Journal of Herpetology, Vol. 38, No. 1, pp. 1–8, 2004 Copyright 2004 Society for the Study of Amphibians and

Cross-Breeding of Distinct Color Morphs of the Strawberry Poison Frog (Dendrobates pumilio) from the Bocas del Toro Archipelago,

1,2,3 4 5 K. SUMMERS, T. W. CRONIN, AND T. KENNEDY

1Department of Biology, East Carolina University, Greenville, North Carolina 27858, USA 2Smithsonian Tropical Research Institute, Apartado 2072, Balboa, Republic of Panama 4Department of Biology, University of Maryland at Baltimore County, Hilltop Circle, Baltimore, Maryland 21250, USA 5Department of Biology, McGill University, Montreal, Quebec H3A 2K6, Canada

ABSTRACT.—Populations ascribed to Dendrobates pumilio, the Strawberry Poison Frog, show extreme variation in color and color pattern among island and mainland locations in the Bocas del Toro Archipelago in Panama. Previous analyses indicate that these different populations are probably members of a single . Here we present data on crosses between several different color and color pattern morphs. Successful crosses were made between different morphs from seven populations: Bocas Island, Nancy Cay, Pope Island, , and Almirante, and the Aguacate Peninsula on the mainland. The resulting offspring were characterized for color and color pattern. Our study indicates that different color morphs can interbreed to produce viable offspring. The offspring typically displayed a mixture of colors but always showed color pattern if one parent showed color pattern. This suggests that color pattern is under single locus control with dominance, whereas coloration may be under polygenic control, or may represent a single locus system with incomplete dominance.

One of the most dramatic examples of color a preliminary investigation of the inheritance of and color pattern polymorphism within a puta- color and color pattern in these populations. tive single species occurs in the Bocas del Toro Color polymorphisms are found in many differ- Archipelago, in Panama (Daly and Myers, 1967; ent species of anurans (Hoffman and Blouin, Myers and Daly, 1983). Populations ascribed to 2000). Color in anurans typically results from Dendrobates pumilio, the Strawberry Poison Frog, pigments contained in chromatophores, whereas display extreme variation in color and color color pattern (hereafter referred to as ‘‘melanistic pattern among island and mainland locations in pattern’’) is usually caused by melanin contained this archipelago (Daly and Myers, 1967; Myers in melanophores (Hoffman and Blouin, 2000). and Daly, 1983). Recent analysis of the spectral The inheritance of color or melanistic pattern reflectances of individuals from different popu- (or both) has been investigated in a number of lations quantitatively documents the extent of the species of anurans (Hoffman and Blouin, 2000). divergence among these populations (Summers In most cases, evidence from single generation et al., 2003). In spite of these extreme differences crosses suggests that melanistic pattern is con- in color and color pattern among populations, trolled by simple Mendelian single-locus inher- evidence from both acoustic analysis of call itance (e.g., Lantz, 1947; Goin, 1960; Browder et similarity (Myers and Daly, 1976) and genetic al., 1966). Similar investigations of the inheri- analysis of DNA sequence similarity (Summers tance of color have been done in a smaller et al., 1997) suggests that the different popula- number of cases. In some studies, inheritance of tions are closely related and probably members color appears to be under single locus control of the same species. However, the most impor- with a dominant and recessive allele (e.g., tant criterion for determining species status Fogelman et al., 1980; Blouin, 1989), whereas in under the biological species concept is ability to other cases, multiple loci are involved (e.g., interbreed and produce viable offspring (Mayr, Mattews and Pettus, 1966). Multiple generation 1942). One goal of this study was to demonstrate studies using F2 or backcrosses (which are that different color and color pattern morphs can necessary to conclusively determine modes of interbreed and produce viable offspring. inheritance) have been done with only a few Another goal of this study was to make species (Hoffman and Blouin, 2000). Studies of the inheritance of color and mela- nistic pattern provide basic information impor- 3 Corresponding Author. E-mail: summersk@mail. tant to investigations of the evolutionary forces ecu.edu producing and maintaining color and melanistic 2 K. SUMMERS ET AL. pattern polymorphisms in anurans, which are PVC pipe, and then feed the tadpoles infertile currently poorly understood (Hoffman and eggs for approximately six weeks (Weygoldt, Blouin, 2000). Both the genetic control of color 1980; Brust, 1993). Some offspring were aban- and melanistic pattern, and the evolutionary doned by their mother, and we attempted to feed mechanisms that have generated the diversity of these tadpoles with chicken egg yolk and frogs color and melanistic pattern among populations eggs gathered from the field. However, none of of D. pumilio, are unknown. In fact, the genetic these tadpoles survived to metamorphosis. basis of color and melanistic pattern polymor- The six offspring that survived transport to the phisms in tropical frogs in general is poorly United States were taken to the University of understood, and relatively few studies have been Maryland at Baltimore County, and their spectral done on tropical species (Hoffman and Blouin, reflectance was measured. The spectral reflec- 2000). tance of a sample of adults from each of the Several of the different Bocas del Toro color populations used in the cross-breeding experi- morphs have been bred in captivity at the United ments was taken in Panama (before the pro- States National Aquarium in Baltimore. These duction of the offspring), to characterize the morphs breed true under uniform environments parental color morphs for each cross. Measure- (R. Gagliardo, pers. comm.), suggesting that ments of spectral reflectance were made with color and melanistic pattern are under genetic a portable spectrometer (Ocean Optics 2000), control. We used cross-breeding between differ- attached to a portable computer (BiLink), at the ent color and melanistic pattern morphs and STRI field laboratory on Bocas Island. Adult spectrometric analysis of color, to investigate the frogs in the laboratory were illuminated with inheritance of color and melanistic pattern in a single halogen lamp (250 watts). The spectrom- these populations. eter was calibrated before measuring each in- dividual, using a white diffuse reflectance MATERIALS AND METHODS standard (Spectralon). Measurements were taken Cross-breeding experiments were carried out with either a 50 lm 3 2 m or 400 lm 3 2 m optic at the Smithsonian Tropical Research Institute’s fiber. Three measurements in separate locations (STRI) research station on Bocas Island, Bocas del were taken from the dorsum of each frog, and Toro, Panama, and took place from 15 May to 7 three from the venter. The offspring from crosses September 2000. Adult frogs were caught in the were measured in the laboratory at the Univer- field in the Bocas del Toro Archipelago. Research sity of Maryland at Baltimore County, using the and collection permits were obtained from the same type of spectrometer used in the field, and Panamanian authorities (ANAM), via the Smith- ambient illumination. For both adults and off- sonian Tropical Research Institute, before the spring, the source of illumination provided onset of collecting. ample light across the full spectrum during all Frogs were caught by hand and kept in plastic reflectance measurements. bags until they were measured. The sex and population of the morphs used, the number of RESULTS pairs maintained, and the total number of The majority of pairs did not produce surviv- offspring produced are shown in Table 1. ing offspring, and only two pairs produced more Terrariums were set up at the Smithsonian than a single surviving offspring. Approximately Tropical Research Institute’s Bocas del Toro half of the offspring produced (i.e., observed as Research Station, on Bocas Island. Each of a total embryos or tadpoles) did not survive to meta- of 35 terraria was prepared for the cross-breeding morphosis. A total of 16 offspring survived experiments by placing a single pair of frogs from through metamorphosis during the four months distinct populations in the terrarium. Each of the study. Unfortunately, 10 of these juveniles terrarium was also provided with leaf litter, died in transit or as a result of stresses induced a bromeliad, a leafy plant, two pools of water during transport from Panama to the United (one in a plastic cup and one in a six-inch piece of States. Three juveniles died in the United States, PVC pipe), and a piece of bamboo as a hiding and three survived the juvenile period in place. The frogs were fed with fruit flies captivity. (Drosophila) grown in cultures or collected from Color was determined for the 16 juveniles that the field. Flies were coated with vitamin powder survived beyond metamorphosis (Table 1). The before being placed in the terrariums. Terrariums offspring showed distinctive color and pattern were misted with water daily to keep the even before metamorphosis. The color of juve- humidity level high. niles in the field appears identical to that of Eggs were laid on leaves (typically bromeliad adults, indicating that there is little change in leaves), and usually reached the tadpole stage in coloration during development from the juvenile approximately two weeks. Females would carry to the adult stage (K. Summers, pers. obs.). their tadpoles to pools in the bromeliads or in the Visually, color appears to remain the same over COLOR INHERITANCE IN PANAMANIAN POISON FROGS 3

TABLE 1. Collection locations of parentals used for the crossing experiments, and color and pattern descriptions for each parental type and the offspring.

Male Color Female Color Fig. Ref. Offspring (number) Bocas Green Dorsum, Nancy Orange Dorsum Fig. 1, top Bronze, Island Spotted, Cay Orange/White Fig. 2, middle Spotted (4) Yellow Venter Venter Bocas Green Dorsum, Nancy Orange Dorsum Fig. 2, bottom Greenish-yellow, Spotted (1) Island Spotted, Cay Orange/White Yellow Venter Venter Nancy Orange Dorsum, Bocas Green Dorsum, Fig. 2, top Bronze, Cay Orange/White Island Spotted Spotted (5) Venter Rambala Yellow, Black Nancy Orange Dorsum, Yellowish- Stripes Cay White Venter orange, Black Stripes (1) Bocas Green Dorsum, Almirante Red Dorsum and Fig. 1, middle Red-Bronze Island Spotted, Venter, Blue Fig. 3, top Dorsum, Yellow Venter Legs Spotted, Bronze Venter (1) Aguacate Blue Dorsum Almirante Red Dorsum Fig. 1, bottom Red Dorsum, Peninsula and Venter and Venter, Fig. 3, middle Blue Venter Blue Legs and Legs (2) Pope Green Dorsum Almirante Red Dorsum Red-Bronze Island and Venter and Venter, Dorsum and Blue Legs Venter (1) Bastimentos Red Dorsum, Almirante Red Dorsum Fig. 3, bottom Red Dorsum, Island Spotted, and Venter, Spotted, White Venter Blue Legs Blue Legs (1) the juvenile period and into adulthood in tances revealed considerable variation among the dendrobatid frogs in captivity (Walls, 1994). offspring, but all of them showed substantial With respect to melanistic pattern, in all cases reflectance in both the green and orange regions in which one of the parental morphs in a cross of the spectrum. showed melanistic pattern, the offspring also Offspring from crosses between other popula- showed melanistic pattern (Table 1, Fig. 1). tions also appeared to be a mixture of both the Offspring from crosses between the Bocas Island parental colors, with one exception. The offspring green-spotted morph with the Nancy Cay orange from the cross between a Bocas Island (spotted morph generally looked similar, with some green) male and Almirante (red with blue legs) variation. The typical offspring from crosses female showed a red-bronze color (Fig. 1, middle between these populations was a bronze orange right), with substantial reflectance in both the color, with spotting on the dorsum, as illustrated red-orange and the green-yellow regions of the by a photo of the cross between a Bocas Island spectrum from the dorsal and ventral surfaces male and a Nancy Cay female (Fig. 1, top right). (Fig. 3, top right). The cross between the Rambala The spectral reflectance measurements from the (striped yellow) male and an Almirante female dorsal and ventral surfaces of this offspring are produced a yellowish-orange offspring with shown in Figure 2 (middle right). black stripes. The offspring from a cross between The color and melanistic pattern displayed by a Pope Island (green) male and an Almirante offspring of crosses in either direction (Bocas female showed a red-bronze dorsum and venter. Island male with Nancy Cay female, or the In contrast, the two offspring from the cross reverse) appeared similar to the offspring in between the Aguacate Peninsula (blue) male and Figure 1 (top right), with some appearing slightly an Almirante female had red dorsal surfaces (Fig. more green and some more orange. Figure 2 1, bottom right), although the extent of red shows the spectral reflectances measured from coloration appeared to be reduced relative to several crosses between orange Nancy Cay that of the Almirante parent. The spectral morphs and green-spotted Bocas Island morphs. reflectances measured from one of these off- The most divergent offspring (Table 1; Fig. 2, spring (Fig. 3, middle right) reveal a pronounced bottom right) was greenish-yellow with spots, peak in the blue region for the ventral surfaces, with a slight orange tinge. The spectral reflec- like the paternal blue morph from the Aguacate 4 K. SUMMERS ET AL.

