Information on resource partitioning in multispecies communities is essential to understanding the dynamics of competitive interactions between species and their coexistence in stable communities (Schoener 1974). Among reptiles, much has been learned about resource partitioning in lizard communities (reviewed in Schoener 1983), but studies on other groups have lagged behind. Resource partitioning studies describe the patterns that occur in interacting communities of animals and help to understand factors that cause these partitioning patterns (Schoener 1977). Bury (1979) noted that there were few studies on turtle communities and 4 decades later the situation is not much better, although partitioning in various parts of the world has been reviewed. The structure and functioning of freshwater turtle communities remain understudied topics, even though the basics of population ecology are known. More studies are needed, especially from tropical regions, where turtle species richness and diversity are high (Luiselli 2008; Luiselli et al. 2011). The availability of food items in a habitat is the principal limiting factor for a forager (Perry and Pianka 1997), and a greater level of understanding about the degree of variation in diets of natural populations increases understanding about food chains (Goodyear and Pianka 2011).

Previous studies on food resource partitioning in freshwater turtles have mostly been limited to comparative studies of a few congeneric species (Mahmoud 1968; Berry 1975; Vogt 1981; Williams and Christiansen 1981). Vogt and Guzman-Guzman (1988) and Moll (1990) analyzed resource partitioning in Neotropical freshwater turtle communities, albeit small communities consisting of 3 and 4 species, respectively. More recently, Fachín-Terán et al. (1995) compared a community of 5 pleurodire species and Cunha et al. (2020) compared an assemblage of 4 Podocnemis species. These studies confirmed that food resource partitioning occurs in turtle communities, that the limits of food niches may vary temporally and geographically, and that the niche width is particularly sensitive to interspecies interactions and population structure (Lindeman 2000). Food partitioning does not imply competition; in order to prove competition, sampling of the entire food supply availability would have to be conducted in the habitat of the turtles at the same time the stomach contents are collected. No one has ever done this, for any turtle assemblage. What is usually done, as in this study, is that turtles living in the same habitat are collected and their dietary samples are compared. Because the same resources are available to all species, they can be compared without inferences being made about competition. It would be impossible to sample all of the food resources available for turtles to eat in a large river system, so that is why we call it food partitioning, i.e., how the turtle species divide the available resources. If some resources have a high caloric value and are in limited supply, some species may or may not choose to eat these food items over others. Most turtle species are generalized omnivores (Lagler 1943; Vogt and Villarreal-Benitez 1997; Legler and Vogt 2013) and utilize the most abundant, easily obtained food in their habitats. Seasonality of available food resources forces turtles to change their preferences both quantitatively and qualitatively throughout the year (Mahmoud 1968; Dreslik 1999; Alcalde et al. 2010). Welsh et al. (2017) completed an elegant study in the tropical Daly River in the Northern Territory, Australia. The stomach contents and microhabitat of 5 species of turtles were compared. They captured the turtles by hand and flushed their stomachs soon after, finding that Carrettochelys insculpta was an omnivore feeding primarily on eel grass and snails. Elseya dentata overlapped somewhat in stomach contents, feeding on algae, but lived in a different microhabitat. Emydura victoriae fed almost exclusively on mollusks and overlapped little with the other 4 species. Emydura subglobosa was an omnivore feeding on sponges. Chelodina oblonga had almost no stomach contents, which the authors attributed to their use of the river only as a refuge during the low-water season. Cunha et al. (2020) found little variation in the plants eaten by 4 sympatric Podocnemis species in the Brazilian Amazon, perhaps because the plants eaten were all abundant species and readily available throughout the year, such that there was no reason to expect food partitioning of these resources.

