and
Lomolino 1998; Gaston and Blackburn 2000), emydids
show lower species richness in tropical regions than tem-
perate regions. This unusual pattern suggests the possi-
bility that some ecological factor makes the tropics a
“harsher” environment for emydids than temperate
regions. However, emydids are among the most abundant
vertebrates (in terms of biomass) in some tropical com-
munities, and some tropical emydids grow to larger sizes
and are more fecund than their close temperate zone rel-
atives (Moll and Legler 1971; Moll and Moll 1990). The
reverse latitudinal gradient in species richness seen in emy-
did turtles, combined with their ability to thrive in tropical
regions, makes a compelling “natural experiment” that
suggests that phylogenetic history (i.e., the time-for-
speciation effect) can be more important than local eco-
logical processes in determining latitudinal patterns of spe-
cies richness. If local ecological conditions were more
important than time, emydids should have higher species
richness in tropical regions despite their recent arrival.
These observations imply that the pattern of high tropical
species richness observed in many organisms might be
caused by those groups originating in tropical regions and
having been present and speciating in tropical regions for
longer periods of time (the museum hypothesis; Gaston
and Blackburn 1996), as was widely believed during the
122 The American Naturalist
Figure 5: Regression of geographic range size and midpoint of geographic
range using (A) raw data and (B) independent contrasts using a gradual
model of character evolution and ACCTRAN optimization of the mor-
phological data.
first half of the twentieth century (Stebbins 1974; Nelson
1978).
Patterns of Species Richness within Eastern North America
We hypothesize that the differences in regional diversity
and average geographic range size between northeastern
and southeastern North America may also reflect the time-
for-speciation effect, in this case acting through local ex-
tinction or emigration of northern emydid populations
caused by glaciation. If speciation is primarily allopatric
(table 1), then we would expect species with large geo-
graphic ranges to be broken up over time into descendant
species with smaller geographic ranges by vicariant spe-
ciation events (even though the range size of descendant
species may eventually increase after speciation). Because
the northern regions of North America were scoured by
glaciers that reached their maximum extent only 20,000
yr ago (Holman 1995; Brown and Lomolino 1998), emy-
dids in northeastern North America must be relatively
recent immigrants to the region, with little time to be split
into new species by vicariance. Conversely, in the south-
east, there has been more time for vicariant speciation
events to simultaneously fragment range sizes and increase
regional species diversity. This hypothesis is also consistent
with the similarity in local species richness between north-
eastern and southeastern North America.
Another possibility is that the disparity in regional spe-
cies richness between southeastern and northeastern North
America reflects greater rates of extinction in the northeast
or higher rates of speciation in certain lineages in the
southeast. Based on fossil evidence, Holman (1995) sug-
gested that most (if not all) of the present-day northeastern
emydid fauna was present before Pleistocene glaciation
(beginning 1.9 million years ago) and that these species
responded to glaciation by shifting their distributions
southward or contracting their ranges to southern refugia
and then reinvading northern regions after glaciers re-
treated. (Note: The presence of emydids in the northeast
before glaciation explains how northeastern North Amer-
ica has high species richness relative to other continent-
scale regions despite being recolonized quite recently.) The
fossil record reveals only one extinct emydid species in all
of North America (Holman 1995). However, even if there
were many northeastern species that went extinct and were
not recorded in the fossil record (which seems unlikely),
the failure of the northeast to catch up to the southeast
in species richness might still reflect the time-for-
speciation effect. Similarly, molecular studies of North
American birds (Zink and Slowinski 1995) suggest that
glaciation slowed the rate of speciation in the Pleistocene
by reducing or displacing habitats rather than causing ex-
tinction. Given that the southeast and northeast may have
been inhabited by emydids for a similar period of time
before glaciation, the higher species richness in the south-
east may be explained largely by southeastern taxa that
speciated during the Pleistocene (when much of the north-
east was periodically rendered uninhabitable for emydids).
