in the Common Musk Turtle, Sternotherus odoratus

Functional Ecology 2001 Physical apertures as constraints on egg size and shape Blackwell Science, Ltd in the Common Musk Turtle, Sternotherus odoratus P. J. CLARK, M. A. EWERT and C. E. NELSON Department of Biology, Indiana University, Bloomington, IN 47405 3700, USA Summary 1. Egg size in turtles often increases with female size, contrary to expectations of optimality. Functional constraints on egg width imposed by the pelvic aperture or the gap between the carapace and plastron (the caudal gap) have been inferred for a few populations but appear inapplicable in others. 2. For Sternotherus odoratus (the Common Musk Turtle), the pelvic aperture was always wider than the width of the female s largest egg by at least 3 7 mm. The caudal gap was narrower than the widest egg for 25 7% of the females. 3. Egg width increased, and elongation (length/width) decreased, as female size and clutch size increased. 4. Females at three ecologically contrasting sites differed appreciably in size but produced eggs of the same mean shape and size, despite the strong within-site changes in both egg size and shape with female size. As the younger females at all sites were of similar age and produced eggs of similar size and shape (again, despite differences in body size), egg size and shape may be age-specific. 5. No optimal egg size prevailed but the scaled residuals of egg size with female mass were less variable than were those for clutch size. Key-words: Chelonia, optimal egg size, pelvic aperture, reproductive strategy Functional Ecology (2001) Ecological Society Introduction In some populations of freshwater turtles, both egg size and clutch size increase with female size. This increase contrasts with the expectation from optimal egg size theory that egg size should be relatively constant while clutch size should vary with the levels of resources available for investment in reproduction (Smith & Fretwell 1974). In the seminal analysis of this phenomenon, Congdon & Gibbons (1987) found increases with female size both for egg width and for the width of the pelvic aperture (through which the eggs must pass during oviposition) for Chrysemys picta in Michigan. Further, the width of the pelvic aperture in small turtles was apparently too narrow to allow them to lay eggs as wide as most of the eggs laid by large turtles. From these data, and from similar data on Deirochelys reticularia, Congdon & Gibbons (1987) concluded that the need for passage through the pelvic aperture had constrained many of the eggs of these two species from reaching otherwise optimal sizes. In contrast, they found for Trachemys scripta that egg width increased much more slowly than pelvic aperture and was distinctly smaller than pelvic aperture, and, thus, seemed not to be limited by it. Author to whom correspondence should be addressed. As further support for the pelvic aperture as potentially limiting in some species, Long & Rose (1989) found that these apertures were larger in females than in males in three diverse species. Gravid females from eight additional populations in four species have been X-rayed subsequent to Congdon & Gibbons (1987) to determine the relationship of egg width to pelvic aperture width (DePari 1988; Iverson 1991; Iverson & Smith 1993; van Loben Sels, Congdon & Austin 1997; Chen & Lue 1999). In one population of C. picta in New Jersey, egg width was only slightly greater than pelvic aperture at small female body sizes and, thus, appeared to be constrained by pelvic aperture (DePari 1988). In each of the seven other populations, although egg width and pelvic aperture width both also increased with female size, the difference between egg width and the pelvic aperture was so large that most or all of the small turtles appeared capable of laying eggs as wide as any produced by large turtles in the same population. An alternative morphological constraint has been suggested. Eggs must pass through the gap at the back of the shell between the carapace and the plastron (the caudal gap) after they clear the pelvis. In Gopherus berlandieri this gap, when measured, was smaller than egg width; however, the posterior plastral lobe in ovipositing females can rotate enough for the eggs to pass (Rose 70

