investment by green turtles nesting on Ascension Island

1098 Reprductive investment by green turtles nesting n Ascensin Island GRAEME C. HAYSJ Department f Zlgy, Universitv f Aberdeen, Aberdeen, AB9 2TN, United Kingdm COUN R. ADAMS Department f Aquaculture, University f Stirling, Stirling, SK9 4LA, United Kingdm JOHN AND R. SPEAKMAN Department f Zlgy, University f Aberdeen, Aberdeen, AB9 2TN, United Kingdm Received Octber 15, 1992 Accepted February 10, 1993 HAYS, G.c., ADAMS, C.R., and SPEAKMAN,J.R. 1993. Reprductive investment by green turtles nesting n Ascensin Island. Can. 1. Zl. 71: 1098-1103. utch sizes and egg sizes were measured fr green turtles (Chelnia mydas) nesting n Ascensin Island in the Suth Atlantic (7 57'S, 14 22'W) in 1992, The mean number f eggs per clutch was 127.5 (SD = 22.9; n = 46 clutches). The number f eggs per clutch increased in larger turtles and declined as the nesting seasn prgressed. Mean egg size was 45.5 mm (SD = 1.45 mm; n = 47 clutches). Three turtles laid significantly smaller eggs than the rest f the sample. When these three clutches were remved frm the analysis, mean egg size increased in larger turtles. When the effect f female bdy size was remved there was n relatinship between the number and size f the eggs in a clutch. The depth frm the tp t the bttm f the egg chamber, the depth frm the tp f the egg chamber t the tpmst egg, and the depth f the bdy pit were all independent f adult size. Egg size varied systematically within clutches, the largest eggs being laid first and the smallest eggs last. This intraclutch variatin in egg size had imprtant cnsequences fr the calculated gradient f the relatinship between egg size and adult size. HAYs, G.c., ADAMS, C.R., et SPEAKMAN,1.R. 1993, Reprductive investment by green turtles nesting n Ascensin lsland. Can. 1. Z!. 71: 1098-1103. Le nmbre d'eufs par cuvee et la taille des eufs nt ete mesures chez des Trtues venes (Chelnia mydas) au curs de la saisn de nidificatin de 1992. sur lile de lascensin, dans le sud de latlantique (7 57'S, 14 22'0). Le nmbre myen d'eufs par prtee a ete evalue a 127,5 (ecart type = 22,9; n = 46 prtees). Le nmbre deufs par prtee augmentait chez les grsses trrues et diminuait a mesure que la saisn de pnte avancait. La taille myenne dun euf etait de 45,5 mm (ecart type = 1,45; n = 47 prtees). Tris trrues nt pndu des eufs significativernent plus petits que les autres trrues de lechantilln. Apres eliminatin de ces tris prtees dans les analyses. la taille myenne des eufs sest averee plus grande chez les grsses trrues. Apres crrectin pur elirniner l'effet de la taille des femelles, il ny avait plus de relatin entre le nmbre deufs et leur taille dans une cuvee. La prfndeur du haut jusquau fnd de lenceinte de pnte. la prfndeur du haut de lenceinte de pnte jusqua leuf sirue sur le dessus dans lenceinre et la prfndeur de la cavite ccupee par la trtue etaient rutes independanies de la taille de ladulte. La taille des eufs variait significativement au sein des cuvees : les eufs les plus grs etaient pndus les premiers et les eufs les plus petits. les derniers. Cene variatin intragrupe de la taille des eufs a une grande imprtance lrsqu'n calcule le gradient de la relatin entre la taille des eufs et la taille des adultes. [Traduit par la redactin] Intrductin There is cnsiderable interest 'in the ultimate factrs that influence egg size (Smith and Fretwell 1974; Brckelman 1975; Parker and Begn 1986). Mdels aiming t predict ptimal egg size assume that the survivrship f eggs is related t their size, with smaller eggs having a lwer chance f survival. Since the prbability f survival must have an upper limit the relatinship between egg survivrship and egg size cannt be linear but must be a curve f decreasing gradient as the upper asympttic limit f survivrship is apprached. Since there must als be a limit t the resurces an individual can partitin t reprductin, if individuals are perating at the limit f their resurce availability there may be a trade-ff between the number and size f eggs that a female can prduce. Hence, when female bdy size is held cnstant a negative relatinship between egg size and number is predicted. 