Although heightened rates of species extinctions in the
northeast are unlikely, it is possible that greater rates of
speciation related to habitat specificity in some south-
eastern emydids increased species richness in southeastern
North America. The southeastern emydid fauna includes
several river specialists (e.g., Graptemys flavimaculata,
Graptemys gibbonsi,Graptemys ernsti,Graptemys oculifera)
that seem to make overland migrations between drainages
much less frequently than other species of aquatic emydids
(Ernst et al. 1994). Because of their reduced propensity
for terrestrial dispersal, these southeastern Graptemys may
have experienced increased rates of allopatric speciation
Species Richness in Emydid Turtles 123
Table 1: Modes of speciation in emydid turtles as inferred from overlap
in the geographic ranges of sister species
Sister species pair % overlap
Clemmys insculpta ⫹Clemmys muhlenbergii 60.7
Graptemys ernsti ⫹Graptemys gibbonsi 0
Graptemys flavimaculata ⫹Graptemys oculifera 0
Graptemys ouachitensis sabinensis ⫹Graptemys versa 0
Pseudemys nelsoni ⫹Pseudemys rubriventris 0
Terrapene carolina ⫹Terrapene coahuila 0
Terrapene nelsoni ⫹Terrapene ornata 0
Trachemys decorata ⫹Trachemys decussata 0
Trachemys scripta callistrosis ⫹Trachemys scripta venusta 0
Trachemys scripta grayi ⫹Trachemys scripta nebulosa 0
Note: The absence of overlap in the ranges of most sister species indicates a
preponderance of allopatric speciation.
caused by vicariant events that affected riverine habitats
(e.g., separation of previously connected drainages) during
the Pleistocene (Lamb et al. 1994). Low levels of divergence
in cytochrome bsequences also indicate that these Grap-
temys have speciated relatively recently (Lamb et al. 1994).
Thus, high rates of speciation in certain southeastern taxa
(associated with habitat specificity) and limited time for
speciation in the northeast both may have contributed to
the higher species richness in southeastern North America.
The Time-for-Speciation Effect versus Variation in
Rates of Speciation and Extinction
A potential criticism of using the time-for-speciation effect
to explain geographic patterns of species richness is that
this hypothesis assumes that rates of speciation and ex-
tinction are generally similar between regions and lineages.
In fact, these rates can vary extensively over space, time,
and taxa (e.g., Sanderson and Donoghue 1994, 1996;
Schluter 2000). For example, many recent studies have
postulated that certain groups have undergone recent
rapid diversification and thus have greater species richness
than expected given their age (e.g., Baldwin and Sanderson
1998; Albertson et al. 1999; Magallo´ n and Sanderson 2001;
Richardson et al. 2001a, 2001b). However, the fact that
rates of speciation in these groups are considered unusual
suggests that they are deviations from a more general re-
lationship between time and species richness. We suggest
that the time-for-speciation effect may be important in
explaining patterns of species richness generally but not
universally. This study may be the first to use phylogenetic
methods to test whether time alone can explain patterns
of species richness between geographic regions rather than
variation in rates of speciation or extinction (but see Gas-
ton and Blackburn 1996). Our results show that time ex-
plains most of the variation in emydid species richness at
the continental scale (fig. 3).
Range Size and Species Richness
Many authors have discussed the relationship between spe-
cies richness and average geographic range size, specifically,
the idea that smaller range sizes allow more species to be
“packed” into a region (e.g., Stevens 1989; Gaston 1996).
However, few have addressed the effects of geographic
range size on different measures of species richness (sensu
Whittaker 1972); namely, adiversity (local species rich-
ness), bdiversity (turnover in species composition be-
tween communities or habitats), and gdiversity (regional
species richness). Brown and Lomolino (1998) stated that
one implication of Rapoport’s rule is that “alpha and beta
diversity appear to be positively correlated” (Brown and
Lomolino 1998, p. 470) in the Northern Hemisphere due
to smaller geographic ranges causing greater species turn-
over in areas that have higher adiversity (i.e., the tropics).