71 Physical apertures as constraints on eggs & Judd 1991). Similarly, although the pelvis is rigidly attached both to the carapace and plastron in side-necked turtles (Pleurodira), in Pseudemydura umbrina an unknown mechanism allows the carapace and plastron to move apart during oviposition (Kuchling 1999, p. 84). Caudal gap was severely constraining in a different case (M. A. Ewert, unpublished data). During captive breeding a female Rhinoclemmys annulata (Gray) broke her first four eggs (i.e. four successive single-egg clutches) after the eggs were protruding roughly a third of their length distal to her cloaca. Once the caudal gap was enlarged by filing the opposing surfaces, she laid her next seven eggs (= seven clutches) without breaking any. However, in each case (i.e. for all 11 clutches) oviposition was induced by oxytocin injection and the caudal gap might be more compliant when turtles nest and oviposit naturally. The present study explored the relationships among egg size, egg shape (length/width = elongation), female size, pelvic aperture and caudal gap in Sternotherus odoratus (Latreille). Females and their clutches were obtained from three ecologically diverse sites where mean female sizes differed appreciably (Clark 2000), increasing the chance of finding informative contrasts. Materials and methods Gravid females of S. odoratus were collected during 1992, at three south-central Indiana (USA) sites: Grassyfork (G; Grassyfork Goldfish Hatchery, Morgan County, IN; 39 30 N, 86 24 W), Yellowwood (Y; Lake Yellowwood, Yellowwood State Forest, Brown County, IN; 39 11 N, 86 21 W) and Crane (C; Lake Greenwood, Crane Naval Weapons Support Center, Martin County, IN; 38 52 N, 86 50 W). The number of gravid females in each sample was: 31 from Grassyfork (all collected on 15 May), 22 from Yellowwood (all on 26 May) and 17 from Crane (collected 6 20 June). Carapace length (CL), caudal gap and live female postovipositional mass (= female spent mass, FSM) were measured directly for each female. Oviposition was induced by injection of oxytocin (Ewert & Legler 1978). Freshly laid eggs were measured for mass, length and width. To obtain the width of the pelvic aperture and clutch size, females were X-rayed on table-top film cassettes at 185 ma and 50 kv from 1 m for 0 07 or 0 05 s. (Gibbons & Greene 1979; see also Hinton et al. 1997). Most females later died in captivity (G, 16 of 31 females, Y, 12 of 22, C, all 17), for reasons unrelated to this study (they froze during a winter power failure), and were dissected so that the width and height of the pelvic aperture could be measured directly. A comparison of direct and X-ray measurements of pelvic aperture width determined that X-ray magnification (Graham & Petokas 1989) ranged only from 0% to 1%, so no correction was needed. The data for various parameters were initially compared graphically using Statview 512 (SAS Institute Inc., Cary, NC, USA). Statistical significance testing used Minitab 8 2 (Minitab Inc., State College, PA, USA), Statpack 4 11 (Australian Assoc. Math. Teach., Adelaide, Australia), Statview 512 and SAS (Proc GLM, with ANCOVA; SAS Institute Inc., Cary, NC, USA). Student Newman Keul or Tukey s HSD were used as post hoc multiple range tests on significant ANOVAs. For residuals analyses, each relevant association was defined by least squares and the data point residuals were examined with non-parametric analyses ( Wilcoxon signed rank or Spearman s α, as appropriate). Results FEMALE BODY PROPORTIONS Pelvic apertures of the 45 dissected females included 19 that were wider and 26 that were higher than perfectly circular but never by more than 4 5 mm (in 22 6 mm) with no significant bias (z = 1 686, P = 0 09, Wilcoxon signed ranks test). The ratio of pelvic aperture width to height ranged from 0 81 to 1 2 (mean ± SD 0 98 ± 0 08) and did not vary significantly with female CL (r = 0 09, slope = 0 001, P < 0 5). Therefore, all further references to pelvic aperture will be in terms of