FIG. 1. Photos of some of the morphs of Dendrobates pumilio used in cross-breeding experiments. From left to right (top row): Bocas Island, Nancy Cay, offspring (second row): Bocas Island, Almirante (mainland), offspring (third row): Aguacate Peninsula, Almirante (mainland), offspring.

Peninsula, and unlike the maternal morph from transportation, and the difficulty of adapting to Almirante. The dorsal surface shows a profile dramatically new environments (particularly by similar to the Almirante red parent, but the levels juveniles). High mortality from long distance of reflectance in the red region are reduced travel is common even for adults. We think that relative to the Almirante parent. inferences concerning the relative viabilities of mixed versus single population offspring will require long-term experiments in a single loca- DISCUSSION tion with controlled comparisons of between and Our breeding experiments demonstrated that within population offspring production. different color morphs of D. pumilio can produce It is possible that the color of offspring from viable offspring. Although this result was not crosses between different morphs would itself unexpected, it provides further evidence in favor lower the survivorship of those offspring. In of the hypothesis that these dramatically differ- many cases, the coloration of the offspring ent color morphs are all members of a single appeared less saturated and less intense than species. Given the low survival rates of embryos that of the parentals, and this could impose a cost to metamorphosis, and the low survival rates of in terms of predation attempts in the wild. In the juveniles, it is tempting to speculate that turn, this could provide a mechanism favoring offspring from the between-morph crosses may genetic isolation between morphs if they come have poor health and vigor relative to ‘‘normal’’ into contact (as some may on the mainland). offspring. We do not think that such an inference Evaluating this hypothesis will require experi- is justified. Low survival to metamorphosis is ments on attack rates by predators. very common in captive D. pumilio (Heselhaus, The offspring from crosses between Bocas 1992). The low survival of the juveniles is not Island and Nancy Cay parentals appeared mixed surprising, given the rigors of long-distance in color. The spectral reflectances of these off- COLOR INHERITANCE IN PANAMANIAN POISON FROGS 5

FIG. 2. Spectral reflectances for parental morphs and their offspring, from crosses between Bocas Island and Nancy Cay Dendrobates pumilio. Each panel shows the percent spectral reflectance as a function of wavelength in nanometers, ranging from 350–750 nanometers. The top right panel shows a ‘‘typical morph’’ for the Nancy Cay parental morph, and the panels down the left side show the typical morph for the Bocas Island parental morph. Each panel on the right shows the spectral reflectance for one of the offspring. Three measurements were taken of the dorsal (thick lines) and the ventral (thin lines) surface of each offspring. 6 K. SUMMERS ET AL.

FIG. 3. Spectral reflectances for parental morphs and the offspring from crosses between several different morphs of Dendrobates pumilio from the Bocas del Toro Archipelago. Panels are set up as in Figure 2. spring also indicate a mixture of the reflectances orange part of the spectrum, respectively (Fig. 2). of the parental morphs. For example, the parental The offspring show substantial reflectance in types from Bocas Island and Nancy Cay each both the green and orange part of the spectrum, produce clearly distinct peaks in the green and indicating that they are producing both green COLOR INHERITANCE IN PANAMANIAN POISON FROGS 7 and orange pigments. They also show some color in D. pumilio may be similar. reflectance in parts of the spectrum where the parental morphs do not. The offspring from Acknowledgments.—We thank the staff of the crosses between other color morphs (with the Smithsonian Tropical Research Institute for exception of the cross between the Aguacate support with all parts of the project, especially blue and Almirante red morphs) also displayed L. Moe and L. and B. Ferenbach. Special thanks a mixture of colors, possibly due to their ability to M. Leone for assistance with permits. We to express both pigment sets. thank the Ministry of the Environment in Previous studies suggest that color and mela- Panama (ANAM) for granting research, col- nistic pattern inheritance in frogs is typically lection and exportation permits for this re- under simple genetic control, involving single search. We thank S. Rand for advice on the loci (e.g., Fogelman et al., 1980), although project. This project was funded by grants from evidence for the influence of multiple loci has the National Geographic Society (grant number been found in some cases (e.g., Matthews and 6702-00), and the National Science Foundation Pettus, 1966). In this study, the fact that the (grant number IBN-9724028). offspring from the cross between green and orange morphs of D. pumilio displayed a mixture of parental colors is consistent with the possibil- ity that these colors are under polygenic control, LITERATURE CITED although incomplete dominance with single loci BLOUIN, M. S. 1989. Inheritance of a naturally occurring is an equally likely alternative. Testing between color polymorphism in the ornate chorus frog, these two hypotheses of genetic control will Pseudacris ornata. Copeia 1989:1056–1059. require further crosses and genetic analysis, BROWDER, L. 1968. Pigmentation in Rana pipiens.I. particularly backcrosses and F2 crosses (Hoffman Inheritance of the speckled mutation. Journal of and Blouin, 2000). Heredity 59:162–166. Offspring from crosses between morphs con- BROWDER, M. W., J. C. UNDERHILL, AND J. C. MERRELL. sistently have melanistic patterns if one parent 1966. Mid-dorsal stripe in the wood frog. Journal of has them. This suggests that this trait may be Heredity 57:65–67. BRUST, D. 1993. Maternal brood care by female under simple genetic control, involving a single Dendrobates pumilio, a frog that feeds its young. locus and dominance of the melanistic pattern Journal of Herpetology 27:96–98. causing allele. Simple genetic control of melanis- DALY,J.W.,AND C. W. MYERS. 1967. Toxicity of tic pattern has been found in most other anurans Panamanian poison frogs (Dendrobates): some bi- in which it has been investigated (Hoffman and ological and chemical aspects. Science 156:970–973. Blouin, 2000). The fact that similar color and FOGELMAN, J., P. CORN, AND D. PETTUS. 1980. The genetic melanistic pattern was displayed by offspring of basis of dorsal color polymorphism in Rana pipiens. crosses in either direction (Bocas male with Journal of Heredity 71:439–440. Nancy Cay female, or the reverse), suggests an GOIN,C.J.1960.Patternvariationinthefrog Eleutherodactylus nubicola Dunn. Bulletin of the absence of sex linkage for genes affecting both State Museum of Biological Sciences color and melanistic pattern. 5:243–258. The Aguacate Peninsula (blue) by Almirante HESELHAUS, R. 1992. Poison-Arrow Frogs: Their Natural (red with blue legs) produced offspring with red History and Care in Captivity. Blandford, London. dorsal color, and the spectral reflectance curve of HOFFMAN, E., AND M. BLOUIN. 2000. A review of colour the red color of the offspring was similar to that and pattern polymorphisms in anurans. Biological of the female parent. However, the extent of red Journal of the Linnaen Society 70:633–665. coloration on the offspring (which had blue LANTZ, L. A. 1947. Note (appendix to H. M. Bruce and venters) was reduced relative to that of the red A. S. Parkes, Observations of Discoglossus pictus Otth). Proceedings of the Royal Society of London parent (which had a red venter). The loss of red Series B Biological Sciences 134:52–58. coloration may be caused by a simple mutation MAYR, E. 1942. Systematics and the Origin of Species. causing loss of function in an enzymatic or Columbia Univ. Press, New York. regulatory pathway, as suggested by the dis- MATHEWS,T.C.,AND D. PETTUS. 1966. Color inheri- covery of completely blue mutants in Costa tance in Pseudacris triseriata. Herpetologica 22: Rican populations of D. pumilio, which are nor- 269–275. mally red with blue legs (M. Donnelly, pers. MYERS,C.W.,AND J. W. DALY. 1976. Preliminary comm.). evaluation of skin toxins and vocalization in At this point, it is unclear whether the blue taxonomic and evolutionary studies of Poison-Dart Frogs (Dendrobatidae). Bulletin of American Mu- color is structurally produced or produced by seum of Natural History 157:173–262. a blue pigment. In ranids, blue color is produced ———. 1983. Poison Dart Frogs. Scientific American by iridescent pigment cells called guanophores 248:120–133. that scatter the shorter wavelengths of light that SUMMERS, K., E. BERMINGHAM,L.WEIGHT,S.MCCAFFERTY, reach them (Browder 1968). The basis of the blue AND L. DAHLSTROM. 1997. Phenotypic and mito- 8 K. SUMMERS ET AL.

chondrial DNA divergence in three species of WALLS, J. G. 1994. Jewels of the Rainforest: Poison Frogs Poison-Dart Frogs with contrasting parental care of the Rainforest. TFH Publications, Neptune, NJ. behavior. Journal of Heredity 88:8–1. WEYGOLDT, P. 1980. Complex brood care and reproduc- tive behaviour in captive Poison-Arrow Frogs, SUMMERS, K., T. W. CRONIN, AND T. KENNEDY. 2003. Color Dendrobates pumilio O. Schmidt. Behavioral Ecology and pattern diversity in Dendrobates pumilio in the Sociobiology 7:349–332. Bocas del Toro Archipelago, Panama. Journal of Biogeography 30:35–53. Accepted: 4 July 2003.

Journal of Herpetology, Vol. 38, No. 1, pp. 8–15, 2004 Copyright 2004 Society for the Study of Amphibians and Reptiles

Sexual Size Dimorphism in the Red Hills Salamander, Phaeognathus hubrichti (Caudata: Plethodontidae: Desmognathinae)

1 KRISTIN A. BAKKEGARD AND CRAIG GUYER

Department of Biological Sciences, Auburn University, Alabama 36849-5407, USA

ABSTRACT.—Body size is an important variable used in life-history and sexual selection theory to predict reproductive, behavioral, and ecological traits. Except for the presence of special skin glands in males, has not been reported in the Red Hills Salamander (Phaeognathus hubrichti), the basal member of the Desmognathinae. These data provide insight into the evolution of SSD (sexual size dimorphism) in the entire subfamily. We conducted multivariate and univariate tests on eight morphological measurements of 92 preserved P. hubrichti. We also examined specimens for broken tails and U-shaped scars, which may indicate bites from conspecifics. Male salamanders were larger than females in all measurements except tail length and had more scars than either females or juveniles. This species exhibited male-biased SSD in shape and size: males were broad and bulky, females were long and thin. Regression analysis showed differences in resource allocation between male and female salamanders. Differences in life-history strategies and sexual selection in the form of male-male combat may explain these differences in body size.