We have shown that in emydid turtles, Rapoport’s rule is
supported (smaller ranges in the south) and yet aand b
diversity appear to be uncoupled: band gdiversity are
higher in southeastern North America than they are in
northeastern North America, but adiversity is similar
between these regions (figs. 1B,4A). We suggest that this
pattern is caused by the effects of allopatric speciation on
species diversity and average geographic range size.
The effect of allopatric speciation on range size and
different types of diversity can be illustrated with a simple
example (fig. 6). Consider three wide-ranging, closely re-
lated species (A, B, C) that occur sympatrically throughout
the same geographic region. In this region, gdiversity is
3, adiversity in each community is 3, and bdiversity is
0. A vicariant event then divides the region in half so that
three pairs of descendant sister species (A
1
and A
2
,B
1
and
B
2
,C
1
and C
2
) are generated by allopatric speciation. The
average geographic range size of species in the region will
decrease, gdiversity of the region will increase to 6, a
diversity will remain 3 in any community, and species
124 The American Naturalist
Figure 6: Effect of vicariant events and allopatric speciation on patterns of geographic range size variation and a,b, and gdiversity. Before the
vicariant event (region on the left), species turnover between any two communities is zero. After the vicariant event, turnover in species composition
between any two communities in the region is 100% or zero, depending on which communities are sampled.
turnover between any two communities will be either
100% or 0, depending on which two communities are
sampled. Thus, allopatric speciation can increase gand b
diversity without increasing adiversity, and it will tend
to reduce the average geographic range size of species in
a region.
These observations imply a general relationship between
range size and species richness. When average geographic
range size is very large, species will tend to overlap with
more of the other species in the regional species pool,
which makes adiversity a greater proportion of gdiversity
and bdiversity small. Conversely, when average geographic
range size is small, species will tend to overlap with few
other species in the region, leading to high bdiversity and
to gdiversity that is many times adiversity. This rela-
tionship can be summarized with the following equation:
A
t
¯
gp#a,
R
where gequals the number of species in a region, equals
¯
a
the average number of species at any locality in the region,
A
t
equals the area of the entire region, and equals theR
average geographic range area of an organism within the
region (this assumes that portions of the range of any
species that lie outside the region are ignored). In fact,
this equation is a simple derivation of the relationships
that Whittaker (1972, p. 232) originally used to define b
diversity.
Based on these general relationships and our results
from emydid turtles, we suggest that species richness will
accumulate in a region in a predictable series of stages,
each typified by different patterns of regional and local
species richness and geographic range size. The first species
in the group to colonize a region (e.g., after glaciation)
will often spread throughout the region and come to in-
habit a large area. This pattern corresponds to the large
average range sizes of emydids in postglacial northeastern
North America. In time, the ranges of some species will
be reduced by vicariance and allopatric speciation, pro-
ducing a variety of range sizes. This will increase band g
diversity but have no initial effect on adiversity (fig. 6).
This pattern corresponds to southeastern North America,
with high regional species richness and species that exhibit
a variety of range sizes (from single drainages to the entire
region) despite local species richness that is no higher than
in the northeast. Eventually some of these new species will
disperse and invade the ranges of other species, raising
average adiversity and increasing average geographic
range size. This situation might correspond to the pattern
seen in many tropical lineages with high local and regional
diversity. In support of this hypothesis, Hecnar (1999)
analyzed latitudinal patterns of geographic range size
across almost all turtles and continents and found that
large geographic range sizes in temperate regions may be
a localized effect of recent glaciation, whereas large geo-
graphic ranges are a general pattern in tropical regions.
Thus, according to this scenario, latitudinal patterns of
geographic range size distributions and species richness
may reflect a temporal progression of events that has gone
further in tropical regions than temperate ones.