its width, as measured from X-ray photographs. Females from Grassyfork averaged largest (FSM, CL), as did their pelvic apertures and caudal gaps (Table 1). Females from Yellowwood and Crane were similar in size, but those from Yellowwood had significantly larger pelvic apertures. After ANCOVA adjustment for body size, relative aperture was significantly larger in females at Yellowwood than at either other site (Table 1). At each site, both pelvic aperture and caudal gap increased with CL, though less strongly in the Crane sample, which contained the fewest turtles and least variability in their size (Fig. 1, Table 1). Caudal gap as a proportion of female CL was isometric (origin lies within 90% confidence limits of caudal gap regressed on CL) and did not differ among sites (Table 1). Pelvic aperture was allometric as a proportion of CL at each site and for all sites combined: pelvic apertures were larger relative to CL in small females (P < 0 001). The caudal gap was smaller than the pelvic aperture in 68 of the 70 females (97 1%) from all sites (the two exceptions were the 2nd and 8th largest turtles). Pelvic aperture and caudal gap differed most in small turtles, e.g. by averages of 5 6 mm in turtles below median size, of 4 7 mm in turtles above median CL (and of only 1 9 mm in the five largest turtles). EGG SIZE AND SHAPE Mean egg size and egg elongation (length/width) did not differ significantly among sites, despite the large differences in average female size (Table 1). Egg width (both as means and maximums per clutch) increased significantly with CL both within sites and for all sites

72 P. J. Clark et al. Table 1. Measurements of female size, clutch size and egg size (± 1 SE) for Sternotherus odoratus from three field sites: Grassyfork (G), Yellowwood (Y), and Crane (C). Egg measurements are the means of clutch means Attribute Grassyfork Yellowwood Crane Comparison among sites* Site means ranked Sample size 31 22 17 70 Female spent mass (g) 187 3 ± 8 8 122 4 ± 6 5 117 6 ± 5 8 F 25 88, P < 0 001 G > Y = C Carapace length (mm) 104 7 ± 1 7 90 9 ± 1 7 89 0 ± 1 4 F 25 90, P < 0 001 G > Y = C Pelvic aperture (mm) 23 8 ± 0 3 22 5 ± 0 4 20 9 ± 0 3 F 15 12, P < 0 001 G > Y > C Pelvic aperture (least squares means) 22 6 23 4 22 0 F 6 40, P < 0 003 G < Y > C Caudal gap (mm) 19 3 ± 0 6 15 9 ± 0 5 16 3 ± 0 5 F 12 32, P < 0 001 G > Y = C Caudal gap (leasts quares means) 17 6 17 1 17 9 F 0 72, P = 0 49 G = Y = C Clutch size (number) 6 7 ± 0 3 3 8 ± 0 2 3 2 ± 0 2 F 43 08, P < 0 001 G > Y = C Mean egg mass (g) 3 95 ± 0 13 4 07 ± 0 15 3 75 ± 0 1 F 1 14, P = 0 33 G = Y = C Mean egg length (mm) 26 8 ± 0 3 27 3 ± 0 3 26 9 ± 0 3 F 0 58, P = 0 56 G = Y = C Mean egg width (mm) 15 4 ± 0 2 15 3 ± 0 2 15 0 ± 0 1 F 0 77, P = 0 47 G = Y = C Maximum egg width (mm) 15 8 ± 0 2 15 6 ± 0 2 15 2 ± 0 2 F 1 97, P = 0 15 G = Y = C Mean elongation 1 75 ± 0 02 1 78 ± 0 02 1 79 ± 0 01 F 1 55, P = 0 22 G = Y = C Mean pelvic clearance (mm) 8 00 ± 0 22 6 95 ± 0 27 5 72 ± 0 35 F 17 48, P < 0 001 G > Y > C Mean caudal clearance (mm) 3 54 ± 0 49 0 33 ± 0 37 1 09 ± 0 44 F 14 57, P < 0 001 G > Y = C *All P-values for F-ratios are based on df = 2,67 except those involving ANCOVA (two cases, indicated as least-squares means), for which df = 2,66. The inequality signs indicate that two or more sites differed at P < 0 05. The equal sign indicates that the adjacent listed sites did not differ by P = 0 05. CL was usually stronger than that with pelvic aperture or caudal gap. Within individual clutches, heavier eggs were more elongate for females from Grassyfork (1 72 vs 1 76; Wilcoxon signed ranks test, z = 2 69, P < 0 01) and Crane (1 75 vs 1 83; z = 2 77, P < 0 01) but not for those from Yellowwood (1 79 vs 1 77; z = 0 73, P > 0 4). Among all individual eggs across sites, elongation did not vary significantly with egg mass (ρ = 0 078 for 340 eggs, P = 0 15). Among clutches for all sites combined, mean egg elongation decreased with clutch size (Table 2). After adjustment for body size (CL), this association was lost (changed from r = 0 36, P = 0 002 to r = 0 033, P = 0 79; residuals analysis). That is, elongation tended to