Sexual size dimorphism (SSD) occurs when- Species within Plethodontidae have one of ever adult male and female have three SSD patterns: female biased (females larger different mean body sizes. Determining the than males), male biased (males larger than cause(s) of SSD in any taxon is challenging females), or unbiased (both sexes the same size; because life-history strategies, sexual selection, Bruce, 2000). Female bias is the most common and ecological differences between the sexes can pattern, especially in the Plethodon (Bruce, all contribute to variation in body size (Slatkin, 2000). However, members of subfamily Desmog- 1984; Halliday and Verrell, 1986; Kozlowski, nathinae (Dusky Salamanders) have a special 1989; Shine, 1990; Andersson, 1994). Selection form of male-biased SSD. Males mature at a can also occur at different life-history stages. smaller size but grow to a larger size than females However, juvenile amphibians generally grow as quickly as possible because once maturity is (Bruce, 1993). Thus, SSD produces larger variance reached, resources shift from growth to pro- in male snout–vent length, except in two dwarf duction of gametes. If there is no selection on species (Desmognathus wrighti and Desmognathus a particular body part, it should continue on its aeneus) in which the degree and direction of SSD juvenile growth trajectory. Thus, the juvenile varies among populations (Bruce, 2000). body plan is a null hypothesis against which The most basal member of the Desmognathi- changes in adult size and shape can be compared. nae (Wake, 1966), the monotypic Phaeognathus We used the Red Hills Salamander (Phaeognathus hubrichti has not been examined for SSD. hubrichti Highton) to explore this idea. Phaeognathus hubrichti is sexually dimorphic; males have large ‘‘hedonic’’ glands concentrated 1 Corresponding Author. Present address: Depart- on the tail and dorsum, and their vent walls are ment of Biology, Utah State University, Logan, Utah papillate; females lack these glands and have 84322-5305, USA; E-mail: [email protected] plicate vent walls (Jordan, 1975; Mount, 1975). SEXUAL SIZE DIMORPHISM IN PHAEOGNATHUS 9

Although morphological and DNA sequence Schwaner and Mount, 1970). Individuals data clearly indicate it is desmognathine (Wake, , 81 mm SVL were classified as juvenile, as 1966; Titus and Larson, 1996), life-history traits were individuals less than 100 mm and lacking of P. hubr ichti are different than the other pigmented glands. Juvenile salamanders were members of the subfamily (Tilley and Bernardo, not referred to either sex. Using body size as 1993). Phaeognathus has large yolky eggs, an a criterion to separate mature and juvenile female entirely terrestrial life cycle and small clutch size, salamanders may have resulted in the misclassi- life-history traits more similar to most Plethodon fication of some small females as juveniles. The than to Desmognathus (Brandon, 1965; Marvin, result of this conservative approach is that the 1996). Based on these similarities, one might mean measurements for females may be slightly predict that Phaeognathus would exhibit female- inflated, thus reducing the differences between biased SSD. However, as in most of the other male and female measurements. We examined desmognathines, Phaeognathus may exhibit male- salamanders for broken tails and semicircular or biased SSD, which would suggest that it is the U-shaped scars that appeared to conform to tooth ancestral trait for the Desmognathinae. Thus, patterns of P. hubrichti. Salamanders with multi- identifying the SSD pattern in P. hubrichti is ple scars in the same area were scored as a single important for understanding the evolution of event. We scored salamanders with tails broken this trait in this group. Here, we assess sexual after or during collection (skin ragged at point of dimorphism in P. hubrichti and interpret the break, no sign of healing [Wake and Dresner, results in light of existing theory. 1967]) only for scars. Scars have been used as an indicator of conspecific aggression in other salamanders (Staub, 1993; Camp, 1996). MATERIALS AND METHODS We radiographed all tails to look for breaks We collected morphological data from 92 pre- not apparent by visual inspection. An indepen- served P. hubrichti (listed in Appendix 1; institu- dent observer knowledgeable in vertebrate tional abbreviations as in Leviton et al. [1985]). morphology also scored tails as complete or We recorded morphological measurements to the broken to confirm our observations. We three nearest 0.1 mm using dial calipers: snout–vent disagreed on one tail, which we examined length (SVL; tip of snout to posterior edge of together and determined was complete. We also vent), overall length (OAL; tip of snout to noted the direction of attack from the orientation posterior tip of tail), head width (HW, taken of bite marks on salamanders’ bodies. For this immediately posterior to the commissure of the analysis, any salamander greater than 81 mm jaw), head length (HL, snout to gular fold), and SVL was classified as an adult and sexed tail diameter (TD, at cloaca, lateral plane). We according to the presence or absence of glands. measured mass to the nearest 0.01 g using an We chose this size because it was the smallest at electronic balance, blotting salamanders dry with which pigmented glands were visible on males absorbent toweling before weighing. We calcu- and therefore may be the size at which sala- lated tail length (TL) by subtracting SVL from manders recognize each other as potential rivals. OAL and tail volume (TV) by using the formula 2 We used a multiresponse permutation pro- for a cone [1/3*p*(TD/2) *TL]. Because of vari- able specimen condition, we could not collect cedure (MRPP) to conduct a multivariate com- every measurement for every salamander. Only parison of adult body sizes using SVL, HW, HL, TL, TD, and mass. It is analogous to Hotelling’s salamanders with a complete body and tail were 2 used in calculations involving mass, overall T with the advantage of being distribution free length, tail length, and tail volume. Data for (Cade and Richards, 2001) and was performed USNM 142486 were taken from Valentine (1963). using BLOSSOM (vers. 2001-8d, Midcontinent The SVLs for AUM 11115, 11224, 11491, 11430, Ecological Science Center, Fort Collins, CO). All and 11092 were taken from Sever and Trauth other statistical analyses were performed using (1990). SAS (vers. 8.01, SAS Institute, Inc., Cary, NC). We placed salamanders into one of two All variables were natural log (ln) transformed categories based on SVL (adult or juvenile), and when we regressed mass on OAL or TV on SVL. one of three age/sex categories, based on the P-values less than 0.05 were considered signifi- presence or absence of tiny, pigmented glands on cant for all statistical tests. the tail and abdomen, which only males have To explore how differences in size and shape (Jordan, 1975). Salamanders . 81 mm SVL and of adults may have developed, we used ANCO- possessing pigmented glands were classified as VA to test for different slopes of juveniles versus male; those . 100 mm SVL but lacking pigmen- adults (separately for each sex) for the indepen- ted glands were classified as female. Phaeognathus dent variables HW, HL, TL, and TD on SVL, ln hubrichti matures at these sizes (Brandon, 1965; TV on ln SVL, and ln mass on ln OAL. We also 10 K. A. BAKKEGARD AND C. GUYER

FIG. 1. Size and shape comparison between mature male and female Phaeognathus hubrichti. These individuals are the same OAL. Male is on the top, female is underneath. used the juvenile regression equations for each for that measurement. Positive values, where the morphological measurement to calculate a pre- 95% confidence interval (C.I.) excluded zero, dicted value for male and female salamanders. indicate increased adult investment in that body Residuals (actual-predicted) were calculated for part (actual values larger than predicted); all data and then used in a one-sample t-test negative values revealed the converse. This with zero as the test value. Any statistically approach allowed us to detect subtle changes significant difference from zero indicates adult in the allocation of resources to different body salamanders deviated from the juvenile pattern parts. We used a G-test (log-likelihood ratio test)

TABLE 1. Morphological measurements of preserved specimens of Phaeognathus hubrichti. Mean 6 1 SD above range (in parentheses); N 5 sample size. * signifies males significantly larger (ANCOVA; P 0.01) for that measurement as determined by least-square means. ** signifies female significantly larger (Tukey’s test; P , 0.03; slopes differed significantly; Zar, 1984).

Males Females Juveniles SVL (mm) 107.2 6 13.5 109.1 6 5.5 75.3 6 20.4 (81.1–138.0) (100.4–120.8) (30.4–98.9) N 5 27 N 5 22 N 5 37 OAL (mm) 201.5 6 21.9 211.6 6 17.3 140.1 6 38.1 (161.0–253.9) (180.0–251.2) (53.8–191.6) N 5 18 N 5 14 N 5 27 Mass (g)* 10.1 6 4.3 8.9 6 2.8 3.6 6 2.1 (4.0–21.6) (6.1–13.8) (0.4–7.5) N 5 18 N 5 13 N 5 27 HW (mm)* 12.3 6 1.4 12.0 6 1.0 8.7 6 2.1 (9.8–16.4) (10.3–13.9) (4.6–11.6) N 5 28 N 5 26 N 5 36 HL (mm)* 20.2 6 2.6 19.9 6 1.1 14.2 6 3.6 (15.7–25.8) (17.7–21.9) (7.4–20.3) N 5 28 N 5 26 N 5 36 TL (mm)** 92.1 6 14.1 102.0 6 12.8 64.5 6 20.1 (68.0–115.9) (75.0–130.5) (20.3–95.0) N 5 18 N 5 14 N 5 27 TD (mm)* 7.4 6 1.2 7.0 6 0.9 4.8 6 1.6 (5.4–10.9) (5.4–8.6) (1.6–6.8) N 5 27 N 5 25 N 5 36 TV (mm3) 1411.5 6 712.4 1293.2 6 458.8 494.1 6 325.5 (612.8–3632.4) (816.9–2353.1) (27.6–1033.1) N 5 18 N 5 13 N 5 25 SEXUAL SIZE DIMORPHISM IN PHAEOGNATHUS 11

FIG. 2. Relationships between body size measurements of Phaeognathus hubrichti. Female (F) regression line represented by dashes and dots, male (M) by a dashed line, and juvenile (J) by a solid line. Open squares represent adult female salamanders, closed circles represent adult males, and open inverted triangles, juveniles. ANCOVA used to test for equality of slopes. ** indicates all lines have unequal slopes (F 5 6.22, df 5 2, P 5 0.004). *** indicates only female and juvenile regression lines with unequal slopes (TV on SVL: F 5 5.3, df 5 1, P 5 0.028; TD on SVL: F 5 4.45, df 5 1, P 5 0.040). All regressions were significant (P , 0.008) except for maleTL on SVL (F 5 2.3, df 5 17, P 5 0.149). 12 K. A. BAKKEGARD AND C. GUYER