'Present address: The Sir Alister Hardy Fundatin Fr Ocean Science, c/ Plymuth Marine Labratry, Prspect Place, The He, Plymuth, PLl 3DH, United Kingdm. This negative relatinship has been fund bth intraspecifically in viparus and viviparus snakes (Frd and Killibrew 1983; Frd and Seigel 1989), the yellw mud turtle (Kinstern n flavescens) (Iversn 1991), and lizards (Stewart 1979; Nussbaum 1981), and interspecifically acrss fishes (Elgar 1990), mammals (Read and Harvey 1989), birds (Rhwer 1988), and turtles (Elgar and Heaphy 1989). Several egg size mdels predict a single ptimal egg size that maximises the number f ffspring surviving; this ptimal size crrespnds t the pint where fr further increases in egg size the marginal increase in survival is ffset by the marginal decrease in the ttal number f eggs prduced (Smith and Fretwell 1974; Brckelman 1975; Parker and Begn 1986). If the relatinship between egg size and survivrship is cnstant, bth spatially and temprally, these mdels predict a spatitemprally stable ptimal egg size and, hence, changes in ttal investment are expected t be achieved by changes in egg number. If, hwever, the nature f the relatinship between egg size and egg survivrship varies, in space and (r) time, there may be spati-ternpral variability in the ptimal egg Printed in Canada I Imprime au Canada I

I... HAYS ET AL. 1099 size. leading t mre cmplex trade-ffs in the apprtining f ttal investment between egg size and clutch size (McGinley et al. 1987). Fr lggerhead sea turtles (Caretta carenai nesting in Greece, egg size is independent f adult size, but changes seasnally, while egg number is psitively related t adult size (Hays and Speakrnan 1991, 1992). These data therefre suggest a temprally changing ptimal egg size, with larger turtles increasing their investment per clutch, at anyne time, by increasing clutch size rather than egg size. Similarly, fr sme species f freshwater turtle (e.g., Pseudemys jlridana and Sterntherus dratusi, increases in clutch vlume in larger females are mediated thrugh increases in egg number and nt egg size (Cngdn and Gibbns 1985). In sme ther species f freshwater turtle (e.g., Kinsternn [lavescens, K. subrubrum, Deirchelys reticularia, Chrysemys picta, and Pseudemys scripta) bth egg number and egg size increase in larger females (Cngdn and Gibbns 1985, 1987; Cngdn et al. 1983. 1987; Iversn 1991). Such differences in egg size between individual females at a given time are nt anticipated frm the simple mdels f ptimal egg size and require explanatin. In tw f these species (C picra and D. reticul.aria) the mean egg diameter per clutch has been shwn t increase with adult size with the same gradient as pelvic width. This has led t the suggestin that in these species. pelvic width may be a cnstraint n egg size; that is. small turtle with a narrwer pelvis lay smaller eggs than thse that ptimise the number f ffspring surviving (Cngdn and Gibbns 1987). There wuld therefre appear t be a dichtmy between species f turtles in which egg size is independent f the adult size (suggesting that egg size may be set at an ptimum) and species in which egg size may be cnstrained by pelvic width. Hwever. the situatin is cmplicated further by ther species in which egg size increases with adult size but where the gradient f this relatinship is significantly less than that between pelvic width and adult size (Cngdn and Gibbns 1987: Iversn 1991). Fr example. in the yellw mud turtle ikinstemn jlavescens) mean egg diameter per clutch increases with adult size with a gradient f 0.119, whereas pelvic width increases with adult size with a significantly higher gradient, O. J 68 (Iversn 1991), Similarly, in Pseudemys scripta mean egg diameter per clutch increases with adult size with a gradient f 0.026. while pelvic width increases with adult size with a significantly higher gradient f 0.096 (Cngdn and Gibbns 1987). In these previus studies examining the relatinship between egg size and female size, mean egg size fr a clutch has been used as a measure f egg size (Cngdn and Gibbns 1987; Iversn 1991). Hwever, if there is systematic intraclutch variatin in egg size, this may influence the interpretatin f egg size - adult size relatinships. In this paper we