be a stronger associate of body size than of clutch size. Fig. 1. Association of pelvic aperture (bold or closed symbols) and caudal gap (open symbols) with female carapace length at three sites: Grassyfork (triangles), Yellowwood (circles) and Crane (crosses). See Table 2 for statistical detail. combined (Table 2). Mean egg elongation by clutch decreased significantly with CL at Yellowwood, for Grassyfork and Crane combined, and for all sites combined (Table 2). A second-order polynomial fit to the data indicated that elongation decreased strongly from small to moderate size females, but changed little from large to very large females (Fig. 2). Thus, smaller turtles tended to have more elongate eggs than did larger ones. They also tended to have smaller clutches (Table 2). The association of elongation with EGG WIDTH VERSUS PELVIC APERTURE AND CAUDAL GAP Maximum egg width increased with both pelvic aperture and caudal gap (Fig. 3), significantly so at Grassyfork and Yellowwood, and for all sites combined (Table 2). For every female, pelvic aperture exceeded maximum egg width by at least 3 7 mm. Caudal gap, however, was smaller than maximum egg width for 25 7% of the females (18/70), equalled it for one, and exceeded it for the rest (Fig. 3). Although some females at each site had maximum egg widths exceeding their caudal gaps (Fig. 3), egg width exceeded caudal gap significantly less frequently at Grassyfork (9 7%; 3/31) than at Yellowwood (50%; 11/22 of females), with Crane (23 5%; 4/17) intermediate (overall chi-square contingency test; χ 2 = 11 01,

73 Physical apertures as constraints on eggs Table 2. Associations between several variables for Sternotherus odoratus from three field sites Association Statistic Grassyfork Yellowwood Crane All sites Sample size 31 22 17 70 Pelvic aperture vs carapace length r 2 0 564 0 735 0 146 0 613 F 37 51 55 46 2 57 107 51 P-value <0 001 <0 001 0 13 <0 001 Caudal gap vs carapace length r 2 l. 0 542 0 344 0 322 0 364 F 34 35 10 5 7 11 38 85 P-value 0 001 0 004 0 018 <0 001 Mean egg width vs carapace length r 2 0 749 0 570 0 449 0 434 F 86 39 54 8 12 23 43 4 P-value <0 001 <0 001 0 003 <0 001 Maximum egg width vs carapace length r 2 0 662 0 577 0 551 0 486 F 56 90 27 26 18 38 64 38 P-value <0 001 <0 001 0 001 <0 001 Egg elongation vs carapace length r 2 0 065* 0 355* 0 261* 0 190* P-value 0 189 0 003 0 036 <0 001 ρ 0 254 0 503 0 387 0 407 P-value 0 160 0 020 0 120 <0 001 Clutch size vs carapace length r 2 0 435 0 422 0 267 0 655 F 22 36 14 58 5 45 128 84 P-value <0 001 0 001 0 034 <0 001 Clutch size vs egg elongation r 2 0 089* 0 038* 0 195* 0 128* P-value 0 103 0 195 0 076 0 002 ρ 0 271 0 153 0 511 0 357 P-value 0 110 0 452 0 041 0 003 Maximum egg width vs pelvic aperture r 2 0 504 0 634 0 069 0 456 F 29 50 34 61 1 11 56 97 P-value <0 001 <0 001 0 309 <0 001 Maximum egg width vs caudal gap r 2 0 459 0 365 0 171 0 348 F 24 59 11 49 3 10 36 35 P-value <0 001 0 003 0 10 <0 001 Pelvic clearence vs carapace length r 2 0 154 0 451 0 003 0 351 F 5 274 16 46 0 005 36 71 P-value 0 029 <0 001 0 945 <0 001 Caudal clearance vs carapace length r 2 0 355 0 073 0 110 0 451 F 15 99 1 569 1 858 49 45 P-value <0 001 0 224 0 194 <0 001 Mean egg mass vs female spent mass r 2 0 754 0 490 0 211 0 338 F 88 99 19 23 4 02 34 78 P-value <0 001 <0 001 0 063 <0 001 *The association is inverse; the others are direct. Fig. 2. Association of mean egg elongation (length/width) with female carapace length at three sites: Grassyfork (triangles), Yellowwood (circles) and Crane (crosses). The curved lines are least-squares fitted polynomials to show the prevailing trends among and across the sites. See Table 2 for statistical detail. Fig. 3. Association of maximum egg width with pelvic aperture (bold or closed symbols) or caudal gap (open symbols) at three sites: Grassyfork (triangles), Yellowwood (circles) and Crane (crosses). The bold dashed line represents the case when either the caudal gap or the pelvic aperture equals egg width. Points on the left and above this line represent cases in which the caudal gap was narrower than the eggs. See Table 2 for statistical detail.