TABLE 2. T-test compared residuals from adult TABLE 3. Damage to Phaeognathus hubrichti by age salamanders (juvenile regression line used to generate class and sex. the predicted value) with zero. C.I. indicates the confidence interval. Bite marks Tail loss YES NO YES NO Males Females Males 10 17 6 21 Regression P 95% C.I. P 95% C.I. Females 5 37 9 33 Mass on OAL 0.001 0.11 0.35 0.835 0.11 0.14 Juveniles 1 18 1 18 HW on SVL ,0.001 0.20 0.66 0.365 0.39 0.15 HL on SVL ,0.001 0.33 1.09 0.479 0.18 0.38 TL on SVL 0.081 14.2 0.92 0.241 2.28 8.28 TD on SVL 0.231 0.11 0.45 0.003 0.64 0.15 DISCUSSION TV on SVL 0.327 0.19 0.07 0.034 0.25 0.012 Phaeognathus hubrichti were sexually dimor- phic in shape and size. Females had longer tails, shorter bodies and were thinner than males, to determine whether there were differences in which had shorter tails, longer bodies and were the relative frequencies of individuals with scars wider. Using Bruce’s (2000) criteria, P. hubrichti or tail breaks among all age/sex groups. had male-biased dimorphism because males were significantly more variable in SVL than females. Thus, male-biased SSD is most likely RESULTS the ancestral condition in Desmognathinae. Male P. hubrichti were significantly larger than Small body size and no or possibly female- females in our six–dimensional parameter space biased SSD in the miniaturized D. wrighti and D. (MRPP standardized test statistic 5 3.19; P 5 aeneus appear to have evolved independently in 0.005; Fig. 1). Males had a larger maximum size those lineages. However, males were not larger than females for all measurements except for TL than females in all measurements; maximum TL (ANCOVA; SVL or OAL as the covariate; Table 1). of females was 11.2% longer than that of the Males had significantly more variance around the maximum male TL. Comparison of adult mor- mean for SVL and HL than females (Levene’s test; phology with that of juveniles suggests that F 5 8.46, df 5 48, P 5 0.006; F 5 10.6, df 5 53, males, at maturity, increase their investment in P 5 0.002 respectively; Table 1). Variances for all mass, HW, and HL. Adult females also increase other measurements were not statistically differ- investment in tail length as compared to ent. Male and female slopes differed significantly juveniles. Thus, as P. hubrichti mature, males for only TL on SVL (Fig. 2). and females exhibit different shape-related Males had greater mass, HW, and HL than changes in growth. predicted by regression of these variables on Male-male combat could explain investment juvenile OAL or SVL (Table 2). Female TD and by male P. hubrichti in larger heads and heavier TV were significantly smaller than predicted by bodies. Males had a greater proportion of bite juvenile regression. Male TL may be smaller (t 5 scars than females or juveniles. Male, but not 1.85, df 5 17, P 5 0.081) than that predicted by female, HW and HL were larger than predicted the juvenile model (Fig. 2., Table 2). This was the by the juvenile model. If selection acted only on only regression that was not significant (F 5 2.3, burrowing or feeding ability, then one would df 5 17, P 5 0.149). All others were significant expect no differences between the sexes or at P , 0.008. juvenile salamanders in head morphology. Scars were not distributed evenly among age/ Whether there is any relationship between SSD sex groups (G 5 8.29, P 5 0.016). Male and social behavior in amphibians is debated salamanders had significantly more scars than (Shine, 1979; Halliday and Verrell, 1986). None- female or juvenile salamanders (G 5 5.61, P 5 theless, Bruce’s (2000) review attributed male- 0.018; G 5 4.59, P 5 0.032, respectively), but biased SSD in Desmognathinae to male-male there was no significant difference in the number combat and aggression, a conclusion consistent of scars between female and juvenile salaman- with our findings. ders (G 5 0.62, P 5 0.432; Table 3). Salamanders Interestingly, we found no bite marks or had bite marks on the tail and body (with injuries to the heads of P. hubrichti. Most attacks approximately equal frequency) but never on the were directed to the side or top of a salamander, head (sample sizes were not large enough for as if one individual ambushed another from side a statistical analysis by body area). No salaman- tunnels. In contrast, a high percentage of marks ders had scars consistent with those expected were found on the head and tail of three species from predators. In contrast, broken tails were of desmognathines (Keen and Sharp, 1984; distributed homogeneously among age/sex Jaeger, 1988; Camp, 1996). Four agonistic-en- groups (G 5 2.28, P 5 0.320; Table 3). counter behaviors have been observed, in the SEXUAL SIZE DIMORPHISM IN PHAEOGNATHUS 13

field, between pairs of P. hubrichti at burrow et al., 1996) and eight specimens of P. hubrichti entrances (Bakkegard, 2002). In each, an intruder (Parham et al., 1996) showed no evidence that salamander traveled across the surface, entered males live longer than females. No data on the thefieldofviewofavideocameraand survival or dispersal of P. hubrichti are available. interacted with a resident salamander at a bur- Therefore, further study is required to determine row entrance. Every time, the resident repelled whether these demographic variables explain the intruder with a head butt or mouth gape. In SSD. a head butt, the resident advanced toward the The contribution of tail length to SSD in intruder, anchoring its hind legs in its burrow salamanders is relatively unstudied (but see entrance. The two salamanders faced off; then Park et al., 1996; Serra-Cobo et al., 2000). Instead, the resident swung its head and upper body so most analyses of SSD in the Plethodontidae have that its head impacted the side of the intruder’s focused on SVL (Bruce, 2000). Snout–vent length head. The intruder tumbled down the slope, out is the standard measure of body size in of view of the video camera. In gaping, the salamanders because they can lose their tails, resident retreated into its burrow (snout inside the entrance) once it became aware of the making OAL an inconsistent measure of body intruder and opened its mouth. After a face off size. However, because tail loss is detectable to (snouts almost touching) lasting about 1 min, the the trained observer (Martof and Rose, 1963), resident closed its mouth as the intruder fell off salamanders with damaged tails should be easy the slope. Thus, gaping or head–butting beha- to eliminate from any statistical analysis (but see viors may explain the lack of bites on salaman- Shaffer, 1978). Our results suggest comparative der’s heads. These observations, coupled with study of TL (relative to SVL) is needed, bite marks, provide the first evidence that P. especially of the subfamily Desmognathinae. hubrichti meet several criteria defining the re- The tail is an important organ, used for energy source-area competition concept of territoriality storage, antipredator defense, courtship and (Mathis et. al., 1995). Gaping has also been respiration (Fitzpatrick, 1976; Brodie, 1977; documented as an intraspecific threat display Maiorana, 1977; Houck, 1982). Additionally, tail among male Desmognathus ochrophaeus (Verrell volume may be a useful index of body condition and Donovan, 1991) (Fraser, 1980). We recommend including tail Female P. hubrichti were significantly less measurements in future studies of body size of variable in SVL than males but had similar plethodontid salamanders. variation in OAL and TL. Females appear to have an upper limit to SVL that is less than that Acknowledgments.—F. S. Dobson, D. Folkerts, of males but no limit on tail length. Hom’s (1988) R. Lishak, the Utah State University herpetology model of optimal reproductive allocation, based group, and two anonymous reviewers provided on D. ochrophaeus, suggests female salamanders helpful comments on the manuscript. A. Cutler have little prospect for further growth after and S. Durham provided statistical assistance. reaching maturity. Female salamanders may We thank D. B. Wake and N. L. Staub for invest in tail length because the tail stores fats a helpful discussion. These individuals loaned or needed for reproduction (Fitzpatrick, 1973; allowed us to examine specimens in their care: Maiorana, 1977). However, female P. hubrichti R. A. Brandon (SIUC), J. A. Campbell (UTA), C. had a smaller TD and less TV than predicted Lydeard (ALA), A. Resetar (FMNH), D. B. Wake from the juvenile model, presumably because of (MVZ), J. J. Wiens (CM), and R. V. Wilson and higher costs of reproduction (Fitzpatrick, 1973; W. R. Heyer (USNM). Marks and Houck, 1989). Additionally, TL in male salamanders was highly variable. Thus, differences in resource allocation may contribute LITERATURE CITED to SSD in P. hubrichti. ANDERSSON, M. 1994. Sexual Selection. Princeton Univ. Sexual size dimorphism in P. hubrichti may Press, Princeton, NJ. also be caused by differential survival of male BAKKEGARD, K. A. 2002. Activity patterns of Red Hills and female salamanders. In a study of five Salamanders (Phaeognathus hubrichti) at their bur- species of Desmognathus, Organ (1961) proposed row entrances. Copeia 2002:851–856. that one sex may grow larger than the other BRANDON, R. A. 1965. Morphological variation and simply because it lives longer. However, his ecology of the salamander Phaeognathus hubrichti. hypothesis was not adequately tested because Copeia 1965:67–71. he used age estimates based on number of BRODIE JR., E. D. 1977. Salamander antipredator testicular lobes, which is an unreliable method postures. Copeia 1977:523–535. (Houck, 1977). Skeletochronological estimates BRUCE, R. C. 1993. Sexual size dimorphism in Desmog- of age in three species of Desmognathus (Castanet nathine salamanders. Copeia 1993:313–318. 14 K. A. BAKKEGARD AND C. GUYER