examine the relatinships between egg number, egg size, and adult size fr green sea turtles (Chelnia mydas) nesting n Ascensin Island in the Suth Atlantic (7 57'S, 14 22'W). We investigate whether a single ptimal egg size exists and whether there is a trade-ff between egg size and egg number. We examine intrac1utch variability in egg size and the effects f this variability n the calculated gradient f the relatinship between egg size and adult size. We als quantify the relatinship between adult size and nest depth t see whether systematic variatin in nest depth dependent n adult size may cause different ptimal egg sizes fr different sized adults. Methds STudy species and STudy site The ppulatin f green turtles n Ascensin Island has been the subject f spradic research ver the last 30 years. There are apprximately 5000-10000 nests n the island each year, an estimated average f three nests in a seasn fr each female; the females return t lay successive clutches with a mean interval f 13.9 days (Mrtimer and Carr 1987). Green turtles usually nest at night, first digging a bdy pit with all fur flippers befre excavating an egg chamber using nly their rear flippers. The green turtle rkery n Ascensin Island is an ideal lcatin at which t make detailed measurements f the variatin f egg size with egg sequence within a clutch. Firstly, this is because the nesting density is very high and, hence, it is pssible t encunter sufficient turtles at the start f egg laying. Secndly, in green turtles, unlike sme ther species such as the lggerhead turtle, the psterir f the carapace des nt usually extend ver the whle f the egg chamber, s it is pssible t bserve the eggs as they are laid and t recrd egg number and size within a clutch. Prtcl During February, March. and April 1992, Lng Beach, ne f the majr nesting beaches n Ascensin Island. was patrlled at night t lcate nesting turtles. Turtles were first bserved at a distance f apprximately 20 m thrugh 8 x 50 binculars t ascertain their rientatin and their stage f nesting. Turtles were then apprached frm behind and watched clsely until they started t lay their eggs. When there were tw bservers in the field we then bth \i) cunted the eggs as they were laid. and (ii) measured the maximum diameter (t an accuracy f 0.1 mm) f apprximately every 10th egg (the first measurement usually being made f abut the 20th t 30th egg laid) using dial callipers (Rabne). We fund it impssible fr a single bserver t bth cunt and measure the eggs as they were laid. Therefre, when there was nly ne bserver either (i) the ttal number f eggs was cunted. r (ii) the diameter f apprximately 10 eggs was measured thrughut egg laying. T maximise precisin the dial callipers were zered prir t each egg measurement. and t verify their accuracy they were crss-calibrated with vernier callipers (Carnlab). Fr a sample f nests we used a semi-rigid tape measure (Stanley Tls) 10 recrd the depth f the bdy pit, the depth frm the tp t the bttm f the egg chamber. and the depth frm the tp f the egg chamber t the tpmst egg. We defined the tp f the egg chamber (and thus als the bttm f the bdy pit) as the psterir tip f the pstcentral scute at the carapace midline. When measuring the bdy pit depth we defined the beach surface by placing an apprximately 3 m lng ruler ver the nest with the ruler resting n the undisturbed beach surface n either side. Bdy pit depth was then measured as the vertical distance frm the ruler t the tp f the egg chamber. We measured the depth f the egg chamber when excavatin was cmplete, the depth f the tpmst egg when egg laying had stpped, and the depth f the bdy pit during egg laying. When egg laying was cmplete and the turtle had started t cver the nest we measured the curve carapace length (CCL) frm the anterir f the precentral scute t the psterir f the pstcentral scute using a flexible 1.5-m tape. T remve any variatin caused by different individuals making measurements, all egg diameters and adult lengths were measured by ne persn (G.c.H.). Data were analysed using Minitab Statistical Sftware (Minitab Inc.). Least squared fit regressins are used thrughut. Results The number f eggs laid was recrded fr 46 clutches. The mean number f eggs per clutch was 127.5 (SD = 22.9, range = 83-170). utch size increased in larger turtles and decreased during the seasn. Of the ttal variatin in the number f eggs laid per clutch, 14.5 % was explained by the size