Variation in food items consumed among turtle populations from different habitats often reflects the available resource options in these environments (Lagler 1943; Fachín-Terán et al. 1995; Pérez-Santigosa et al. 2011). Body size also influences turtle feeding behavior (Mahmoud 1968; Bury 1986; Vogt and Villarreal-Benitez 1997). Feeding behavior influences body size in kinosternid turtles. For example, individuals that consume more animal protein grow larger and megacephaly evolves in populations that feed on mollusks (Vogt and Guzman-Guzman 1988; Iverson 1999).

As part of an integrated study of the ecology of freshwater riverine turtle communities in the southern United States, we analyzed food resource partitioning in the turtle communities of the Cahaba, Chickasawhay, and Pearl rivers. These represent some of the most diverse turtle communities in North America, if not the world. For each community, we compared food resource utilization by the 7 most abundant turtle species. We used quantitative analyses to measure overlap in resource utilization and diet specialization. The goal of the study was to quantitatively document food resource partitioning in complex turtle communities.

METHODS

Study sites (Fig. 1) were established on the Cahaba River near Sprott, Perry County, Alabama (32°40′5.0016″N, 87°14′30.0012″E [datum NAD83]; Fig. 2); the Chickasawhay River at Leakesville, Greene County, Mississippi (31°8′54.96″N, 88°32′52.98″E; Fig. 3); and the Pearl River at Georgetown Water Park, Copiah County, Mississippi (31°52′31.98″N, 90°8′17.988″E; Fig. 4). We selected a sampling area on a 4–5-km stretch of each river where we had previously sampled and confirmed abundant populations of turtles. The 3 rivers at the study sites are similar in character: narrow and winding with moderate current, abundant snags and tree trunks in the channel, and numerous sandbars. The Cahaba River is narrower, shallower, and has clearer, colder water than the other 2 rivers. The rivers are subject to massive flooding in winter and spring, so all collecting was done in the low-water period of May through August. We collected turtles with an array of unbaited fyke nets, using the techniques described by Vogt (1980). The same fyke nets were used in all 3 rivers, with the same fyke net in the same locality every time we sampled a river. We sampled for 10 d at each site and then rotated to another site for 10 d. The nets remained in the same locality for 10 d during each capture period at each site and the nets were set in the greatest possible variety of habitats at each study site.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

We checked nets twice daily (0600 hrs and 1800 hrs), removing dead fish and floating debris from the nets to avoid attracting carrion-feeding species and biasing the stomach content data. We took captured turtles to camp for processing. Camps were located on the riverbanks in approximately the middle of each study area so that we could watch over the nets and easily return to camp when boat motors failed. Each turtle was identified, sexed, measured, weighed, and marked with plastic Floy™ spaghetti or cinch-up tags. Stomach contents were extracted using the flushing technique of Legler (1977) and preserved in 75% ethyl alcohol (EtOH) for subsequent study. We verified whether complete extraction of stomach contents was successful by dissection of preserved specimens that we had previously flushed in the field. The contents of a total of 719 stomachs were identified and measured: 606 (84%) from stomachs flushed in the field and 113 (16%) from specimens collected at our field sites during 1978–1979 and preserved to examine gonads (Table 1). Efficiency of stomach flushing was tested by preserving 20 turtles after stomach flushing had failed to produce stomach contents. Dissection of these turtles later in the laboratory proved these turtles to have empty stomachs, suggesting that our flushing methods were successful in documenting empty stomachs as well as collecting stomach contents, if there were contents in the stomach. Food items were separated, counted, identified, and measured volumetrically by fluid displacement in graduated cylinders. Stomach flushing in turtles is successful given that turtles engulf everything whole or in large portions, i.e., they do not masticate their food, and digestion takes place in the small intestine (Bjorndal and Bolten 1993). The stomach is a reservoir holding food until it can be passed to the small intestine. Herbivorous species have an elongated cecum to help digest plant material, slowing the movement of food through the gut and consequently facilitating the results of flushing their stomachs (Legler 1977).