Local and Regional Species Richness and
Saturation of Communities
Many studies have tried to determine whether the species
richness of communities tends to be limited more by the
size of the regional species pool or by the niche space
available in communities (e.g., Terborgh and Faaborgh
Species Richness in Emydid Turtles 125
1980; Cornell 1985; Tonn et al. 1990; Cornell and Lawton
1992; Hawkins and Compton 1992; Caley and Schluter
1997; Shurin et al. 2000). A common approach to this
problem is to examine the relationship between local and
regional species richness for a set of communities (re-
viewed in Cornell and Lawton 1992; Cornell 1993). In
some cases, a plot of regional versus local species richness
bears an asymptote such that local species richness no
longer increases, regardless of regional species richness
(i.e., “Type II” communities; Cornell and Lawton 1992).
Communities along such an asymptote are said to be “sat-
urated” with species, and the number of species in these
communities are assumed to be limited by competition
(Terborgh and Faaborg 1980; Cornell and Lawton 1992;
Aho and Bush 1993; Caley and Schluter 1997). In this
study, we have shown that regional species richness in-
creases from northeastern to southeastern North America
but that local species richness does not. This pattern could
be interpreted as evidence that communities in south-
eastern North America are saturated with emydids. How-
ever, we have described how allopatric speciation can in-
crease regional species richness without affecting local
species richness (fig. 6). This mechanism would produce
a pattern resembling saturation even in the absence of
competition. This mechanism seems to explain patterns
of regional and local species richness in emydids in eastern
North America and may explain patterns resembling sat-
uration in other systems as well.
Local Processes and the Time-for-Speciation Effect
At the largest spatial scale, patterns of emydid species rich-
ness seem to have been determined directly by the time-
for-speciation effect, with species richness in various
regions being tightly correlated with how long emydids
have been present in each region (fig. 3). At a smaller
scale, within eastern North America, the impact of the
time-for-speciation effect is more indirect and is seemingly
mediated through temporary local extinction caused by
glaciation. We suggest that the time-for-speciation effect
may often influence patterns of species richness at smaller
spatial scales by acting through a variety of local ecological
processes, such as extinction, competition, adaptation, and
predation. For example, Brown et al. (2000) and McPeek
and Brown (2000) recently used a phylogenetic approach
to explore patterns of species richness in lakes containing
damselfly larvae (genus Enallagma). They found that lakes
where dragonflies are the top predator have much lower
diversity than lakes where fish are the top predator. Map-
ping habitat type onto the phylogeny suggests that the
dragonfly lake habitat has only recently been invaded by
damselflies. Thus, the low diversity in dragonfly lakes can
be attributed to the limited time for speciation within this
habitat (although McPeek and Brown [2000] allowed that
unusually high extinction rates in dragonfly lakes was also
possible). It is easy to imagine similar scenarios in which
the time-for-speciation effect is mediated through other
local-scale processes.
Acknowledgments
For comments on the manuscript we thank J. Chase, R.
Espinoza, S. Kalisz, B. Livezey, Z. Long, M. McPeek, R.
Relyea, S. Schnitzer, and two anonymous reviewers. We
acknowledge the National Science Foundation (DEB
0129142 to J.J.W.) for financial support in the latter stages
of the project.
Literature Cited
Aho, J. M., and A. O. Bush. 1993. Community richness
in parasites of some freshwater fishes from North Amer-
ica. Pages 185–193 in R. E. Ricklefs and D. Schluter,
eds. Species diversity in ecological communities: his-
torical and geographic perspectives. University of Chi-
cago Press, Chicago.
Albertson, R. C., J. A. Market, P. D. Danley, and T. D.
Kocher. 1999. Phylogeny of a rapidly evolving clade: the
cichlid fishes of Lake Malawi, East Africa. Proceedings
of the National Academy of Sciences of the USA 96:
5107–5110.
Andersen, N. M. 1999. The evolution of marine insects:
phylogenetic, ecological and geographical aspects of spe-
cies diversity in marine water striders. Ecography 22:
98–111.