74 P. J. Clark et al. df = 2, P < 0 005; post hoc 2 2 contingency tests: G Y χ 2 = 10 76, df = 1, P < 0 001; G C χ 2 = 1 69, df = 1, P > 0 1; Y C χ 2 = 2 84, df = 1, P > 0 05). Pelvic clearance (i.e. pelvic aperture minus maximum egg width) and caudal clearance (caudal gap minus maximum egg width) both differed significantly among sites, both were largest at Grassyfork (Table 1) and both tended to increase with female size (Table 2). Discussion CAUDAL GAP MOST CONSTRAINING Caudal gap was smaller than pelvic aperture height in every turtle and was smaller than pelvic aperture width in all except two of the largest turtles. The width of the widest egg was at least 3 7 mm narrower than the pelvic aperture in each female but was 3 7 mm narrower than the caudal gap in only one-fifth of the females. The widest egg was actually wider than the caudal gap in a quarter of the females. Further, the caudal gap was most consistently constraining in smaller females. The eggs of S. odoratus are brittle-shelled and cannot be deformed without breaking. To have laid the egg whose width most exceeded the caudal gap, plastral kinesis must have allowed expansion of the caudal gap by at least 3 6 mm (32%) over the measured value (11 1 mm). Slight, ventrally directed kinesis of the posterior lobe of the plastron is physically possible along the xiphoplastral hypoplastral suture in Sternotherus (Bramble, Hutchinson & Legler 1984). DETERMINANTS OF EGG WIDTH At Yellowwood, egg width (and mass) increased strongly with increasing female size, maximum egg width within a clutch exceeded the caudal gap in 50% of the females, and, thus, egg size was severely constrained by the female s anatomy (i.e. her caudal gap). At Grassyfork, in contrast, egg width (and mass) again increased strongly with increasing female size, but female body sizes were distinctly larger and maximum egg width exceeded the caudal gap in only 10% of the females, thus providing distinctly less evidence of a constraint due to the caudal gap. Further, mean egg size at Grassyfork did not differ significantly from that at Yellowwood, despite the differences in body size and the parallel increases of egg size with body size at both sites. This set of relationships would be coherent if egg size were age-specific rather than size-specific. At our three sites, females of similar age, in fact, may be laying eggs of similar size, with both egg size and body size increasing with age within each site. Differences among sites in age of maturation (estimated from females that retained the juvenile growth annuli; see Germano & Bury 1998) were small (G = 5 1 years, n = 13 females; Y = 5 2 years, n = 13; C = 5 4 years, n = 20). The youngest reproductive females at each site were of similar age (i.e. had similar counts of growth annuli on the plastron; five annuli at G, four at Y, five at C) and laid eggs of similar size even though the youngest females at Grassyfork were noticeably larger than the youngest at the other sites. Although the largest females were quite different in size across sites, the eggs they laid were again similar in size (see Clark 2000 for more details). These larger turtles were among the oldest at each site, as judged by deterioration of the head pattern, irregularity of seams on the plastron, and degree of apparent wear on shells. The differences among sites in body size for the youngest females and in mean body size probably reflected differences in ecological conditions. Grassyfork is an intensely managed goldfish hatchery. Forage was more abundant there than at the other two, unmanaged sites, and growth rates for the first two seasons posthatching were higher there also (Clark 2000). Differences in early juvenile growth also correspond to differences in size at sexual maturity in another turtle (C. picta) studied at several local sites (Rowe 1997). Halliday & Verrell (1988) found that differences in growth rate prior to maturation were a major determinate of adult size in a majority of amphibians and reptiles for which data were available. Chrysemys picta is the only other species where published data allow among-site comparisons of egg width with the size of the reproductive canal and with female size. In C. picta, too, egg width tends to increase with body size within local populations, regardless of whether the egg width is similar to or decidedly smaller than the pelvic aperture (Fig. 4). Our finding that mean egg width is not sensitive to the large differences in female size among sites in S. odoratus, and the similar results evident in Fig. 4 for C. picta, raise an interesting question of mechanism. In both species, physiological control of egg width apparently does not directly reflect the constraints imposed by the female s own pelvic aperture or caudal gap. Perhaps follicle and oviduct sizes vary