1100 CA". J. ZOOL. VOL. 71. 1993 50... 48 0>- (j)e EE 46 cu--- :Om Ol 44 COl CUO> 0>. :20 42 40 105 110 115 120 eel (cm) [] 125 130 FIG. 1. Relatinship between mean egg diameter and the curve carapace length (Cf.L. cm) f the adult turtle. Ignring the three data pints marked as pen squares, mean egg diameter was 47.74 mm (SD = 1.13: /1 = 44 clutches) and increased significantly in larger turtles: mean egg diameter (mm) = 0.102 eel (cm) + '33.8 (fj1.42) = 11.8, P < 0.01, r2 = 0.22). Nne f the residual variatin in egg size was explained by the date n which the clutch was laid (stepwise regressin, p > 0.05). The three clutches with small eggs were significantly smaller than predicted n the basis f the adult length (Z = 5.3,4.3.2.8, p <ii 0.01). f the turtle and 13.9% by the date. Defining the date as the number f days since March 1, the cmplete equatin was [1] number f eggs per clutch = 1.9 C - 0.95 date - 36.3 (Fi2.43) = 8.6. p < 0.01, r2 = 0.28) The depth f the bdy pit averaged 33.2 cm (SD = 11.1; n = 83), the depth frm the tp t the bttm f the egg chamber averaged 58.5 cm (SO = 2.2; n = 48) and the depth frm the tp f the egg chamber t the tpmst egg averaged 32.0 cm (SO = 6.2; n = 63). All three measures were independent f turtle size (C) (Fi1.46) = 2.5, Fil.61) = 1.0, and Fil.81) = 2.1, P > 0.05). ' Egg diameters were recrded fr 47 clutches. Mean egg diameter was 45.5 mm (SO = 1.45 mm; ranse = 41.0-48.5 mm: n = 47 clutches). When three clutches with very small eggs were remved frm the analysis (i.e., a restricted sample f 44 clutches was used), mean egg size increased in larger adults but did nt change during the seasn (stepwise regressin, p > 0.05) (Fig. I). We calculated the deviatin (Z scre) f the three clutches with anmalusly small eggs frm the regressin equatin relating egg size t adult size fr the remaining 44 clutches (Fig. 1). Fr each f these three clutches with small eggs, Z > 2.8, i.e.. the prbability f getting even ne clutch with eggs this small frm the bserved distributin f increasing egg size with adult size was less than ne in a hundred. Bth the ttal number f eggs and mean egg diameter were recrded fr 24 clutches. Using the egg number and mean egg diameter predicted n the basis f adult length we calculated the residuals (actual value minus predicted value) f egg number and mean egg diameter fr these 24 clutches. There was n relatinship between the residuals f egg number and mean egg size (F[I.22] = 0.04, p > 0.05). Fr the 22 clutches fr which bth the diameter and the number f that egg within the clutch were recrded, egg diameter was nt independent f the number f the egg. Instead Ē OlE _ 0>... 0>0> >-.- 0> re -cu 0>.- 0::'0-1 '2~--~-'--~--,---~~--~~ 40 80 120 160 Egg number frm start f clutch FIG. 2. Relatinship between the relative egg size (egg size minus mean egg size fr that clutch) and the number f the egg within the clutch (/1 = 22 clutches). Each value is the mean fr a 1O-egg interval. (Egg number 1 = the first egg laid.) -- larger eggs tended t be laid at the start f a clutch and smaller eggs at the end. Thus, eggs that were amng the first 75 t be laid had diameters, n average, slightly «0.5 mm) larger than the mean size fr the clutch, while the l40th t 150th eggs had diameters, n average, > 1.6 mm smaller than the mean diameter fr the clutch (Fig. 2). the decline in egg size between the start f a clutch (the mean diameter f the 20th t 40th eggs laid) and the end f a clutch (the mean diameter f the last 20 eggs laid) averaged 1.21 mm (SE = 0.25; n = 22 clutches). This decline in egg size within clutches was entered int a stepwise regressin against the number f eggs laid in the clutch, the size f the turtle (C), and the mean egg diameter f the 20th t 40th eggs laid. The nly factr that entered the resulting regressin equatin was the number f eggs laid. the largest decline in egg size being fund in the largest clutches (Fig. 3). Nne f the residual variatin in the intraclutch decline in egg size was explained