Avise, J. C. 2000. Phylogeography: the history and for-
mation of species. Harvard University Press, Cambridge,
Mass.
Axelrod, D. I. 1952. A theory of angiosperm evolution.
Evolution 6:29–59.
Baldwin, B. G., and M. J. Sanderson. 1998. Age and rate
of diversification of the Hawaiian silversword alliance.
Proceedings of the National Academy of Sciences of the
USA 95:9402–9406.
Barraclough, T. G., and A. P. Vogler. 2000. Detecting the
geographical pattern of speciation from species-level
phylogenies. American Naturalist 155:419–434.
Brown, J. H. 1995. Macroecology. University of Chicago
Press, Chicago.
Brown, J. H., and M. V. Lomolino. 1998. Biogeography.
2d ed. Sinauer, Sunderland, Mass.
Brown, J. M., M. A. McPeek, and M. L. May. 2000. A
phylogenetic perspective on habitat shifts and diversity
in the North American Enallagma damselflies. System-
atic Biology 49:697–712.
Bury, R. 1979. Population ecology of freshwater turtles.
126 The American Naturalist
Pages 571–602 in M. Harless and H. Morlock, eds. Tur-
tles: perspectives and research. Wiley, New York.
Bush, A. O., J. M. Aho, and C. R. Kennedy. 1990. Eco-
logical versus phylogenetic determinants of helminth
parasite community richness. Evolutionary Ecology 4:
1–20.
Caley, M. J., and D. Schluter. 1997. The relationship be-
tween local and regional diversity. Ecology 78:70–80.
Chesser, R. T., and R. M. Zink. 1994. Modes of speciation
in birds: a test of Lynch’s method. Evolution 48:
490–497.
Christie, J., A. Ewington, H. Lewis, P. Middleton, and B.
Wilkelman, eds. 1991. The New York Times atlas of the
world, in collaboration with the Times of London.
Times Books, New York.
Colwell, R. K., and D. C. Lees. 2000. The mid-domain
effect: geometric constraints on the geography of species
richness. Trends in Ecology & Evolution 15:71–76.
Cornell, H. V. 1985. Species assemblages of cynipid gall
wasps are not saturated. American Naturalist 126:
565–569.
———. 1993. Unsaturated patterns in species assem-
blages: the role of regional processes in setting local
species richness. Pages 243–252 in R. E. Ricklefs and D.
Schluter, eds. Species diversity in ecological commu-
nities: historical and geographic perspectives. University
of Chicago Press, Chicago.
Cornell, H. V., and J. H. Lawton. 1992. Species interac-
tions, local and regional processes, and limits to richness
of ecological communities: a theoretical perspective.
Journal of Animal Ecology 61:1–12.
Cronquist, A. 1968. The evolution and classification of
flowering plants. Houghton Mifflin, Boston.
Darlington, P. J. 1957. Zoogeography: the geographical dis-
tribution of animals. Wiley, New York.
Ernst, C. H., and R. W. Barbour. 1989. Turtles of the world.
Smithsonian Institution, Washington, D.C.
Ernst, C. H., J. E. Lovich, and R. W. Barbour. 1994. Turtles
of the United States and Canada. Smithsonian Insti-
tution, Washington, D.C.
Feldman, C. R., and J. F. Parham. 2002. Molecular phy-
logenetics of emydine turtles: taxonomic revision and
the evolution of shell kinesis. Molecular Phylogenetics
and Evolution 22:388–398.
Felsenstein, J. 1985a. Confidence limits on phylogenies:
an approach using the bootstrap. Evolution 39:783–791.
———. 1985b. Phylogenies and the comparative method.
American Naturalist 125:1–15.
Fischer, A. G. 1960. Latitudinal variation in organismal
diversity. Evolution 14:64–81.
Garland, T., Jr., P. H. Harvey, and A. R. Ives. 1992. Pro-
cedures for the analysis of comparative data using phy-
logenetically independent contrasts. Systematic Biology
41:18–32.