more directly with female age than with differences in female size produced by disparate ecological conditions. Then, egg size could increase with female size withinpopulations and, simultaneously, be independent of any among-population differences in female size that were not a result of substantially different age structures. DETERMINANTS OF EGG SHAPE It appears that the oviducts can differentially elongate and otherwise shape eggs under differing proximate conditions (Ewert 1985, p. 78; Iverson & Ewert 1991; Tucker & Janzen 1998). That heavy eggs are more elongate than lighter eggs within the same female S. odoratus, even when the apertures are far from constraining, suggests that the oviducts and, to some extent, the follicles, rather than the physical apertures, have proximate control over egg width. Accordingly,

75 Physical apertures as constraints on eggs Fig. 4. Associations of pelvic aperture width (lines a ) or egg width (lines b ) with carapace length at 10 sites including Sternotherus odoratus (this study) and Chrysemys picta (published data). Some data in the original sources have been converted from plastron length to carapace length (see Iverson & Smith 1993). Numbered entries: S. odoratus: (1) Grassyfork, (2) Yellowwood, (3) Crane; C. picta: (4) Congdon & Gibbons (1987), (5) DePari (1988), (6) Rowe (1994; Dobbins site), (7) Rowe (1994; Hansen site), (8) Rowe (1994; Swan site), (9) Rowe (1994; Beem site), (10) Iverson & Smith (1993). a female may ovulate follicles of slightly different sizes, and the oviducts may then constrain the width more for large follicles than for small ones, resulting in heavier eggs that are more elongate, rather than simply larger in all dimensions. The oviducts could be involved, as well, in the change in egg shape from being relatively elongate in small females to less elongate in medium-to-large females (Fig. 2). The oviducts in small (or younger) females could be making egg shape more elongate than otherwise optimal to facilitate passage of heavier eggs. An alternative hypothesis, in which eggs become less elongate as the oviducts become packed with large clutches, is obscured because the initially significant association of egg elongation with clutch size disappears with adjustment for body size. This latter hypothesis parallels that of Elgar & Heaphy (1989) but takes it from a multispecies analysis to plasticity within a species. EGG SIZE AND CLUTCH SIZE Optimal egg size theory predicts that egg size should reflect hatchling ecology and be consistent across females within a site (Smith & Fretwell 1974). In contrast, within each of our sites, egg mass, like egg width, increased with FSM. Thus, S. odoratus had not attained a female-size independent, optimal egg size at these sites. The same conclusion has been reached for a number of other organisms where females continue to grow after reaching sexual maturity (Bernardo 1996). The failure of turtles to show consistently an optimal egg size independent of female size has led to a search for alternative approaches for testing for optimal egg size. Specifically, Roosenburg & Dunham (1997) compared the coefficient of variation (CV) for egg mass (13 1%) with that for clutch size (22 4%) in Malaclemys terrapin and concluded the difference was consistent with optimal egg size theory. Similarly, Nieuwolt-Dacanay (1997) found that egg width (not mass) varied little (CV = 4 1%) compared with clutch size (CV = 27 6%) in Terrapene ornata (see also Iverson & Smith 1993). In the present study of S. odoratus, mean egg width and egg mass were each less variable than clutch size at each site (CV for egg width G = 6%; Y = 7%; C = 5%; for egg mass, G = 18%; Y = 17%; C = 13%; and for clutch size, G = 27%, Y = 29%; C = 31%). Using this criterion of relative variability, egg size (both width and mass) was relatively constant at each site in this study. As clutch size and egg size both varied with female mass (Table 1), the effects of female size could be removed by comparing the residuals of each of the two associations with FSM. These two residuals, then, if scaled as a proportion of the expected values also remove the effects of different units (mm or g vs number of eggs) and reduce the effects of differences in the range of measurement (i.e. of egg mass varying by a factor of 2 and clutch size by a factor of 5). Five potential relationships were examined by comparing the scaled residuals on body size (CL or FSM, as appropriate) for two measures of egg size width or mass with the scaled residuals on body size for clutch size. The five were: (1) clutch size more variable and largely independent of egg size; (2) egg size more variable and largely independent of clutch size; (3) a trade-off between egg size and clutch size; (4) a simultaneous increase in egg size and in clutch size; and (5) no relationship between the egg size or clutch size, with neither appreciably more variable (Fig. 5a,b). The scaled residuals (for each female) for egg mass and those for clutch size were essentially uncorrelated (Fig. 5a; G, r 2 = 0 02, F = 0 54, df = 1,29, P = 0 47; Y, r 2 = 0 02, F = 0 39, df = 1,20, P = 0 54; and C, r 2 = 0 005, F = 0 07, df = 1,15, P = 0 79). Thus, when the effect of female mass was removed, egg mass and clutch size varied essentially independently of each other, showing neither a trade-off nor a parallel increase (i.e. showed very flat slopes and very weak correlations that strongly support the null). The scaled residuals of clutch size on CL were more dispersed than were those for egg width at each site (Fig. 5b; Wilcoxon signed ranks test, G, z = 4 7, P < 0 0001; Y, z = 4 0, P < 0 0001; C, z = 3 6, P < 0 0005) and were more dispersed than were those for egg mass (Fig. 5a) at Grassyfork (z = 3 3, P < 0 002) and Crane (z = 3 1, P < 0 002) but not at Yellowwood (z = 1 2, P > 0 2). Treated parametrically, the variances of the scaled residuals for clutch size (G, s 2 = 0 33; Y, s 2 = 0 49; C, s 2 = 1 17) were significantly larger than those for

76 P. J. Clark et al. Fig. 5. Scattergrams for scaled residuals of egg size vs clutch size following least-squares adjustment for body size: (a) comparing egg mass vs female spent mass; (b) comparing egg width vs carapace length. Lines 1 4 represent various axes defining hypothetical data. See text for explanation and statistical detail (Grassyfork, triangles; Yellowwood, circles; Crane, crosses). egg mass (G, s 2 = 0 05; Y, s 2 = 0 05; C, s 2 = 0 02) at each site (test of homogeneity of variances: P < 0 05 for each). Thus, each of the two size-corrected measures of egg size (width and mass) was, or tended to be, less variable than that for clutch size. Conclusion In at least some populations of S. odoratus (especially the one at Yellowwood), as in some of C. picta (specifically, those studied by Congdon & Gibbons 1990; and by DePari 1988), many small females cannot lay appreciably larger eggs without them breaking or becoming lodged in the reproductive tract. For these examples, we agree with Congdon & Gibbons (1990) that natural selection, by setting a minimum for egg width, may be setting a minimum for the breadth of the reproductive canal and, thus, for female body size at initial maturity. Moreover, selection acting on small-bodied populations may influence egg width for the species on a regional basis, if not as a whole. Thus, in both S. odoratus and C. picta, egg sizes in populations where female morphology appears non-constraining are quite similar to those in populations where morphology is constraining. For further study, we suggest examining reproductively mature, young turtles of many species in habitats offering poor growth potential and in smaller-bodied geographical variants. In our study, if at each site the female size for which the regression line of caudal gap on CL crosses that for egg width on CL can be taken as indicative of reproductive maturity, then females of S. odoratus from southern Indiana should mature at 77 82 mm CL. Plastral flexion would allow a slightly smaller adult size. Our smallest reproductive female in 1992 was 74 mm CL and the smallest we ever encountered from these sites was 69 7 mm CL (both from Yellowwood). Given the evidence for a convergence of the caudal gap width and egg width in S. odoratus and for a much lesser constraint from pelvic aperture, we also recommend that X-ray measurements of pelvic apertures and measurements of egg sizes be supplemented with direct measurements of the caudal gaps in evaluating egg size body size relationships across different species and populations of turtles. These additional data, too, will help evaluate the extent of physical constraints on egg size within populations, and, thence, the extent to which egg sizes from diverse habitats throughout a region correspond to those observed in the most physically constrained populations. Acknowledgements We thank Lynn Andrews at Crane Naval Weapons Support Center and the management of Grassyfork Goldfish Hatchery for permitting the collection of turtles (Indiana Department of Natural Resources, Scientific Collecting Permit no. 714). We also thank Ron and Molly Clark for assistance with the collection and care of turtles, the College Mall Animal Clinic for use of their X-ray equipment, and Lynda Delph, John Phillips and Donald Whitehead for valuable input. This research was funded in part by grants from the Indiana Academy of Science and the Department of Biology, Indiana University. References Bernardo, J. (1996) The particular maternal effect of propagule size, especially egg size: patterns, models, quality of evidence and interpretations. American Zoologist 36, 216 236.