by the date n which each nest was laid (stepwise regressin, p > 0.05).. Fr the 22 clutches fr which we had cmplete data n egg diameters and egg number within a clutch, we investisated hw the decline in egg size within a clutch influ'~nced the gradient f the relatinship between mean egg size per clutch and adult size. Firstly. we calculated this gradient when nly the last 40 eggs laid in a clutch were used in the analysis. the last 50 eggs laid, the last 60 eggs, and s n. Secndly, we calculated this gradient using nly the first 20 eggs laid in a clutch, the first 30, and s n, until all the measured eggs were included in the analysis. The gradient f the relatinship between mean egg diameter per clutch and adult size was lw «0.07) when nly eggs laid at the end f a clutch were included in the analysis, high (> 0.135) when nly eggs laid at the start f a clutch were included, and O. J08 when all the measured eggs were included (Fig. 4). Discussin In a previus study f clutch sizes fr green turtles n Ascensin Island, Simn and Parkes (1976) reprted that the mean number f eggs per clutch was 116.3 in 1973 (n = 169 nests) and 127.9 in 1974 (n = 163 nests). These figures are clse t the value we recrded (127.5 eggs; n = 46 nests). As in ur study, Simn and Parkes (1976) fund that there was a significant psitive relatinship between the number f j /

> HAYS ET AL. 1101 Q)..-.. ce '~g enq) C'ON Q).- ten Q) 5 4 3 2-1+-----~----~----.-----.-----. 80 100 120 140 160 180 Number f eggs laid FIG. 3. Decline in egg size between the start f a clutch (mean diameter f the 20th t 40th eggs laid) and the end f a clutch (mean diameter f the last 20 eggs laid) plned against the ttal number f eggs laid in the clutch. Egg decline (mm) = 0.0334 (number f eggs laid) - 2.83 (Fjul = 13.0, P < 0.01, r2 = 0.39). eggs in a clutch and the size f the adult turtle. By using numbered flipper tags t allw individual turtles t be identified, Mrtirner and Can (1987) shwed that the number f eggs per clutch tended t decrease as a turtle laid successive clutches. This is cnsistent with the seasnal decline in the number f eggs per clutch that we fund (eq. I), since turtles wuld have been laying successive clutches as the seasn prgressed. It has been suggested that the larger clutches laid by larger turtles may be accmmdated by larger turtles laying deeper nests (Bjrndal and Carr 1989). The relatinship between adult size and egg chamber depth has nt, hwever. been previusly quantified. We fund that the depth t the bttm f the egg chamber shwed little variatin and was independent f the size f the turtle. The depth t the tpmst egg in the egg chamber was mre variable but was again independent f turtle size. Thus. larger turtles did nt dig deeper nests. nr did they fill their nests fuller with eggs. Hwever. larger turtles lay mre eggs (this study; Simn and Parkes 1976) and this wuld suggest that these larger clutches are accmmdated by increasing the width f the nest. The mean egg size fr several spatially separated green turtle ppulatins has been reviewed by Hirth (1980). Means include 45.7 mm at Trtuguer, Csta Rica: 45.0 mm at Suriname; 44.7 mm at Eurpa Island, Indian Ocean: 44.6 mm at Trmelin Island, Indian Ocean; 46.3 mm at Aldabra Atll; 42.3 mm at Sharma Beach, Suth Yemen; 46.0 mm at Hern Island. Australia: and 44.0 mm at Hawaii. The variatin in the mean egg size fr different green turtle rkeries may indicate spatial variability in the relatinship between egg size and survivrship and, hence, selectin fr different ptimal egg sizes. Whereas the maximum reprted mean egg diameter fr a ppulatin away frm Ascensin Island is 46.3 mm fr green turtles nesting n AJdabra Atll, the previusly reprted mean egg diameter fr green turtles n Ascensin Island was 54.6 mm (n = 10 nests, with JO eggs measured in each nest) (Carr and Hirth 1962), which is substantially larger than the mean egg diameter reprted here (45.5 mm) and fr ther green turtle rkeries. Yet while the mean egg size measured by Carr and Hirth is > 8 mm larger than that fr any ther ppulatin, the mean hatchling size they measured is in the middle f the range reprted fr varius green turtle ppulatins (reviewed by Hirth 1980). Our callipers were zered 0.16... 