Table 1. List of species found in each river (n = 2511 turtles captured). The total number for each species in each river is the total number of each species captured, with total males/females (M/F) captured and M/F captured at each site. S is the number of stomach contents of M/F analyzed at each site. Rank is by number of total individuals captured, not counting recaptures. Note that we did not include stomach contents of Chelydra serpentina, Macrochelys temminckii, Sternotherus odoratus, or Graptemys pseudogeographica kohnii in the analysis because numbers were so low and they were not collected in all rivers.

View Fullscreen

For analysis, we separated food items into 6 categories: insects, crustaceans, mollusks, fish, vegetation, and unidentified. We used these broad categories in that insects were mainly caddisfly cases, ephemenopteran nymphs, and other insect parts; thus, there were no large categories of insect species we could list; crustaceans were just 1 species of crayfish; mollusks were bits of meat and shell; fish were bits and pieces of fish; and plants were mainly leaves and algae. Our sample sizes were not large enough for any species to warrant reporting samples by size or sex of the turtle; the data were taken for each individual turtle, but there were no obvious differences between the sexes or sizes of the turtles whose stomach contents were flushed to warrant separate analysis by sex or size of the turtle. The original data of the stomach contents are no longer available because, after C.J.M.'s death, we were not able to locate the original data, so no further analysis was possible. Ontogenetic diet shifts have been reported in some freshwater turtles (Parmenter and Avery 1990), but as our trapping technique caught relatively few juvenile turtles, the probability of biased results due to ontogenetic shifts in diet should be small. Also, Eisemberg et al. (2017) reviewed dietary studies of Amazon river turtles published over the last 40 yrs comparing body size, phylogeny, season, and habitat; they found no correlation between maximum carapace size and plant material consumed.

The percent volume of prey items representing each of the 6 food classes at each locality was obtained by summing the volumes for each food class for all individuals of a species and dividing by food volume for all classes in the sample (× 100; Tables 2–4). Food class volume as a percentage of total food volume was the basis for calculations of diet overlap and dietary diversity (Berry 1975). We justify the use of broad categories of food items in that Cummins (1967) showed that species within broad groups, such as insects, crustaceans, and mollusks, showed very little variation in the caloric equivalents in studies of ecological energetics.

Table 2. Food class volume as a percentage of total diet and dietary diversity (H′) for aquatic turtles in the Cahaba River. Unid = unidentifiable matter. n is total number of stomach contents analyzed for each species followed by number of males/females (M/ F).

View Fullscreen

Table 3. Food class volume as a percentage of total diet and dietary diversity (H′) for aquatic turtles in the Chickasawhay River. Unid = unidentifiable matter. n is total number of stomach contents analyzed for each species followed by number of males/females (M/ F).

View Fullscreen

Table 4. Food class volume as a percentage of turtles in the total diet and dietary diversity (H′) for aquatic turtles in the Pearl River. Unid = unidentifiable matter. n is total number of stomach contents analyzed for each species followed by number of males/females (M/ F).

View Fullscreen

We calculated H′ to quantify specialized vs. generalized feeding (Schoener 1968; Berry 1975):

where p_i is the frequency of use of the ith category. The lowest values of H′ result from the lowest diversity of food classes, hence the most specialized feeders should have low values of H′. We calculated H′ for each species at each locality (Tables 2–4). The techniques of Schoener (1968), as adapted by Berry (1975) and Vogt and Guzman-Guzman (1988), measured degree of overlap in resource utilization. The amount of overlap between pairs of species in utilization of the 6 food classes at each locality was estimated by calculating Morisita's modified index of niche overlap (Cλ; Horn 1966; Krebs 1989). Cλ varies from 0, where no food classes are in common, to 1, where all food classes are used in the same proportions. We calculated Cλ values for all species pairs at each locality (Tables 5–7). We are aware of the potential for error introduced by comparing samples from dissimilar habitats in which resources may be differently distributed (Colwell and Futuyma 1971). Nevertheless, we agree with Berry (1975) that Cλ from pooled samples provides at least a rough approximation of the degree of overlap.