77 Physical apertures as constraints on eggs Bramble, D.M., Hutchinson, J.H. & Legler, J.M. (1984) Kinosternid shell kinesis: structure, function and evolution. Copeia 84, 456 475. Chen, T.-H. & Lue, K.-Y. (1999) Population characteristics and egg production of the yellow-marginated box turtle, Cuora flavomarginata flavomarginata, in northern Taiwan. Herpetologica 55, 487 498. Clark, P.J. (2000) Reproductive strategies of turtles: population, latitudinal, and phylogenetic comparisons. PhD thesis, Indiana University, Bloomington, IN. Congdon, J.D. & Gibbons, J.W. (1987) Morphological constraint on egg size: a challenge to optimal egg size theory? Proceeding of the National Academy of Sciences of the USA 84, 4145 4147. Congdon, J.D. & Gibbons, J.W. (1990) Turtle eggs: their ecology and evolution. Life History and Ecology of the Slider Turtle (ed. J. W. Gibbons), pp. 109 123. Smithsonian Institution Press, Washington, DC. DePari, J.A. (1988) Overwintering in the nest chamber by hatchling painted turtles, Chrysemys picta, in Northern New Jersey. PhD thesis, Rutgers University, Newark. Elgar, M.A. & Heaphy, L.J. (1989) Covariation between clutch size, egg weight and egg shape: comparative evidence from chelonians. Journal of Zoology, London 219, 137 152. Ewert, M.A. (1985) Embryology of turtles. Biology of the Reptilia Vol. 14 (eds C. Gans, F. Billett & P. F. A. Maderson, pp. 73 267. Wiley-Interscience, New York. Ewert, M.A. & Legler, J.M. (1978) Hormonal induction of oviposition in turtles. Herpetologica 34, 314 318. Germano, D.J. & Bury, R.B. (1998) Age determination in turtles: evidence from annual deposition of scute rings. Chelonian Conservation and Biology 3, 123 132. Gibbons, J.W. & Greene, J.L. (1979) X-ray photography: a technique to determine reproductive patterns of freshwater turtles. Herpetologica 35, 86 89. Graham, T.E. & Petokas, P.J. (1989) Correction for magnification when taking measurements directly from radiographs. Herpetological Review 20, 46 47. Halliday, T.R. & Verrell, P.A. (1988) Body size and age in amphibians and reptiles. Journal of Herpetology 22, 253 265. Hinton, T.G., Fedderman, P.D., Lovich, J.E., Congdon, J.D. & Gibbons, J.W. (1997) Radiographic determination of fecundity: is the technique safe for developing turtle embryos? Chelonian Conservation and Biology 2, 409 414. Iverson, J.B. (1991) Life history and demography of the yellow mud turtle, Kinosternon flavescens. Herpetologica 47, 373 395. Iverson, J.B. & Ewert, M.A. (1991) Physical characteristics of reptilian eggs and a comparison with avian eggs. Egg Incubation: Its Effects on Embryonic Development in Birds and Reptiles (eds D. C. Deeming & M. W. J. Ferguson), pp. 87 100. Cambridge University Press, Cambridge. Iverson, J.B. & Smith, G.R. (1993) Reproductive ecology of the painted turtle (Chrysemys picta) in the Nebraska Sandhills and across its range. Copeia 1993, 1 21. Kuchling, G. (1999) The Reproductive Biology of the Chelonia. Springer-Verlag, Berlin. van Loben Sels, R.C., Congdon, J.D. & Austin, J.T. (1997) Life history and ecology of the Sonoran mud turtle: a preliminary report. Chelonian Conservation and Biology 2, 338 344. Long, D.R. & Rose, F.L. (1989) Pelvic girdle relationships in three turtle species. Journal of Herpetology 23, 315 318. Nieuwolt-Dacanay, P.M. (1997) Reproduction in the western box turtle, Terrapene ornata luteola. Copeia 1997, 819 826. Roosenburg, W.M. & Dunham, A.E. (1997) Allocation of reproductive output: egg and clutch size variation in the diamondback terrapin. Copeia 1997, 290 297. Rose, F.L. & Judd, F.W. (1991) Egg size versus carapaceplastron aperture size in Gopherus berlandieri. Journal of Herpetology 25, 248 250. Rowe, J.W. (1994) Egg size and shape variation within and among Nebraskan painted turtle (Chrysemys picta bellii) populations: relationships to clutch and maternal body size. Copeia 1994, 1034 1040. Rowe, J.W. (1997) Growth rate, body size, sexual dimorphism and morphometric variation in four populations of painted turtle (Chrysemys picta bellii) from Nebraska. American Midland Naturalist 138, 174 188. Smith, C.C. & Fretwell, S.D. (1974) The optimal balance between size and number of offspring. American Naturalist 108, 499 506. Tucker, J.K. & Janzen, F.J. (1998) Order of oviposition and egg size in the red-eared slider turtle (Trachemys scripta elegans). Canadian Journal of Zoology 76, 3777 3380. Received 11 February 2000; revised 26 July 2000; accepted 1 August 2000