0 C Q) 0.14 0 0-0 C'O... 0.12 '0 Q)... 0.10 C'O :l 0 0.08 C'O 0.06 20 40 60 80 100 120 140 160 FIG. 4. Gradient f the relatinship Egg number between egg diameter and adult size when different parts f clutches were included in the analysis.., maximum egg number frm the end f the clutch that was included in the analysis (ME). Gradient = 0.0809 lg(me) - 0.0636 (FjI.IOI = 317, P < 0.01, r2 = 0.97). 0, maximum egg number frm the start f the clutch that was included in the analysis (MS). Gradient = (142176-0.OO0047(MS)4) x 10-6 (Fjl.l21 = 48.5. P < 0.01, r2 = 0.80). The maximum measured egg number frm the end f a clutch was 144 (4th egg laid frm a clutch f 148) while the maximum measured egg number frm the start f a clutch was 159 (l59th egg laid frm a clutch f 161 eggs). Thus, the final pint n each plt represents all the measured eggs used in the analysis. hence the calculated gradients fr these tw pints are the same (0.108). The regressin equatins frm which the gradients were calculated were all significant at p < 0.05. befre every measurement, and crss-calibrated. hence we are cnfident in bth their accuracy and their precisin, whereas Carr and Hirth give n details cncerning hw their measurements were made. The discrepancy with Carr and Hirth ' s previus measurement remains unreslved. but may reflect either a real tempral change in the mean egg size r simply a methdlgical difference between the tw studies. Eggs in 3 f a ttal f 48 clutches had signi ficantly smaller mean diameters (Fig. I). Egg size in turtles typically shws a cnsistent pattern within ppulatins. either increasing r being independent f adult size (Cngdn and Gibbn, 1985: Hays and Speakrnan 1991). The finding that three turt les laid significantly smaller eggs than the rest f the sample was therefre unexpected. These clutches with small egus may represent reprductive senescence. The turtles prducing the small eggs were at the large end f the size distributin. Hwever, marine turtles grw nly very slwly, r nt at all. nce they have reached sexual maturity (e.g.. 0.2 cm per year fr lggerhead turtles in Australia (Limpus 1985), and green turtles in the Indian Ocean shw n significant increase (Le Gall et a!. 1985)). Adult size is therefre prbably nt influenced primarily by age but by the size at sexual maturity. Thus, the large size f the three turtles that laid small eggs des nt necessarily mean that they were lder than the rest f the sample. A secnd ptential explanatin fr these bservatins is that these three clutches were laid by turtles that had mved t Ascensin Island frm anther rkery characterised by turtles that laid smaller eggs. This hypthesis wuld require a majr breakdwn in nesting beach fidelity by these turtles, since the nearest alternative nesting beaches are > 2000 km frm Ascensin Island. Such mvement between nesting beaches has been dcumented. Fr example, a green turtle in the I

1102 CAI'. J. ZOOL. VOL. 71. 1993 Indian Ocean nested n beaches 2250 km apart (Le Gall and Hughes 1987). Further evidence fr pr nesting beach fidelity in sme turtles is prvided by bservatins that every year green turtles attempt t nest n Saint Helena (15 55'S, 5 43'W), an islated Atlantic Island 1100 km sutheast f Ascensin Island, despite the fact that the rcky castline and lack f sandy beaches preclude successful nest excavatin and egg laying (Edwards 1990). If there are discrete subppulatins f turtles nesting n Ascensin Island that represent turtles frm ther rkeries, this may be reflected in genetic hetergeneity in the turtles. Bwen et al. (1992) sequenced the mitchndrial DNA f hatchlings frm 35 different nests n Ascensin Island and fund that 34 f the 35 nests were characterised by ne haplphyte, while ne nest was characterised by a secnd haplphyte, supprting the suggestin that this ppulatin may be genetically hetergeneus. Hw these genetic differences relate t differences in egg size between individuals remains t be investigated. Bth egg size and egg number increased in clutches laid by larger turtles n Ascensin Island, while egg size' decreased mntnically frm the start t the end f clutches. This wuld suggest that if there is a single ptimal egg size fr this ppulatin, it