———. 2000. Sexual size dimorphism in the Pletho- MARVIN, G. A. 1996. Life history and population dontidae. In R. C. Bruce, R. G. Jaeger, and L. D. characteristics of the salamander Plethodon kentucki Houck (eds.), The Biology of Plethodontid Sala- with a review of Plethodon life histories. American manders, pp. 243–260. Kluwer Academic/Plenum Midland Naturalist 136:385–400. Publishers, New York. MARTOF,B.S.,AND F. L. ROSE. 1963. Geographic CADE, B. S., AND J. D. RICHARDS. 2001. User Manual for variation in southern populations of Desmognathus BLOSSOM Statistical Software. Midcontinent Eco- ochrophaeus. American Midland Naturalist 69:376– logical Science Center, Biological Resources Di- 425. vision, U.S. Geological Survey, Fort Collins, CO. MATHIS, A., R. G. JAEGER,W.H.KEEN,P.K.DUCEY,S.C. CAMP, C. D. 1996. Bite scar patterns in the black-bellied WALLS, AND B. W. BUCHANAN. 1995. Aggression and salamander, Desmognathus quadramaculatus. Journal territoriality by salamanders and a comparison with of Herpetology 30:543–546. the territorial behavior of frogs. In H. Heatwole and CASTANET, J., H. FRANCILLON–VIEILLOT, AND R. C. BRUCE. B. K. Sullivan (eds.), Amphibian Biology. Vol. 2. 1996. Age estimation in Desmognathine salaman- Social Behaviour, pp. 634–676. Surrey Beatty and ders assessed by skeletochronology. Herpetologica Sons, ChippingNorton, New South Wales,Australia. 52:160–171. MOUNT, R. H. 1975. The Reptiles and Amphibians of FITZPATRICK, L. C. 1973. Energy allocation in the Alabama. Alabama Agricultural Experiment Sta- Allegheny Mountain Salamander, Desmognathus tion, Auburn. ochrophaeus. Ecological Monographs 43:43–58. ORGAN, J. A. 1961. Studies of the local distribution, life ———. 1976. Life history patterns of storage and history, and population dynamics of the salaman- utilization of lipids for energy in amphibians. der genus Desmognathus in Virginia. Ecological American Zoologist 16:725–732. Monographs 31:189–220. FRASER, D. F. 1980. On the environmental control of PARHAM, J. F., C. K. DODD JR., AND G. R. ZUG. 1996. oocyte maturation in a plethodontid salamander. Skeletochronological age estimates for the Red Hills Oecologia 46:302–307. Salamander, Phaeognathus hubrichti.Journalof HALLIDAY,T.R.,AND P. A. VERRELL.1986.Sexual Herpetology 30:401–404. selection and body size in amphibians. Herpeto- PARK, S., D. PARK, AND S. Y. YANG. 1996. Courtship, logical Journal 1:86–92. fighting behaviors and sexual dimorphism of the HOM, C. L. 1988. Optimal reproductive allocation in salamander, Hynobius leechii. Korean Journal of female dusky salamanders: a quantitative test. 39:437–446. American Naturalist 131:71–91. SCHWANER, T. D., AND R. H. MOUNT. 1970. Notes on the HOUCK, L. D. 1977. Life history patterns and re- distribution, habits and ecology of the salamander productive biology of Neotropical salamanders. In Phaeognathus hubrichti Highton. Copeia 1970: D. Taylor and S. I. Guttman (eds.), The Reproduc- 571–573. tive Biology of Amphibians, pp. 43–72. Plenum SERRA-COBO, J., J. F. UIBLEIN, AND J. P. MARTINEZ-RICA. Publ. Co., New York. 2000. Variation in sexual dimorphism between two ———. 1982. Male tail loss and courtship success in the populations of the Pyrenean salamander Euproctus plethodontid salamander Desmognathus ochro- asper from ecologically different mountain sites. phaeus. Journal of Herpetology 16:335–341. Belgian Journal of Zoology 130:39–45. JAEGER, R. G. 1988. A comparison of territorial and non- SEVER, D. M., AND S. E. TRAUTH. 1990. Cloacal anatomy territorial behaviour in two species of salamanders. of female salamanders of the plethodontid sub- Behaviour 36:307–310. family Desmognathinae (Amphibia: Urodela). JORDAN JR., J. R. 1975. Observations on the Natural Transactions of the American Microscopical Society History and Ecology of the Red Hills Salamander, 109:193–204. Phaeognathus hubrichti Highton (Caudata: Pletho- SHAFFER, H. B. 1978. Relative predation pressure on sala- dontidae). Unpubl. master’s thesis. Auburn Univ., manders (Caudata: Plethodontidae) along an altitu- Auburn, AL. dinal transect in . Copeia 1978:268–272. KEEN,W.H.,AND S. SHARP. 1984. Responses of SHINE, R. 1979. Sexual selection and sexual dimorphism a plethodontid salamander to conspecific and in the Amphibia. Copeia 1979:297–306. congeneric intruders. Animal Behaviour 32:58–65. ———. 1990. Proximate determinants of sexual differ- KOZLOWSKI, J. 1989. Sexual size dimorphism: a life ences in adult body size. American Naturalist history perspective. Oikos 54:253–255. 135:278–283. LEVITON, A. E., R. H. GIBBS JR., E. HEALS, AND C. E. SLATKIN, M. 1984. Ecological causes of sexual dimor- DAWSON.1985.Standardsinherpetologyand phism. Evolution 38:622–360. ichthyology. Part I. Standard symbolic codes for STAUB, N. L. 1993. Intraspecific agonistic behavior of institutional resource collections in herpetology and the salamander Aneides flavipunctatus (Amphibia: ichthyology. Copeia 1985:802–832. Plethodontidae) with comparisons to other pletho- MAIORANA, V. C. 1977. Tail autotomy, functional dontid species. Herpetologica 49:271–282. conflicts and their resolution by a salamander. TILLEY, S. G., AND J. BERNARDO. 1993. Life history Nature 265:533–535. evolution in plethodontid salamanders. Herpeto- MARKS, S. B., AND L. D. HOUCK. 1989. Partial cost of logica 42:154–163. spermatophore production in the salamander Des- TITUS,T.A.,AND A. LARSON. 1996. Molecular phyloge- mognathus ochrophaeus. Journal of Herpetology netics of Desmognathine salamanders (Caudata: 23:81–84. Plethodontidae): a reevaluation of evolution in SEXUAL SIZE DIMORPHISM IN PHAEOGNATHUS 15

ecology, life history, and morphology. Systematic ZAR, J. H. 1994. Biostatistical Analysis. 2nd ed. Biology 45:451–472. Prentice-Hall, Inc., Englewood Cliffs, NJ. VALENTINE, B. D. 1963. The plethodontid salamander Phaeognathus: external morphology and zoogeogra- Accepted: 1 October 2003. phy. Proceedings of the Biological Society of Washington 76:153–158. VERRELL,P.,AND A. DONOVAN. 1991. Male–male APPENDIX 1 Material Examined aggression in the plethodontid salamander Desmog- AUM 2244–2246, 2755–2575, 2760, 4514–4516, 9000– nathus ochrophaeus. Journal of Zoology, London 9003, 9005, 11091,11092, 11114, 11115, 11224–11226, 223:203–212. 11381, 11428–11430, 11490, 11491, 11494, 22700, 25077, WAKE, D. B. 1966. Comparative osteology and evolu- tion of the lungless salamanders, family Pletho- 35147, 35148, 35150, 35514, 35515; CM 39509, 39510, dontidae. Memoirs of the Southern California 90099, 135683–135693; FMNH 140227, 140229–140231, Academy of Sciences 4:1–111. 246873; MVZ 92220, 169267, 169268, 201242, 201250; WAKE, D. B., AND I. G. DRESNER. 1967. Functional SIUC unnumbered (3), 13, 17, 19, 22–24, 2182; ALA morphology and evolution of tail autotomy in 3077, 3078; USNM 142486, 162416, 255248, 291867– salamanders. Journal of Morphology 122:265–306. 291878, 468519; UTA 30104, 53107–53109.

Journal of Herpetology, Vol. 38, No. 1, pp. 15–21, 2004 Copyright 2004 Society for the Study of Amphibians and Reptiles

South Pacific : Human Impacts and a New Species

1,2 3 GREGORY K. PREGILL AND DAVID W. STEADMAN

1Department of Biology, University of San Diego, 5998 Alcala´ Park, San Diego, California 92110, USA 3Florida Museum of Natural History, University of Florida, P.O. Box 117800, Gainesville, Florida 32611, USA

ABSTRACT.—The diversity and distribution of Pacific island iguanas were altered drastically following human colonization around 2800 years ago. A giant recovered from archaeological sites in the Ha’apai group of islands, Kingdom of Tonga, became extinct within a century of human arrival. We describe this iguana as a new species of , a genus with two small arboreal species found today in Fiji (, Brachylophus vitiensis) and parts of Tonga (Brachylophus fasciatus). Additional evidence suggests that B. fasciatus was probably introduced to Tonga (the type locality) by prehistoric people 2000 years after extinction of the giant form. Lapitiguana impensa, described in 2003 from Fiji by G. K. Pregill and T. H. Worthy was an even larger extinct iguana that also succumbed to human impact. The two living species are relicts of a much richer evolutionary history than previously known.

ThefirstpeopletocolonizethePacific the giant tortoises consumed by 18th-century encountered birds and reptiles that had evolved mariners on islands in the Indian Ocean on islands free of mammalian predators. Hu- (Arnold, 1979). In the Pacific, extinction and mans preyed on these naive species, many of extirpation of insular correlate with which became extinct, sometimes abruptly prehistoric human colonization, but evidence (Martin and Steadman, 1999). New Zealand’s of direct predation is lacking or circumstantial large, flightless moas are a stunning example of (Pregill and Dye, 1989; Pregill, 1993, 1998). Most unrestrained exploitation by prehistoric people species of Pacific island lizards are small (body (Holdaway and Jacomb, 2000), yet the list of mass typically , 15 g; snout–vent length , 100 birds similarly extirpated throughout the Pacific mm), cryptic, and would be of little interest to also includes megapodes, rails, pigeons, parrots, early hunters. In the southwest Pacific islands of passerines, and many seabirds (Steadman, Fiji and Tonga, however, two species of giant 1995). Among insular reptiles, by contrast, most iguanas did attract attention. One of these, known examples of extinction through human Lapitiguana impensa,wasrecentlydescribed predation took place historically. Notable are from Fiji (Pregill and Worthy, 2003). The other is a species of Brachylophus, described below, 2 Corresponding Author. E-mail: pregill@sandiego. that apparently was rapidly consumed to edu extinction by Tonga’s first human colonizers. 16 G. K. PREGILL AND D. W. STEADMAN

TABLE 1. Selected measurements (millimeters) of fossil bones, Brachylophus gibbonsi, new species and Brachylophus fasciatus (SDSNH 60429, SVL 180 mm).

gibbonsi fasciatus Cranial parietal width-frontal border 23.6 12.5 basale height 15.2 8.5 basale length 15.1 8.0 quadrate height 14.0 7.4 premaxilla width 10.7, 15.6 4.2 premaxilla length 15.6 6.7 dentary tooth row 30.0, 32.8 16.0 Prescral vertebrae centrum length* 10.6, 11.3, 6.3 12.4, 13.8 Sacral vertebra width across diapophyses 30.3 15.1 Girdles scapula height 27.2, 27.5 10.1 ilium length 32.6, 41.9 15.0 acetabulum diameter 13.9, 14.3 6.0 Limbs humerus shaft length 50.5, 50.8 31.7 FIG. 1. The Ha’apai Island Group of Tonga, show- humerus distal width 16.4, 20.0 7.8 ing location of archaeological sites with bones of Bra- femur shaft length 62.6, 65.0 35.7 chylophus gibbonsi. femur proximal width 14.2, 15.0 6.0 tibia length 55.6, 61.5 26.5 *forfasciatus caudal no. 15. The living and extinct iguanas (Brachylophus spp., Lapitiguana) of the southwest Pacific are members of a monophyletic lineage () In Tonga, the existence of a third species of of large mainly herbivorous squamates other- Brachylophus, much larger than its living con- wise confined to the Americas (Etheridge, 1982; geners, was first revealed by fossils from the de Queiroz, 1987). Of the two extant Pacific Tongoleleka archaeological site on Lifuka Island species, the Fijian Banded Iguana, Brachylophus in the Ha’apai group (Pregill and Dye, 1989). At fasciatus, was discovered by Europeans in the that time, the comparatively few bones re- late 18th century. It occurs today on islands covered were undiagnostic beyond referral to scattered across Fiji and parts of Tonga (Gib- Brachylophus. Further excavations (1995–1997) in bons, 1981; Zug, 1991). The Crested Iguana, Ha’apai by D. V. Burley, D. W. Steadman, and Brachylophus vitiensis, unknown to science until field parties yielded more than 1300 bones of 1979, is restricted to Yadua Taba and a few Brachylophus other small islands in Fiji’s Yasawa group this large from five cultural sites (Gibbons, 1984; Laurie et al., 1987; Harlow on five islands (Fig. 1). Particularly at the and Biciloa, 2001). Both living species are Tongoleleka site, careful stratigraphic control cryptic, arboreal, and endangered. The two and radiometric dating demonstrated that this extinct forms were much larger than either large was a food item and became extinct extant species. Lapitiguana impensa was found in along with various birds almost immediately late Quaternary deposits on Fiji’s main island after first human contact (Steadman et al., Viti Levu (Worthy et al, 1999). Like Brachylophus 2002b). it appears to be a basal iguanid, although a sister-group relationship is unverified (Pregill MATERIALS AND METHODS and Worthy, 2003). Because a few of the fossils Field methods and site descriptions are were associated with cultural artifacts, the detailed elsewhere (Burley, 1999; Dickinson et extinction of Lapitiguana must be subsequent al., 1999; Steadman et al., 2002a,b). Measure- to human settlement of Viti Levu, although the ments (Table 1) were rounded to the nearest 0.1 chronology is imprecise. mm. Osteological characters and characterstates EXTINCT BRACHYLOPHUS FROM TONGA 17

TABLE 2. Summary of bones, Brachylophus gibbonsi, new species, from cultural sites in the Ha’apai Group, Kingdom of Tonga. FO1 5 Faleloa site, Li7 5 Tongoleleka site, HA1 5 Ha’ano, Ui4 5 ‘Uiha, HF1 5 Mele Hevea site, Ha’afeva.