Table 5. Diet overlap (Morisita's index of niche overlap, Cλ) among aquatic turtles in the Cahaba River.

View Fullscreen

Table 6. Diet overlap (Morisita's index of niche overlap, Cλ) among aquatic turtles in the Chickasawhay River.

View Fullscreen

Table 7. Diet overlap (Morisita's index of niche overlap, Cλ) among aquatic turtles in the Pearl River.

View Fullscreen

RESULTS

Comparative Diet Diversity by Species. — We trapped for 2428 trap days during the periods 1 July to 13 August 1978 and 30 April to 4 August 1979, capturing 2707 turtles of 16 species (Table 1). The aquatic turtle faunas of the 3 rivers consisted of either 9 (Cahaba and Chickasawhay rivers) or 10 (Pearl River) species. The number of conspecifics captured at a given site ranged from 8 to 93 (mean number of individuals per species per site = 34.2). Only the 7 most abundant species in each river are treated here. Numbers of alligator snapping turtles (Macrochelys temminckii), eastern snapping turtles (Chelydra serpentina), the eastern musk turtle (Sternotherus odoratus), and Mississippi map turtles (Graptemys pseudogeographica kohnii) were too low to include in analyses. Also, although we collected stripe-necked musk turtles (Sternotherus peltifer) in the Chickasawhay River upstream from Leakesville in Wayne County (McCoy et al. 1978), none were collected at the Leakesville site.

Cahaba River. — In the Cahaba River (Table 2), the generalist red-eared slider (Trachemys scripta elegans) had the most diverse diet, followed by S. peltifer. Although T. s. elegans ate few insects (7% by volume), it utilized far more crustaceans than any other species at the site (17% of total intake). The remaining portion of the diet that we could identify consisted of approximately equal volumes of fish and plant material. Trachemys scripta is a scavenger and much of the fish volume undoubtedly represents carrion. The species was always differentially attracted to nets that contained dead fish. Sternotherus peltifer had the greatest relative volume of mollusks (19%) at the Cahaba River site, but fish were the most common item in its diet (40% of volume). The relatively high dietary diversity of S. peltifer is also due to relatively large mollusk and plant intakes (19% and 13% of volume, respectively). The third most diverse diet in this community was that of the Gulf coast smooth softshell (Apalone mutica calvata). Apalone mutica calvata consumed mainly insects at this site with the remainder of the diet composed of near equal quantities of mollusks and plant matter. Its congener, the Gulf Coast spiny softshell (Apalone spinifera aspera), ranked fourth in dietary diversity. Apalone spinifera relied heavily on fish (60% of volume), most of which were probably caught alive. Also, a relatively large volume of plant matter (23%) was consumed by this species. The diet of the black-knobbed sawback (Graptemys nigrinoda) at this site consisted mainly of insects (42% of volume) and plants (54% of volume). The remaining two species at this site, the Alabama map turtle (Graptemys pulchra) and the eastern river cooter (Pseudemys concinna concinna) had very low and similar diet diversities, yet highly specialized and very different diet compositions. The strong specialization of P. c. concinna on plant matter is clearly reflected in these data. At this site, the diet of P. c. concinna consisted of 92% plant material with only small amounts of insects, mollusks, and fish. Graptemys pulchra at this site relied almost exclusively on insects with a small amount of plant matter (9% of volume). However, most of the G. pulchra captured at this site were adult males.