Gaston, K. J. 1996. Species-range-size distributions: pat-
terns, mechanisms and implications. Trends in Ecology
& Evolution 11:197–201.
Gaston, K. J., and T. M. Blackburn. 1996. The tropics as
a museum of biological diversity: an analysis of the New
World avifauna. Proceedings of the Royal Society of
London B, Biological Sciences 263:63–68.
———. 2000. Pattern and process in macroecology. Black-
well Science, Malden, Mass.
Gaston, K. J., T. M. Blackburn, and J. I. Spicer. 1998.
Rapoport’s rule: time for an epitaph? Trends in Ecology
& Evolution 13:70–74.
Goldman, N. 1993. Statistical tests of models of DNA sub-
stitution. Journal of Molecular Evolution 36:182–198.
Hawkins, B. A., and S. G. Compton. 1992. African fig wasp
communities: undersaturation and latitudinal gradients
in species richness. Journal of Animal Ecology 61:
361–372.
Hecnar, S. J. 1999. Patterns of turtle species’ geographic
range size and a test of Rapoport’s rule. Ecography 22:
436–446.
Hillis, D. M., B. K. Mable, and C. Moritz. 1996. Appli-
cations of molecular systematics: the state of the field
and a look to the future. Pages 515–543 in D. M. Hillis,
C. Moritz, and B. K. Mable, eds. Molecular systematics.
2d ed. Sinauer, Sunderland, Mass.
Holman, J. A. 1995. Pleistocene amphibians and reptiles
in North America. Oxford University Press, New York.
Huelsenbeck, J. P., and K. A. Crandall. 1997. Phylogeny
estimation and hypothesis testing using maximum like-
lihood. Annual Review of Ecology and Systematics 28:
437–466.
Iverson, J. B. 1992. A revised checklist with distribution
maps of the turtles of the world. Green Nature Books,
Homestead, Fla.
Jablonski, D. 1987. Heritability at the species level: analysis
of the geographic ranges of Cretaceous mollusks. Sci-
ence (Washington, D.C.) 238:360–363.
Lamb, T., C. Lydeard, R. Walker, and J. W. Gibbons. 1994.
Molecular systematics of map turtles (Graptemys): a
comparison of mitochondrial restriction site versus se-
quence data. Systematic Biology 43:543–559.
Legler, J. 1990. The genus Pseudemys in Mesoamerica: tax-
onomy, distribution, and origins. Pages 82–105 in J. W.
Gibbons, ed. Life history and ecology of the slider turtle.
Smithsonian Institution, Washington, D.C.
Lenk, P., U. Fritz, U. Joger, and M. Winks. 1999. Mito-
chondrial phylogeography of the European pond turtle,
Emys orbicularis (Linnaeus 1758). Molecular Ecology 8:
1911–1922.
Letcher, A. J., and P. H. Harvey. 1994. Variation in geo-
Species Richness in Emydid Turtles 127
graphical range size among mammals of the Palearctic.
American Naturalist 144:30–42.
Losos, J. B., and D. Schluter. 2000. Analysis of an evolu-
tionary species-area relationship. Nature 408:847–850.
Lovette, I. J., and E. Bermingham. 2000. Explosive spe-
ciation in the New World Dendroica warblers. Proceed-
ings of the Royal Society of London B, Biological Sci-
ences 266:1629–1636.
Lovich, J. E., and C. J. McCoy. 1992. Review of the Grap-
temys pulchra group (Reptilia: Testudines: Emydidae),
with descriptions of two new species. Annals of Carnegie
Museum 61:293–315.
Lynch, J. D. 1989. The gauge of speciation: on the fre-
quencies of modes of speciation. Pages 527–553 in D.
Otte and J. A. Endler, eds. Speciation and its conse-
quences. Sinauer, Sunderland, Mass.