Element FO1 Li7 HA1 Ui4 HF1 Total Cranial 22 13 1 1 1 38 Axial presacral 54 35 9 – – 98 sacral 5 1 – – – 6 caudal 98 67 6 – 3 174 vertebral scraps 20 22 3 – – 45 ribs 8 12 – 3 – 23 Total axial 185 137 18 4 4 348 Appendicular humerus 29 5 4 – – 38 radius 3 1 – – – 4 ulna 4 1 – 1 1 7 femur 11 17 1 – – 29 tibia 9 6 1 – – 16 fibula 4 5 – 1 1 11 scapulocoracoid 7 1 1 – – 9 pelvis 9 7 1 – – 17 metapodial 100 57 13 8 8 186 phalanges 140 71 10 8 8 237 shaft scraps 43 6 15 1 1 66 misc. scraps 162 139 12 1 1 315 Total app. 521 316 58 20 20 935 Total Bones per Excavation Unit 728 466 77 25 25 1321

are those of de Queiroz (1987), Frost and 11, Faleloa Archaeological Site, Foa Island, Etheridge (1989), Norell and de Queiroz (1991) Ha’apai Group, Kingdom of Tonga. Late Holo- and Pregill and Worthy (2003). Comparative cene, approximately 2800 years B.P. skeletons examined were Amblyrhynchus crista- Paratypes.—Approximately 1320 bones and tus SDSNH 45157, Brachylophus fasciatus bone fragments (UF, Florida Museum of Natural SDSNH 55289, 60429, UF 37578, Brachylophus History) from cultural sites dating from 2800– vitiensis MCZ 158238, 160254, sp. 2600 years B.P. on islands in the Ha’api Group GKP uncataloged, cornuta UF 99017, (Foa, Lifuka, Ha’ano, ‘Uiha, Ha’afeva), Kingdom 34752, Dipsosaurus dorsalis WP 381, 510, Iguana of Tonga (Table 2). iguana GKP uncataloged, Sauromalus obesus WP Diagnosis.—Brachylophus gibbonsi is distin- 40, 74, 163, and Sauromalus varius WP 178. The guished from the two extant species (B. fasciatus, Brachylophus fossils are cataloged in the Verte- B. vitiensis) by incomplete closure of Meckels brate Paleontology Collection of the Florida canal at the anterior end of the dentary. In B. fasciatus and B. vitiensis, the canal is exposed Museum of Natural History, University of as a pin-hole size opening near the tip of the Florida. mandibular symphysis. The opening is larger in some other Iguanidae (e.g, species of Ctenosaura) but in no instance is as large as in B. gibbonsi, SYSTEMATIC PALEONTOLOGY where the opening is a conspicuous foramen Reptilia below the second to fifth tooth. Brachylophus Lepidosauria gibbonsi is distinguished further from B. fasciatus and B. vitiensis by substantially larger size, Iguanidae having a snout–vent length almost twice that Brachylophus Cuvier of the two extant species. Brachylophus gibbonsi, new species Etymology.—To recognize the late J. R. H. Holotype.—UF 212300 right dentary. Collected Gibbons for his exemplary work on the ecology by D. V. Burley et al., June 1997. and systematics of living species of Brachylophus, Locality and Horizon.—Excavation unit 17, level gibbonsi is a proper noun. 18 G. K. PREGILL AND D. W. STEADMAN

mm SVL. Based on proximal limb bones the SVL was 375–444 mm, and on length of presacral centra approximately 340 mm. Taking 350 mm as a working estimate, the Tongan lizard was 1.8–1.9 times the SVL of adult B. fasciatus and about 1.7–1.8 times that of B. vitiensis.

DISCUSSION FIG. 2. Holotype right dentary (UF 212300) of Bra- Osteological Characters.—Most of the diagnos- chylophus gibbonsi, new species. Lingual view. tic elements of the skull (e.g., frontal, palatine, pterygoid, complete mandible) are not repre- sented in the referred material, and the only DESCRIPTION confirmed synapomorphy with Brachylophus is Dentary.—Of nine dentaries, the holotype (Fig. the absence of autotomy septa in the caudal 2) is the most nearly complete. The tooth row is vertebrae (also absent in Amblyrhynchus, Con- olophus Iguana delicatissima 30.0 mm and has placements for 21 pleurodont , and ). Referral to Brachylophus rests on confirmation of skeletal teeth. Counting from the symphysis, those details and the absence of conflicting character present are numbers 3, 5–8, 10, 12–15, 17. The states. In fact, osteological characters that di- third tooth is weakly tricuspid, whereas all agnose Brachylophus are either plesimorphic following teeth are fully tricuspid. Meckel’s (e.g., posterior marginal teeth simple and tri- canal is an elliptical pore on the ventrolingual cuspid, not polycuspid or serrate) or homoplas- surface from below the position of the fourth tic with respect to other Iguanidae (e.g., absence tooth. Posteromedially, the dentary is recessed of autotomy septa) (de Queiroz, 1987; Pregill and for the splenial to a forward position below tooth Worthy, 2003). Brachylophus is one of the basal 16. Six mental foramina penetrate the labial face Iguanidae, which along with Lapitiguana (Pregill between tooth 3 and 13. The articular surface and Worthy, 2003) and Dipsosaurus þ Armandi- marking overlap of the coronoid extends anteri- saurus (Norell and de Queiroz, 1991) are outside orly to below tooth 20. the clade encompassing the other genera (Am- All but one of the eight paratypical dentaries blyrhynchus, , Ctenosaura, Cyclura, consist of anterior fragments from various sized Iguana, Sauromalus, Pumilia). Brachylophus differs individuals. Each possesses the large opening of from Lapitiguana in having the parietal foramen Meckel’s canal. One intact specimen has a tooth located at the frontopareital suture (ancestral), row length of 32.8 mm and positions for 22 teeth, the anterior inferior alveolar foramen located of which numbers 3 and 5 are present. (usually) within the dentary (derived), the Other Cranial Elements.—Additional skull absence of autotomic septa in the caudal bones include partial maxillae, two premaxillae, vertebrae, and smaller size (Pregill and Worthy, a complete braincase, a parietal, and two 2003). Brachylophus gibbonsi was approximately quadrates. Other than their larger size, the fossils intermediate in adult SVL between Lapitiguana are essentially indistinguishable from the same (approximately 500 mm) and B. fasciatus (ap- elements in living Brachylophus. proximately 190 mm). Postcranial Elements.—Limb bones, girdles, Biogeography.—Brachylophus and Lapitiguana and vertebrae are abundantly represented (97% are members of a monophyletic Iguanidae (de of specimens, Table 1), although very few are Queiroz, 1987; Frost and Etheridge, 1989; Norell entire. None differs qualitatively from the same and de Queiroz, 1991) that ranges from southern elements in B. fasciatus or B. vitiensis, except that North America to northern South America, the the larger limb bones (e.g., femur, tibia) are West Indies, and the Gala´pagos Islands. Roughly proportionately more massive at their articular 5000 km of ocean separate the ends. Like extant Brachylophus but unlike Lap- mainland from Fiji and Tonga. Iguanas occur itiguana, caudal vertebrae from the proximal naturally nowhere else in Oceania, nor is there series lack autotomic fracture planes. fossil evidence to suggest that they ever did. Fiji Body Size.—Estimates of snout–vent length and Tonga are oceanic in origin (Martinez and (SVL) are obtained by extrapolating from ratios Taylor, 2002), and long-distance dispersal from of measurements taken from comparative skel- the Americas, incredible as that seems, is the etons of known size. Estimates vary because of only viable hypothesis that explains their pres- allometry and variation in skeletal proportions ence in the southwest Pacific (Cogger, 1974; (Pregill and Worthy, 2003). Cranial elements Gibbons, 1981). A hypothetical iguana-bearing typically yield lower estimated SVLs than do raft from South America captured by the girdles and limbs. For example, the length of the counter-clockwise gyre of the Humboldt–South dentary tooth row, length of the braincase, and Equatorial current probably would drift west- width of the parietal place B. gibbonsi at 310–340 ward between 58Nand58S before possibly EXTINCT BRACHYLOPHUS FROM TONGA 19