Chickasawhay River. — In the Chickasawhay River (Table 3), the greatest dietary generalist was A. m. calvata, which ate about 20% each by volume of insects and crustaceans and 41% by volume of fish. The razor-backed musk turtle (Sternotherus carinatus) ranked second in dietary diversity, although the diet of this species was high in plant material (55% by volume). Apalone spinifera was third in dietary diversity owing to large proportions of both fish and plant material (55% by volume). Trachemys scripta ranked fourth in dietary diversity at this site. In this river, the diet of T. scripta was composed mainly of vegetation (60% of volume) with only small amounts of other food types. The 2 species of Graptemys in this community, the yellow-blotched sawback (Graptemys flavimaculata) and the Pascagoula map turtle (Graptemys gibbonsi), ranked fifth and sixth in dietary diversity, with very similar levels of trophic specialization but substantially different diets. The diet of G. flavimaculata was composed mainly of plant matter (63% of volume), but also included the largest proportion of insects at this site (32% of volume). Graptemys gibbonsi, on the other hand, specialized on mollusks, which made up 82% of its diet. The most specialized diet at this site was that of P. concinna, which consumed plant matter almost to the exclusion of all other foods (99% of volume).

Pearl River. — The pattern of relative dietary diversity seen in the Pearl River community (Table 4) was similar to that found in the Chickasawhay River, although the proportions of the various classes of food items varied somewhat and diet diversity was higher overall. At this site, A. mutica was again the most generalist feeder. However, at this site the majority of this species diet was insects, but with a substantial amount of plant material (29% of volume), which was only eaten in small quantities in the Chickasawhay River. The second most diverse feeder at this site was A. spinifera, which consumed mostly plant material. At the Pearl River site, the diet of A. spinifera included a much larger proportion of insects (23% of volume) and a much smaller proportion of fish (25% of volume) than was seen at the other sites. At this site, the generalist feeder T. scripta ranked third in dietary diversity. Trachemys scripta consumed mainly fish and plant matter, in approximately equal amounts, with insects and mollusks in smaller quantities. The fourth most diverse diet at this site belonged to S. carinatus. The diet of S. carinatus was highly similar to the diet of T. scripta, differing only by a higher proportion of plant material and somewhat smaller proportions of the other food classes. The 2 species of Graptemys found at this site had highly similar dietary diversities, although the composition of the diets differed. The Pearl map turtle (G. pearlensis) and the ringed sawback (G. oculifera) also showed substantially higher dietary diversity at the Pearl River site than the Graptemys at either of the other sites. In the Pearl River, the main constituent of the diet of G. oculifera diet was plant matter (38% of volume), although insects constituted a similar proportion (33% of the volume). Mollusks were also present in substantial quantity, while fish were only a small fraction of the diet. Graptemys pearlensis also consumed substantial proportions of insects and mollusks but ate only a small amount of plant matter. The most represented food in the diet of G. pearlensis at this site was fish (44% of volume). Again, the plant specialist P. concinna had the least diverse diet at this site.

Diet Overlap. — Given that we considered 3 highly diverse communities sharing fairly limited food bases, the results of the calculation of Morisita's index of niche overlap (Cλ) show a pattern of surprisingly little dietary overlap at these 3 sites (Tables 5–7). Most surprisingly, at 2 sites there was relatively little overlap between the species with the most generalized diets. In the Cahaba River, T. scripta, A. mutica, and S. peltifer showed the most generalized diets, with dietary diversities substantially greater than any of the other species at this site. However, dietary overlap among these species ranged from a low of 0.440 between T. scripta and A. mutica to a high of only 0.667 between T. scripta and S. peltifer. Similarly, in the Chickasawhay River, the 3 most generalized feeders (A. mutica, S. carinatus, and A. spinifera) showed overlap values ranging from 0.374 between A. mutica and S. carinatus to 0.778 between A. spinifera and S. carinatus.

In the Pearl River, in contrast, the overlap between dietary generalists was appreciably larger. The 3 species with the least specialized diets at this site were A. mutica, A. spinifera, and T. scripta. At this site a fourth species, S. carinatus, had only a slightly less generalized diet. Dietary overlap was substantial among these 4 species, ranging from a low of 0.753 between S. carinatus and A. mutica to a high of 0.928 between S. carinatus and A. spinifera. In addition, at this site there was substantial overlap between the congeners A. mutica and A. spinifera.