is attained neither by all individuals nr fr all the eggs laid by the same individual. Similarly, fr green turtles nesting in Csta Rica, bth egg number per clutch and mean egg diameter increased psitively with adult carapace length (Bjrndal and Carr 1989). This pattern fr green turtles cntrasts with that fund fr lggerhead turtles, fr which increases in reprductive investment per clutch in larger turtles were assciated nly with increases in the number and nt in the size f eggs (Hays and Speakman 1991), but is similar t that reprted fr sme freshwater turtles fr which increases in bth egg number and egg size ccur in larger individuals (Cngdn and Gibbns 1985, 1987; Iversn 1991). Parallel gradients f the relatinships between (i) egg size and adult size and (ii) pelvic width and adult size suggest that in sme species f freshwater turtle, egg size is cnstrained by the width f the pelvic canal and hence individuals cannt achieve the ptimal egg size (Cngdn and Gibbns 1987). This cnstraint may als be perating in green turtles, althugh there are n data n the relatinship between pelvic width and adult size fr this species with which t test this hypthesis. Variability in egg size in relatin t adult size amng individuals des nt necessarily reflect cnstraint by pelvic width. Fr example. if nest depth were dependent upn the size f the turtle and ptimal egg size increased psitively with depth, ne wuld anticipate an increase in ptimal egg size with turtle size. In the current ppulatin we can reject this interpretatin, since there was n relatinship between nest depth and adult size. There was n inverse relatinship between the residual number f eggs in a clutch and the residual mean egg diameter, and hence n evidence fr the trade-ff between egg size and egg number predicted by egg size ptimality mdels (e.g., Smith and Fretwell 1974; Parker and Begn 1986). The absence f this trade-ff suggests that ttal clutch vlume fr green turtles nesting n Ascensin Island is nt generally perating at the limit f an individual's resurce availability (e.g., bdy cavity vlume). Further evidence in supprt f this hypthesis is fund in the large differences in egg number per clutch previusly reprted fr nests n Ascensin Island laid by the same individuals within a seasn. Thus fr individuals recrded nesting 3 r mre times, the number f eggs in the smallest clutch averaged nly 70.6% (range = 40.6-91.4%; n = 12 turtles) f the number in the largest clutch (Mrtirner and Carr 1987). The decrease in egg size within a clutch may have an adaptive explanatin, e.g., in terms f differing physicchemical characteristics within the nest. Alternatively, the decrease in egg size within a clutch may simply reflect a cnstraint assciated with the physilgy f egg prductin, marine turtles being unable t prduce an entire clutch f eggs f the same size. There has been n cnsideratin f hw the gradient f the relatinship between egg and adult size is affected by intraclutch variability in egg size. We fund that this gradient was. significantly influenced by which eggs within a clutch were sampled (Fig. 4). Hence, when cmparisns are made with the gradient f the relatinship between adult size and pelvic width, it may be mre insightful t select that part f the clutch where the gradient between egg and adult size is maximal (e.g., the start f each clutch). If nly the mean egg diameter fr an entire clutch is cnsidered, the gradient f the relatinship between egg size and adult length may be underestimated. Acknwledgements We thank the Administratr f Ascensin Island, Mr. Brian Cnnelly, fr permitting this wrk t be cnducted. We are indebted t Cable and Wireless, the Ryal Air Frce, and the Office f the Administratr fr prviding accmmdatin n Ascensin Island, and t Steve and Tina Cx and Sam Turtle fr all their help when we were cnducting fieldwrk. G.C.H. was supprted by grant GR9/585 frm the Natural Envirnmental Research Cuncil. We thank Jeanne Mrtimer and Brian Bwen fr stimulating discussins cncerning this wrk and three annymus reviewers fr their cnstructive cmments. Bjrndal, K.A., and Carr, A. 1989. Variatin in clutch size and egg size in the green turtle nesting ppulatin at Trruguerc. C la Rica. Herpetlgica. 