FIG. 3. Relative body size of the two living (Brachylophus fasciatus, Brachylophus vitiensis) and the two extinct (Brachylophus gibbonsi, Lapitiguana impensa) species of iguanas in the southwest Pacific. turning southwesterly toward Tonga and Fiji. inhabited small (1.8–13.3 km2), topographically Thus, but for passing close to Galapa´gos and uncomplicated (11–20 m elevation) islands in perhaps the Marquesas, the raft would miss the central Tonga when people arrived around 2800 hundreds of intervening islands of eastern and years ago. These two vertebrates immediately central Polynesia. Iguanas (Amblyrhynchus, Con- became an important food resource. Unprepared olophus) did reach Gala´pagos, but evidently not for human predators, the extinction of both the Marquesas, where lizard bones from several species was astonishingly rapid. Well-dated archaeological sites include only extant skinks stratigraphic sequences at the Tongoleleka site and geckos (GKP, unpubl.). show that the iguana and megapode both For a squamate , the formidable cross- disappear in a time interval too brief (a century ing of the Pacific would entail surviving several or less) to be resolved by radiocarbon dating months of potentially lethal physiological stress. (Steadman et al., 2002b). Alternatively, eggs deposited in a hollow log or Despite being known as the ‘‘Fijian’’ banded similar refuge might face better odds if they iguana, the first specimens of B. fasciatus came passed all or most of the voyage incubating. from Tonga (Gibbons and Wadkins, 1982; Zug, Significantly, living Brachylophus deposit large 1991). Gibbons (1981) traced the specimen(s) to eggs in small clutch sizes (3–4 eggs) and have the naturalist Riche, who was aboard the French extremely long incubation periods, about 8þ vessel Esperance when it stopped at Tongatapu in months in B. vitiensis, which is two to three times 1792. Among the goods obtained in trade were that reported for other iguanids (Gibbons and iguanas the natives claimed were good to eat Wadkings, 1982). (Gibbons, 1981). The fossil record strongly Given the low probability of rafting such argues, however, that the smaller B. fasciatus a long distance successfully, Fiji and Tonga and larger B. gibbonsi never coexisted in Tonga. probably were invaded by iguanas only once. Among the 100,000þ prehistoric bones of fish, Hence, Lapitiguana and Brachylophus diverged reptiles, birds, and mammals from sites in from a common ancestor subsequent to coloni- Ha’apai, Tongatapu, and ‘Eua (Pregill, 1993; zation. The two living species are relicts of Steadman, 1993; Koopman and Steadman, a radiation that also included the two much 1995; Steadman et al., 2002a,b) not a single bone larger and presumably terrestrial species (Fig. 3). of B. fasciatus has been found. The absence of The arboreal habits and smaller size of B. B. fasciatus in the Tongan bone record suggests fasciatus and B. vitiensis may have evolved early that it was not present in Tonga until 2000 years or even be ancestral traits, but in either case after the extinction of B. gibbonsi. A late pre- these attributes likely saved them from extinc- historic (, 500 years ago) human introduction tion following human contact. would account for the fact that Tongan B. Brachylophus gibbonsi and the extinct mega- fasciatus are essentially indistinguishable from pode, Megapodius alimentum (Steadman, 1989) Fijian populations (Zug, 1991) and for its Tongan 20 G. K. PREGILL AND D. W. STEADMAN name (vokai) being cognate with Fijian fokai. The maximal body size. Proceedings of the National nearest source of B. fasciatus is the Lau Group, an Academy of Sciences USA 98:14518–14523. island arc about equidistant between Tongatapu COGGER, H. G. 1974. Voyage of the Banded Iguana. and Fiji’s main island of Viti Levu. Archaeolog- Australian Natural History 18:144–149. ical and ethnohistoric studies indicate substan- DE QUEIROZ, K. 1987. Phylogenetic systematics of tial prehistoric contact between the people of iguanine lizards: a comparative osteological study. Univ. of California Publications, Zoology 118:1–203. Lau and Tonga (Best, 1984). Our recent archae- DICKINSON, W. R., D. V. BURLEY, AND R. SHUTLER JR. 1999. ological excavations on other islands in Lau Holocene paleoshoreline record in Tonga: geomor- (Aiwa, Nayau) show that B. fasciatus was present phic features and archaeological implications. prehistorically. Journal of Coastal Research 15:682–700. The emerging fossil record of Pacific island ETHERIDGE, R. 1982. Checklist of the Iguanine and reptiles reveals a pattern of Holocene extinctions, Malagasy iguanid lizards. In G. Burghardt and A. S. extirpations, and translocations in which large- Rand (eds.), Iguanas of the World, pp. 7–37. Noyes bodied species were especially vulnerable (see Publications, Park Ridge, NJ. also Pregill, 1986; Burness et al., 2001; Mead et FROST, D. R., AND R. ETHERIDGE. 1989. A phylogenetic al., 2002). For Pacific iguanas, the two largest analysis and of Iguanian lizards (Repti- species, including a monotypic genus, were lia: Squamata). Univ. of Kansas Museum of Natural wiped out by early human settlers in Fiji and History Miscellaneous Publications 81:1–65. Tonga, leaving behind only bones. Conversely, GIBBONS, J. R. H. 1981. The biogeography of Brachylo- phus (Iguanidae), including the description of a new the smallest species was taken by people in late species, B. vitiensis, from Fiji. Journal of Herpetol- prehistoric times from Fiji to Tonga, where it still ogy 15:255–273. survives. Thus, in spite of losing the two largest ———. 1984. Iguanas of the South Pacific. Oryx 18: species, iguanas have fared better than other 82–91. groups of large terrestrial reptiles in the Pacific, GIBBONS, J. R. H., AND I. WATKINS. 1982. Behavior, the meiolaniid turtles and the mekosuchine ecology and conservation of the South Pacific crocodiles that have been lost to human pre- Banded Iguanas, Brachylophus, including a newly dation throughout their known ranges of New discovered species. In G. Burghardt and A. S. Rand Caledonia, Vanuatu, and Fiji (Mead et al., 2002). (eds.), Iguanas of the World, pp. 418–444. Noyes Modern distributions of Pacific island reptiles, Publications, Park Ridge, NJ. much like those for birds, can be understood HARLOW, P. S., AND P. N. BICILOA. 2001. Abundance of only after considering the effects of the past three the Fijian crested iguana (Brachylophus vitiensis)on two islands. Biological Conservation 98:223–231. millennia of human presence. HOLDAWAY, R. N., AND C. JACOMB. 2000. Rapid extinction of the Moas (Aves: Dinornithiformes): model, test, Acknowledgments.—Comparartive skeletons and implications. Science 287:2250–2254. were made available by D. Auth, Florida KOOPMAN,K.F.,AND D. W. STEADMAN. 1995. Extinction Museum of Natural History (UF); J. Hanken and biogeography of bats on ‘Eua, Kingdom of and J. Rosado, Museum of Comparative Zoology Tonga. American Museum Novitates 3125:1–13. (MCZ); San Diego Natural History Museum LAURIE, W. A., H. URYU, AND D. WATLING. 1987. A faunal (SDSNH); and W. Presch (WP). Support was survey of Yaduataba Island Reserve with particular provided by National Science Foundation Grant reference to the crested iguana (Brachylophus EAR-97714819 (DWS), University of Florida vitiensis Gibbons 1981). Domodomo Fiji Museum Division of Sponsored Research Grant RDA 1- Quarterly 1:16–28. 23-95-96 (DWS), and a University of San Diego MARTIN,P.S.,AND D. W. STEADMAN. 1999. Prehistoric extinctions on islands and Continents. In R. Faculty Research Grant (GKP). MacPhee (ed.), Extinctions in Near Time, pp. 17–55. Kluwer Academic/Plenum Publishers, New York. LITERATURE CITED MARTINEZ,F.,AND B. TAYLOR. 2002. Mantle wedge ARNOLD, E. N. 1979. Indian Ocean giant tortoises: their control of back-arc crustal accretion. Nature 416: systematics and island adaptations. Philosophical 417–420. Transactions of the Royal Society London B Biolog- MEAD, J. I., D. W. STEADMAN,S.H.BEDFORD,C.J.BELL, ical Sciences 286:127–145. AND M. SPRIGGS. 2002. New extinct mekosuchine BEST, S. 1984. Lakeba: The Prehistory of a Fijian Island. crocodile from Vanuatu, South Pacific. Copeia 2002: Unpubl. Ph.D. diss., Univ. of Auckland, Auckland, 632–641. New Zealand. NORELL, M. A., AND K. DE QUEIROZ. 1991. The earliest BURLEY, D. V. 1999. Lapita settlement to the east: new iguanine lizard (Reptilia: Squamata) and its bearing data and changing perspectives from Ha’apai on Iguanine phylogeny. American Museum Nov- (Tonga) prehistory. In J.-C. Galipaud and I. Lilley itates 2997:1–16. (eds.), The Pacific from 5000 to 2000 BP: Coloniza- PREGILL, G. K. 1986. Body size of insular lizards: tion and Transformation, pp. 189–200. Editions de a pattern of Holocene dwarfism. Evolution 40: IRD, Paris. 997–1008. BURNESS, G. P., J. DIAMOND, AND T. FLANNERY. 2001. ———. 1993. Fossil lizards from the late Quaternary of Dinosaurs, dragons, and dwarfs: the evolution of ‘Eua, Tonga. Pacific Science 47:101–114. EXTINCT BRACHYLOPHUS FROM TONGA 21

———. 1998. Squamate reptiles from prehistoric sites STEADMAN, D. W., A. PLOURDE, AND D. V. BURLEY. 2002a. in the Mariana Islands. Copeia 1998:64–75. Prehistoric butchery and consumption of birds in PREGILL, G. K., AND T. DYE. 1989. Prehistoric extinction the Kingdom of Tonga, South Pacific. Journal of of giant iguanas in Tonga. Copeia 1989:505–508. Archaeological Science 29:571–584. PREGILL, G. K., AND T. H. WORTHY. 2003. A new iguanid STEADMAN, D. W., G. K. PREGILL, AND D. V. BURLEY. lizard (Squamata, Iguanidae) from the late Quater- 2002b. Rapid prehistoric extinction of iguanas and nary of Fiji, Southwest Pacific. Herpetologica 59: birds in Polynesia. Proceedings of the National 57–67. Academy Sciences USA 99:3673–3677. STEADMAN, D. W. 1989. New species and records of WORTHY, T. H., A. J. ANDERSON, AND R. E. MOLNER. 1999. birds (Aves: Megapodiidae, Columbidae) from an Megafaunal expression in a land without mam- archaeological site on Lifuka, Tonga. Proceedings of mals—the first fossil faunas from terrestrial de- the Biological Society of Washington 102:537–552. posits in Fiji (Vertebrata: Amphibia, Reptilia, Aves). ———. 1993. Biogeography of Tongan land birds Senckenbergiana Biologica 79:237–242. before and after human impact. Proceedings of ZUG, G. R. 1991. The lizards of Fiji: natural history and the National Academy Sciences USA 90:812–822. systematics. B. P. Bishop Museum Bulletins Zool- ———. 1995. Prehistoric extinctions of Pacific island ogy 2:1–136. birds: biodiversity meets zooarchaeology. Science 267:1123–1131 Accepted: 1 October 2003.

Journal of Herpetology, Vol. 38, No. 1, pp. 21–26, 2004 Copyright 2004 Society for the Study of Amphibians and Reptiles

Intraguild Predation on Congeners Affects Size, Aggression, and Survival among Ambystoma Salamander Larvae

ROBERT BRODMAN

Biology Department, Saint Joseph’s College, Rensselaer, Indiana 47978, USA; E-mail: [email protected]

ABSTRACT.—I designed laboratory and mesocosm experiments to test the hypotheses that consumption of a congener as supplemental food can increase the size variance, size, aggression rates, and mortality of intraguild predator populations. Experimental populations of Marbled Salamander (Ambystoma opacum), Jefferson Salamander (Ambystoma jeffersonianum) and Tiger Salamander (Ambystoma tigrinum) larvae that were initially fed a smaller congener Spotted Salamander (Ambystoma maculatum) as a food supplement developed larger SVL and size variation after seven days than control larvae that were not fed a congener. Experimental treatment populations had greater initial size variation because some larvae ate a congener and others did not. Treatment larvae had increased SVL, aggression rates and mortality after 60 days compared to control populations. These results suggest that intraguild predation on a congener can affect fitness and population dynamics of predator populations. I hypothesize that intraguild predation on congeners results in size variation and then increased aggression and mortality caused by cannibalism.