Those species with more specialized diets generally displayed relatively low overlap values, although there was again less differentiation among the species at the Pearl River site. In the Cahaba River, the plant specialist P. concinna had very low overlap with all other members of the community except G. nigrinoda. The large dietary overlap between these species was largely due to a relatively large proportion of plant material consumed by G. nigrinoda at this site (although part of the material we identified as plants could have been sponges; Lindeman 2016). The two species of Graptemys at this site, despite showing specialized diets, had a high degree of overlap with each other and with other members of the community.

The specialist species in the Chickasawhay River community display an interesting pattern of niche separation. At this site P. concinna, despite showing extreme dietary specialization, showed high overlap with S. carinatus, T. scripta, G. flavimaculata, and A. spinifera due to large proportions of plant material in the diets of all these species. Although G. flavimaculata also displayed substantial dietary specialization at this site, this species also had high overlap with several other members of the community; again, this was largely the result of large proportions of plant matter in the diets of several species. At this site, plant matter represented over 50% of the diet in 4 of the 7 species. The greatest degree of niche separation in this community was displayed by G. gibbonsi. This mollusk specialist displayed very low overlap with all other species at this site.

Reflecting the greater overall diet diversity found in the Pearl River, even the most specialized feeder is this community displayed high niche overlap. The species with the lowest overlap values at this site was P. concinna. This was also the only species with a diet diversity (H′) value indicating a specialized diet. All other species at this site had fairly generalized diets and high overlap.

DISCUSSION

This study, conducted over 40 yrs ago, represents one of the first quantitative analyses of food resource partitioning within complex aquatic turtle communities. The 3 aquatic turtle communities examined in this study are among the most diverse in the world. The pattern of resource utilization within each of these communities is likewise highly complex. Also, despite overall similarities in species composition, there are important differences in food resource partitioning among these communities. Our data for the stomach contents of each species within each river are comparable because all species had the same resources available to choose from. It was not necessary to sample the resource base to show these differences, as we were not intending to test for competition for food resources, which could only be proven if a resource was in limited supply and 1 species consumed it in a higher quantity than did others. Although we compared what the same species ate in each river, the differences we report could occur because the availability of resources in each river is different.

Two of these communities, in the Cahaba and Chickasawhay rivers, consist of species that can clearly be classified as either generalist or specialist feeders. Several species within each of these communities were clearly omnivorous, with diets that include substantial proportions of several types of food. Even among these generalist feeders, however, there was enough specialization that the degree of niche overlap was not extreme. These generalists tended to have moderate overlap with many other species, but very high niche overlap with only very few (or none) of the other species. Among the specialized feeders, on the other hand, there was a pattern of overlap with other species that tended to be either extensive or very slight. Specialist species tended to show extensive niche overlap with any other species that consumed a large proportion of the specialists' preferred foods. Niche overlap for these specialists will be very slight, with species that do not consume much of the specialists' preferred foods. Surprisingly, both of these communities included a pair of specialist feeders whose diets overlapped substantially. The diet of P. concinna, a plant specialist, overlapped considerably with the diets of G. nigrinoda in the Cahaba River and G. flavimaculata in the Chickasawhay River. Although both of these species of Graptemys had relatively specialized diets, both consumed large proportions of plant matter, which resulted in the high overlap with the diet of P. concinna.

The aquatic turtle community in the Pearl River displayed a markedly different pattern of resource partitioning and utilization. Within this community, only a single species (P. concinna) had a diet that might be classified as specialized. The other 6 species had high diet diversities and appeared to be largely omnivorous. The analysis of food class volume as a percentage of total diet for each species showed a clear pattern of wide utilization of each food class by several species. In this community, each food class except crustaceans constituted a substantial proportion of several species diets. Crustaceans represented only a very small dietary proportion in any of the turtles in this community.