Mabee, P. M., and J. Humphries. 1993. Coding polymor-
phic data: examples from allozymes and ontogeny. Sys-
tematic Biology 42:166–181.
Maddison, W. P., and D. R. Maddison. 1992. MacClade
version 3.04: analysis of phylogeny and character evo-
lution. Sinauer, Sunderland, Mass.
Magallo´n, S., and M. J. Sanderson. 2001. Absolute diver-
sification rates in angiosperm clades. Evolution 55:
1762–1780.
Martins, E. P. 2001. COMPARE version 4.4: computer pro-
grams for the statistical analysis of comparative data.
http://compare.bio.indiana.edu/. Department of Biol-
ogy, Indiana University, Bloomington.
Martins, E. P., and T. Garland, Jr. 1991. Phylogenetic anal-
yses of the correlated evolution of continuous charac-
ters: a simulation study. Evolution 45:534–557.
Mayden, R. L. 1988. Biogeography, parsimony, and evo-
lution in North American freshwater fishes. Systematic
Zoology 37:329–355.
McPeek, M. A., and J. M. Brown. 2000. Building a regional
species pool: diversification of the Enallagma damselflies
in eastern North American waters. Ecology 81:904–920.
Moll, D., and E. Moll. 1990. The slider turtle in the neo-
tropics: adaptation of a temperate species to a tropical
environment. Pages 152–168 in J. W. Gibbons, ed. Life
history and ecology of the slider turtle. Smithsonian
Institution, Washington, D.C.
Moll, E., and J. M. Legler. 1971. The life history of a
neotropical slider turtle, Pseudemys scripta (Schoepff),
in Panama. Bulletin of the Los Angeles County Museum
of Natural History Science 11.
Morin, P. J. 1999. Community ecology. Blackwell Science,
Malden, Mass.
Nelson, G. 1978. From Candolle to Croizat: comments on
the history of biogeography. Journal of Historical Bi-
ology 11:269–305.
Pianka, E. R. 1966. Latitudinal gradients in species diver-
sity: a review of concepts. American Naturalist 100:
33–46.
———. 2000. Evolutionary ecology. 6th ed. Addison-
Wesley, San Francisco.
Posada, D., and K. A. Crandall. 1998. MODELTEST: test-
ing the model of DNA substitution. Bioinformatics 14:
817–818.
Price, T. D., A. J. Helbig, and A. D. Richman. 1997. Evo-
lution of breeding distributions in the Old World leaf
warblers (genus Phylloscopus). Evolution 51:552–556.
Rambaut, A., and M. Charleston. 2001. TreeEdit,
phylogenetic tree editor. Version 1.0a8. http://
evolve.zoo.ox.ac.uk/software/TreeEdit/TreeEdit.html.
Richardson, J. E., F. M. Weltz, M. F. Fay, Q. C. B. Cronk,
H. P. Linder, G. Reeves, and M. W. Chase. 2001a. Rapid
and recent origin of species richness in the Cape flora
of South Africa. Nature 412:181–183.
Richardson, J. E., R. T. Pennington, T. D. Pennington, and
P. M. Hollingsworth. 2001b. Rapid diversification of a
species-rich genus of Neotropical rain forest trees. Sci-
ence (Washington, D.C.) 293:2242–2245.
Rodrı´guez, F., J. L. Oliver, A. Marı´n, and J. R. Medina.
1990. The general stochastic model of nucleotide sub-
stitution. Journal of Theoretical Biology 142:485–501.
Rohde, K. 1986. Differences in species diversity of Mono-
genea between the Pacific and Atlantic Oceans. Hydro-
biologia 137:21–28.
———. 1992. Latitudinal diversity gradients in species
diversity: the search for the primary cause. Oikos 65:
514–527.
Rosenzweig, M. L. 1995. Species diversity in space and
time. Cambridge University Press, Cambridge.
Roth, J., K. Haycock, J. Gagnon, C. Soper, and J. Caldarola.