45: 181-189. Bwen. B.W., Meylan, A.B.. Rss. 1.P.. Limpus. C.L.. Balazs. G.H.. and Avise, 1.C. 1992. Glbal ppulatin structure and natural histry f the green turtle (Chelnia mydas) in terms f matriarchal phylgeny. Evlutin (Lawrence, Kans.), 46' 865-881. Brckelman, W.Y. 1975. Cmpetitin. the fitness f ffsp mg, and ptimal clutch size. Am. Nat, 109: 677-699. Carr. A.. and Hirth, H. 1962. The eclgy and migratir s f sea turtles: 5 cmparative features f islated green turtle clnies. Am. Mus. Nvit. N. 2091. Cngdn. 1.0., and Gibbns. 1.W. 1985. Egg cmpnents and reprductive characteristics f turtles: relatinships t bdy size. Herpetlgica, 41: 194-205. Cngdn, 1.0., and Gibbns, LW. 1987. Mrphlgical cnstraint n egg size: a challenge t ptimal egg size thery') Prc. Natl. Acad. Sci. U.S.A. 84: 4145-4147. Cngdn, 1.0., Gibbns, 1.W., and Green. 1.L. 1983. Parental investment in the chicken turtle ideirchelys reticulariay. Eclgy, 64: 419-425. Cngdn, 1.0., Breitenbach, G.L., van Lben Sels, R.C., and Tinkle, O.W. 1987. Reprductin and nesting eclgy f snapping turtles (Chelydra serpenrina) in sutheastern Michigan. Herpetlgica. 43: 39-54. Edwards, A. 1990. Fish and fisheries f Saint Helena Island. Centre fr Trpical Castal Management Studies, University f Newcastle upn Tyne, Newcastle upn Tyne. U.K. Elgar, M.A. 1990. Evlutinary cmprmise between a few large and many small eggs: cmparative evidence in telest fish. Oiks, 59: 283-287. Elgar, M.A., and Heaphy, L.l. 1989. Cvariatin between clutch /

HAYS ET AL. 1103 size. egg weight and egg shape: cmparative evidence fr chelnians. J. Zl. (Lndn), 219: 137-152. Frd. N.B.. and Killibrew, D.K. 1983. Reprductive tactics and female bdy size in Butler's garter snake, Thamnphis butleri. J. Herpetl. 17: 271-275. Frd, N.B., and Seigel, R.A. 1989. Relatinships amng bdy size, clutch size, and egg size in three species f viparus snakes. Herpetlgica, 45: 75-83. Hays, G.C., and Speakrnan, J.R. 1991. Reprductive investment and ptimum clutch size f lggerhead sea turtles (Caretta carettav. J. Anim. Ecl. 60: 455-462. Hays, G.c., and Speakrnan, J.R. 1992. utch size fr Mediterranean lggerhead turtles (Caretta carettai. J. Z 1. (Lndn), 226: 321-327. Hirth, H.F. 1980. Sme aspects f the nesting behavir and reprductive bilgy f sea turtles. Am. Zl. 20: 507-523. lversn, J.B. 1991. Life histry and demgraphy f the yellw mud turtle, Kinsternn [lavescens. Herpetlgica, 47: 373-395. Le Gall, J.-Y., and Hughes, G.R. 1987. Migratins de la trtue vene Chelnia mydas dans l'ocean Indien Sud-Ouest bservees a partir des marquages sur les sites de pnte Eurpa et Trmelin (1970-1980). Amphib.-Reptilia, 8: 272-282. Le Gall, J.-Y., Chateau, D., and Bsc, P. 1985. Rhythme de reprductin interannuel des TOl1Uesvertes Chelnia mydas sur les sites de pnte Trmelin et Eurpa (Ocean Indien). C.R. Acad. Sci. Paris, tme 301, Ser. rn,s: 195-196. Limpus, c.l. 1985. A study f the lggerhead sea turtle. Caretta caretta, in eastern Australia. Ph.D. thesis. University f Queensland, SI. Lucia, Australia. McGinley, M.A., Temme, D.H., and Geber, M.A. 1987. Parental investment in ffspring in variable envirnments: theretical and empirical cnsideratins. Am. Nat. 130: 370-398. Mrtimer, J.A., and Carr, A. 1987. Reprductin and migratins f the Ascensin Island green turtle (Chelnia mydas). Cpeia, 1987: 103-113. Nussbaum, R.A. 1981. Seasnal shifts in clutch size and egg size in the side-bltched lizard, Uta stansburiana Baird and Girard. Oeclgia, 49: 8-13. Parker, G.A., and Begn, M. 1986. Optimal egg siz.eand clutch size: effects f envirnment and maternal phentype. Am. Nat. 128: 573-592. Read, A.F., and Harvey, P.H. 1989. Life histry differences amng the eutherian radiatins. J. Z I. (Lndn), 219: 329-353. Rhwer, F.C. 1988. Inter- and intraspecific relatinships between egg size and clutch size in waterfwl. Auk, 105: 161-176. Sirnn, M.H., and Parkes, A.S. 1976. The green turtle (Chelnia mydas): nesting n Ascensin Island, 1973-1974. J. Zl. (Lndn), 179: 153-163. Smith, c.c., and Fretwell, S.D. 1974. The ptimal balance between size and number f ffspring. Am. Nat. 108: 499-506. Stewart, J. R. 1979. The balance between number and size f yung in the live-bearing lizard Gerrhntus ceruleus. Herpetlgica, 35: 342-350.