Lindeman (2016) found extreme food partitioning between G. nigrinoda and G. pulchra in the Alabama River, with sponges representing 75% of the diet in juvenile G. nigrinoda, 54% in males and juvenile females, and 35% in adult females. In G. pulchra, sponges were much less important, with only trace amounts found in the feces of juveniles, males, and subadult females and 5% of the volume in adult females. Clams were the most prevalent food item in the feces of adult females with 99% by volume, 90% in subadult females, 66% in males, and 21% in juveniles. Lindeman found clams in only 3% of the volume of adult female G. nigrinoda feces. Insects were most abundant (75%) in G. pulchra juveniles, while comprising 24% in males and only 8% in juvenile females and 0.95% in adult females. Filamentous algae made up a high volume of feces—33% in adult female G. nigrinoda and 12% in males, but very little in G. pulchra at any size category. Shealy (1976) documented the ontogenetic dietary shift from insects to mollusks in a closely related species, Graptemys ernsti.

The data obtained from feces examination (Lindeman 2016) differs greatly from the results presented here from stomach flushings and stomach contents. While we agree that much can be learned about the identification of the food items from feces, this evidence only includes the undigested remains. It is difficult if not impossible to extrapolate the total volume of the organisms eaten from the volume of clam shells or opercula of snails found in feces, as all of the muscle and other soft tissue which are represented in the stomach contents are digested before being passed as feces (Caputo and Vogt 2008). Some turtle researchers, however, note that feces samples provide better results than stomach flushings, as they represent a collection of many days in the gut. Uncertainties around whether wild turtles eat every day and what time of day they feed may lead to inconsistencies in results of stomach contents. For example, if they feed in the late morning, stomach contents may have emptied into the intestines by the time a turtle is taken out of the trap at 0700 hrs the next day. In contrast, feces contents are more reliably recovered given their longevity in the gut.

The results from Lindeman (2016) contrast to our data of stomach flushings of G. pulchra from the Cahaba River, where we found 90% of the diet to be insects and 9% plants. While part of this plant material may have actually been sponges, we did not explicitly identify sponges in any of the stomach contents of any species in our study. Kofron (1991) similarly did not find sponges in the stomach contents of G. oculifera examined from museum specimens from throughout their range, finding “. . . only caddis flies, dipteran flies, mayflies, beetles and plant material. Many stomachs contained small pieces of wood, suggesting that fallen tree trunks were a ‘grazing' substrate.” He did not mention mollusks. We agree with Kofron (1991) that the small pieces of wood found in the stomach flushings of Graptemys represent grazing behavior, but pertaining to caddisfly feeding, not sponges. Selman and Lindeman (2015) analyzed feces from 4 male and 8 female G. gibbonsi to study their diet in the Leaf River. They found a much higher percentage of mollusks in the feces of the 4 females (nearly 97% by volume) than in the 8 males (13%). These mollusks, however, were primarily invasive Asian Corbicula. Selman and Lindeman (2018) also found Corbicula in 4% of feces by volume from female G. flavimaculata in the Leaf River, but none in the feces of males or in any G. flavimaculata, males or females, from the Pascagoula River. However, they did find native mussels in the feces of males and females (10% and 32% of volume, respectively) from the Pascagoula River. This population seems to be selecting the native mollusks over the more abundant invasive species, perhaps due to softer shells or taste preferences. In contrast, the abundance of Corbicula in the diet of G. gibbonsi suggests that they are a preferred food item for this species. Additionally, G. gibbonsi may be outcompeting G. flavimaculata for this prey, or Corbicula may be too big and hard for G. flavimaculata to consume.