1995. Statview. Version 4.51. Abacus Concepts, Berkeley.
Sanderson, M. J. 1997. A nonparametric approach to es-
timating divergence times in the absence of rate con-
stancy. Molecular Biology and Evolution 14:1218–1231.
Sanderson, M. J., and M. J. Donoghue. 1994. Shifts in
diversification rate with the origin of angiosperms. Sci-
ence (Washington, D.C.) 264:1590–1593.
———. 1996. Reconstructing shifts in diversification rates
on phylogenetic trees. Trends in Ecology & Evolution
11:15–20.
Schluter, D. 2000. The ecology of adaptive radiation. Ox-
ford University Press, New York.
Seidel, M. E. 1988. Revision of the West Indian emydid
turtles (Testudines). American Museum Novitiates
2918:1–41.
———. 1994. Morphometric analysis of cooter and red-
bellied turtles in the North American genus Pseudemys
(Emydidae). Chelonian Conservation and Biology 1:
117–130.
Shurin, J. B., J. E. Havel, M. A. Leibold, and B. Pinel-
128 The American Naturalist
Alloul. 2000. Local and regional zooplankton species
richness: a scale-independent test for saturation. Ecol-
ogy 81:3062–3073.
Smith, H. M., and R. B. Smith. 1979. Synopsis of the
herpetofauna of Mexico. Vol. 6. Guide to Mexican tur-
tles. Bibliographic addendum 3. John Johnson, North
Bennington, Vt.
Stebbins, G. L. 1974. Flowering plants: evolution above
the species level. Harvard University Press, Cambridge,
Mass.
Stevens, G. C. 1989. The latitudinal gradient in geograph-
ical range: how so many species coexist in the tropics.
American Naturalist 133:240–256.
Stone, L., T. Dayan, and D. Simberloff. 1996. Community-
wide assembly patterns unmasked: the importance of
species’ differing geographical ranges. American Natu-
ralist 148:997–1015.
Swofford, D. L. 2002. : phylogenetic analysis using
∗
PAU P
parsimony ( and other methods). Version 4.0b8a. Sin-
∗
auer, Sunderland, Mass.
Swofford, D. L., and W. P. Maddison. 1987. Reconstructing
ancestral character states under Wagner parsimony.
Mathematical Biosciences 87:199–229.
Terborgh, J. W., and J. Faaborg. 1980. Saturation of bird
communities in the West Indies. American Naturalist
116:178–195.
Tonn, W. M., J. J. Magnuson, M. Rask, and J. Toivonen.
1990. Intercontinental comparison of small-lake fish as-
semblages: the balance between local and regional pro-
cesses. American Naturalist 136:345–375.
Vanzolini, P. E. 1995. A new species of turtle, genus Trach-
emys, from the state of Maranha˜o, Brazil (Testudines,
Emydidae). Revista Brasileira de Biologia 55:111–125.
Ward, J. P. 1984. Relationships of chrysemyd turtles of
North America (Testudines: Emydidae). Special Publi-
cations of the Museum Texas Tech University 21:1–50.
Whittaker, R. H. 1972. Evolution and measurement of
species diversity. Taxon 21:213–251.
Wiens, J. J. 1995. Polymorphic characters in phylogenetic
systematics. Systematic Biology 44:482–500.
———. 1999. Polymorphism in systematics and compar-
ative biology. Annual Review of Ecology and Systematics
30:327–362.
Willig, M. R., and S. K. Lyons. 1998. An analytical model
of latitudinal gradients of species richness with an em-
pirical test for marsupials and bats in the New World.
Oikos 81:93–98.
Willis, J. C. 1922. Age and area. Cambridge University
Press, Cambridge.
Zink, R. M., and J. B. Slowinski. 1995. Evidence from
molecular systematics for decreased avian diversification
in the Pleistocene epoch. Proceedings of the National
Academy of Sciences of the USA 92:5832–5835.
Associate Editor: Jonathan B. Losos