In the Pearl River, G. pearlensis, the sister species to G. pulchra and G. gibbonsi, showed stomach flushings of 24% insects, 24% mollusks, and 44% fish, while the sister species to G. nigrinoda, G. oculifera, showed 33% insects, 22% mollusks, and 38% plants by volume in stomach flushings. It is noteworthy that G. oculifera had 22% mollusks in its diet while G. nigrinoda had only 0.2% mollusks and G. flavimaculata had 4% mollusks in our study. Part of the feeding behavior of these species may be controlled by what is available in each river; our data suggest that the Cahaba River is depauperate in mollusks, as none were present in the 11 G. pulchra stomach flushing samples and only 0.2% in G. nigrinoda, but S. peltifer (n = 28) and A. mutica (n = 9) in the Cahaba River had 19% and 14% of the volume of the stomach contents as mollusks, respectively. This may be evidence for competition, suggesting that mollusks are a limited resource, and that S. peltifer and A. mutica are more adept at harvesting mollusks than are Graptemys spp. Sternotherus peltifer is a bottom-walking species, gleaning mollusks along the rocky stream bottom, while A. mutica is an expert swimmer that feeds in the water column as well as in sandy bottom areas. Nickerson and Pitt (2012), snorkeling in clear water, observed Graptemys geographica foraging for snails in the White River in Missouri. The turtles would move along the bottom and flip over rocks to specifically find and consume small snails. The water in the southern rivers of this study is too murky to allow for observations of turtles feeding.

This lack of specialization in diet produced a different pattern of niche overlap among the species in these communities. Among most species in the Pearl River, there was considerable overlap in diet. Only P. concinna showed any degree of niche separation, yet overlap with some other species was quite high. As there was no obvious difference in the species composition of this community, the observed difference in resource utilization was probably due to differences in resource availability.

Four of the species considered in this study occurred at all 3 sites, providing the opportunity to examine shifts in feeding niches due to differences in community structure or food bases. Of these 4 species, P. concinna displayed the most consistency in diet among sites; at all sites P. concinna specialized on plants and had very low diet diversity. Pseudemys concinna was the most specialized feeder at 2 of the sites and the second most specialized at the third site. Diet diversity for this species was lowest in the Chickasawhay River where no fish or insects appear in the diet, although both were consumed in small quantities at other sites. This extreme specialization does not appear to be due to diet overlap with other species, because both of these resources were more heavily utilized by species in the other two communities.

Trachemys scripta, widely known as a generalist, displayed substantial shifts in diet among these sites. In both the Cahaba and Pearl rivers, the diet of T. scripta was highly generalized despite the composition being quite different. In the Chickasawhay River, T. scripta displayed less diet diversity, while still being the fourth most generalized feeder.

The 2 species of Apalone (A. mutica and A. spinifera) displayed an interesting pattern of resource utilization among these sites. In all 3 rivers, these species showed a pattern of diet differentiation very similar to that described by Williams and Christiansen (1981). In all 3 rivers, A. mutica consumed a higher proportion of insects while A. spinifera consumed more fish and plant matter. However, the actual composition of the diets for these species in the 3 communities studied here varied substantially. In the Cahaba and Pearl rivers, A. mutica consumed only very small amounts of crustaceans, which are a large part of their diet in the Chickasawhay River. While A. spinifera consumed a smaller proportion of insects than did A. mutica in all 3 rivers, insects make up a large proportion of the diet of A. spinifera diet in the Pearl River. Both of these species had higher diet diversities in the Pearl River than at the other 2 sites.

Our results demonstrate that resource partitioning does occur within highly complex aquatic turtle communities. Not surprisingly, the interactions observed were also highly complex. However, distinct patterns of differences in resource utilization were observed among the species studied. There were also differences among the communities in the patterns of resource utilization. This suggests that even in these highly diverse communities, resource utilization is subject to differences in community structure and resource availability. Future studies on populations of these turtles should attempt to follow individual turtles throughout the feeding season to see if individual turtles have specific food preferences or if they are ingesting what is available at any one time of the year. We had some preliminary data in that some specimens of P. concinna and T. scripta were flushed several times during the course of this study, but the data did not suggest they were feeding differently during the several times they were flushed.