Pond dwelling salamander larvae of the genus timing of breeding or developmental rates can Ambystoma experience density-dependent effects also lead to intraguild predation, where early- on survival and growth (Wilbur, 1972; Scott, breeding or fast-growing species can prey upon 1990; Smith, 1990; Van Buskirk and Smith, 1991). larvae of late-breeding or slow-growing forms When several species coexist, they often differ in (Stenhouse et al., 1983; Wissinger, 1992; Brod- the time of egg laying, hatching, and develop- man, 1996; Jaskula and Brodman, 2000). ment (Cortwright 1988; Brodman 1996; Jaskula Mesocosm experiments on interspecific inter- and Brodman, 2000). These differences have actions of Ambystoma salamander larvae have been hypothesized to reduce competition among shown reduced survival and increased growth of larvae and contribute to mutual coexistence the intraguild predator in the presence of conge- among species (Keen et al., 1984; Talentino and neric prey (Brodman, 1996; Wildy et al., 1998; Landre, 1991). However, animals with complex Jaskula and Brodman, 2000). Size variation within life cycles often show large variation in size and a population can result in interference competi- timing of metamorphosis (Wilbur and Collins, tion among individuals of different size that 1973; Peckarsky et al., 2001). Differences in results in decelerated growth of smaller larvae 22 ROBERT BRODMAN

(Stenhouse et al., 1983; Smith, 1990; Ziemba and Egg masses of A. maculatum, A. opacum, A. Collins, 1999). However, studies on food and jeffersonianum, and A. tigrinum were collected space limitation failed to demonstrate that these from natural ponds in Ohio and Indiana and observations are the result of interspecific com- brought into the lab. Eggs were kept in an petition for resources (Brodman, 1999). environmental chamber at 5–108C until they One explanation for these disparate results is hatched. Hatchlings were housed individually in that predation on a smaller congener during early 250 mL plastic cups with 200 mL of pond water. development could result in accelerated growth Twenty milliliters of water was removed and because of eating a relatively large meal (Lannoo replaced with 20 mL of pond water containing et al., 1989; Wildy et al., 1998) or because of a large zooplankton or 20 mL of dechlorinated tap dose of the growth hormone thyroxine (Gorbman, water containing brine shrimp (Artemia) nauplii 1964). If some but not all members of a population every other day. prey upon congeners, then this could lead to Food Supplement Experiment.—The SVL of 160 growth spurts in a portion of the population and five-month-old A. opacum larvae, 160 one- result in larger size variation. Because size month-old A. jeffersonianum larvae and 160 one- variation determines the rate of cannibalism (Polis month old A. tigrinum larvae were measured. 1981; Larson et al., 1999: Maret and Collins, 1994; Larvae of each species were assigned to control Ziemba and Collins, 1999), which is widespread (80 larvae of each species) and treatment groups among Ambystoma salamander larvae (Crump (80 larvae of each species) such that there was no 1992), intraguild predation in Ambystoma sala- significant difference in SVL between the control mander larvae could result in increased predation and treatment groups for each species (one-way on smaller conspecifics. This would result in ANOVA, df 5 1,158; F 5 0.060; P 5 0.806). All higher mortality but larger size of the intraguild larvae were maintained in individual cups on an predators that survive when two or more species ad libitum diet of zooplankton for one week; of Ambystoma salamander larvae coexist. however, all treatment larvae were also given The purpose of this study was, first, to use one A. maculatum hatchling as a food supple- a laboratory experiment to test the hypothesis ment on day one of the experiment. SVL of each that eating a congener as a supplemental meal larva was measured on day seven. I also can increase the size of Ambystoma larvae during recorded which treatment larvae ate the spotted early development. In a second experiment, I salamander hatchling during the seven days tested the hypothesis that increased size variance trial. after eating congeners results in increased size, Data were tested for normality using the rates of intraspecific aggression, and mortality Kolmogorov-Smirnov test and tested for homo- during larval development in mesocosms. geneity of variances using the Levene statistic. SVLs were loge-transformed to satisfy conditions for parametric tests. The one-way ANOVA (a 5 MATERIALS AND METHODS 0.05) was used to test the null hypothesis that the Study Animals.—The Spotted Salamander SVL of treatment and control larvae are not (Ambystoma maculatum) was chosen as a repre- significantly different after seven days. The F- sentative late-breeding, slow-growing species to test (a 5 0.05) was used to test the null be potential prey for intraguild predators. hypothesis that the variance of SVL between Marbled Salamanders (Ambystoma opacum), Jef- treatment and control larvae was not significant- ferson Salamanders (Ambystoma jeffersonianum) ly different after seven days. The laboratory and Tiger Salamanders (Ambystoma tigrinum) experiment was also used as a means of were chosen as representative intraguild preda- generating size variation among larvae for the tor species because each is a relatively early- mesocosm experiments that follow. breeding or fast-growing species that is known Mesocosm Experiment.—Mesocosms were cre- to prey on spotted salamander larvae (Wilbur, ated in thirty 85L (0.15 m2) plastic tubs filled to 1972; Stenhouse et al., 1983; Cortwright, 1988). a depth of 47 cm with 70 liters of filtered pond Larvae of these species are also cannibalistic water and a 10 cm layer of leaf litter covering the (Lannoo et al, 1989; Smith, 1990; Brodman 1996). tub bottom. Each mesocosm was inoculated with Ambystoma opacum is a fall breeder with larvae one liter of concentrated pond water containing that overwinter and become the earliest larval algae and colonies of zooplankton and aquatic cohort in spring (Keen at al., 1984; Cortwright, insects. The pond water was filtered through dip 1988). Ambystoma jeffersonianum is a winter nets (seven mesh/cm) to remove large pre- breeder (Minton, 1972) with a relatively fast daceous insects. Mesocosms were placed out- larval growth rate (Brodman, 1995, 1996). doors within an oak wood lot and covered with Ambystoma tigrinum larvae have faster growth aluminum screens (seven mesh/cm) to exclude rates than other Ambystoma species (Keen et al., predators. Supplemental food (e.g., Tubifex 1984). worms, zooplankton, mosquito larvae, or tad- INTRAGUILD PREDATION ON CONGENER SALAMANDERS 23

FIG. 2. Mean SVL growth 6 SE (mm) of treatment FIG. 1. Mean SVL 6 SE (mm) of Ambystoma opacum, Ambystoma opacum, Ambystoma jeffersonianum,and Ambystoma jeffersonianum,andAmbystoma tiginum Ambystoma tiginum larvae after seven days in the food larvae after seven days in the food supplement supplement experiment. The F- and P-values from experiment. The F- and P-values from separate one- separate one-way ANOVA tests of treatment larvae way ANOVA tests of control versus treatment (a 5 that ate a congener versus treatment larvae that did not 0.0167 after Bonferroni adjustment) are F1,79 5 291.2, eat a congener (a 5 0.0167 after Bonferroni adjustment) P 5 0.0001 (A. opacum); F1,79 5 43.9, P 5 0.0001 are F1,79 5 21.8, P 5 0.0003 (A. opacum); F1,79 5 41.3, (A. jeffersonianum); and F1,79 5 110.8, P 5 0.0001 P 5 0.0001 (A. jeffersonianum); and F1,79 5 24.0, (A. tigrinum). P 5 0.0001 (A. tigrinum). pole hatchlings) was added in equal amounts to Each of the 30 mesocosms was observed daily each mesocosm once a week. Salamander larvae for mortality and all dead larvae were removed. alwayshadfoodavailablethroughoutthe To measure aggression rates, the behavior of experiment and were thus fed ad libitum. larvae was recorded during 30-min observation During daily observations any predatory insects periods in each mesocosm during the second (Odonate and Coleoptera larvae larger than 0.5 and third weeks of the study. Observations were cm) observed in a mesocosm were removed made on two to four mesocosms per evening using a small dip net prior to attaining a size that within the first two hours after sunset. Light could kill salamander larvae. conditions were maintained using flashlights After the food supplement experiment, control with 1-LUX lights and red cellophane filters. A. opcaum, A. jeffersonianum, and A. tigrinum Following Walls (1998), the total number of acts larvae that were not given supplemental food, of cannibalism, bites and lunges toward sala- and treatment larvae that were given spotted mander larvae during the 30-min observation salamander hatchlings were randomly selected periods were used to measure aggression rates. for the mesocosm experiment. Ten control larvae Data was tested for normality using the of a single species were randomly selected and Kolmogorov-Smirnov test and tested for homo- placed together in control mesocosms. Treatment geneity of variances using the Levene statistic. mesocosms included a mixture of seven larvae SVL and aggression rates were loge-transformed randomly selected from the treatment group of and percent survival was arcsine-transformed to the food supplement experiment that had eaten satisfy conditions for parametric tests. A two- an A. maculatum larva and three larvae randomly way MANOVA (a 5 0.05) was used to test the selected from the food supplement treatment null hypothesis that the multiple response group that were given an A. maculatum larva but variables were not significantly different among did not eat it within one week. The 7:3 ratio was species and between the control and treatment chosen because approximately 70% of larvae in groups. Separate ANOVA (a 5 0.0167 after the treatment group for each species ate the Bonferroni adjustment) tests were used to test Spotted Salamander hatchling and 30% did not. the null hypothesis that the SVL, survival and Therefore the treatment groups in the mesocosm rates of aggression of treatment and control experiment began with greater size variation larvae were not significantly different. than the control groups. There were five repli- cates of control and treatment mesocosms for RESULTS each species (30 mesocosms) set up as a 3 3 2 factorial design (three species and control-treat- Food Supplement Experiment.—In the food ment). Survival and SVL were measured for all supplement treatments, 65% of the A. opacum, surviving larvae in each mesocosm after 60 days. 74% of the A. jeffersonianum, and 71% of the A. The initial density of 67 larvae/m2 in each tigrinum that were offered an A. maculatum ate it mesocosm was within the range of natural within seven days. There was no significant densities for Ambystoma species (Brodman, difference in SVL between control (11.2 6 0.1 1995; Stangel, 1988; Van Buskirk and Smith, mm) and treatment larvae (11.2 6 0.1 mm) at 1991; pers. obs.). the beginning of the laboratory experiment 24 ROBERT BRODMAN

TABLE 1. Mean 6 SE percent survivoral, SVL (millimeters) and rate of aggression (per 30 min) of Ambystoma opacum, Ambystoma jeffersonianum, and Ambystoma tigrinum larvae in mesocosms after 60 days. The F- and P- values are from separate one-way ANOVA (df, 1,8) tests of control versus treatment (a 5 0.0167* after Bonferroni adjustment).

Species % Survival SVL Aggression A. opacum Control 62.5 6 4.1 19.9 6 0.1 0.25 6 0.06 Treatment 46.0 6 3.7 22.1 6 0.1 0.55 6 0.08 F-value 12.68 9.91 9.25 P-value 0.007* 0.014* 0.016* A. jeffersonianum Control 80.8 6 5.4 19.8 6 0.1 0.31 6 0.03 Treatment 48.5 6 2.2 22.8 6 0.1 0.49 6 0.04 F-value 30.87 36.88 11.13 P-value 0.001* 0.0001* 0.010* A. tigrinum Control 46.9 6 4.6 25.2 6 0.2 0.57 6 0.06 Treatment 12.5 6 0.0 40.1 6 0.3 1.11 6 0.02 F-value 55.00 26.85 76.96 P-value 0.0009* 0.001* 0.0001*

(ANOVA, F1,239 5 0.06, P 5 0.806). Treatment in A. tigrinum and A. jeffersonianum and once in salamander larvae (13.6 6 0.3 mm) grew A. opacum). significantly larger than control (11.6 6 0.1 mm) larvae after seven days (ANOVA, F1,239 5 43.81, P , 0.001). This difference was also significant DISCUSSION for each of the three species (Fig. 1). Within the The results of this study demonstrate that treatment group salamanders that ate an A. eating a congener during early larval develop- maculatum grew significantly more after seven ment affects growth, size variation, aggression days than those that did not eat an A. maculatum and mortality in three species of Ambystoma (Fig. 2). The SVL and growth of salamander salamanders. These data are consistent with larvae from the treatment that did not eat an A. other studies that observed increased larval size maculatum larva were not significantly different but decreased survival of intraguild predators from the control after seven days. Treatment when raised with congeners (Cortwright, 1988; larvae also had significantly greater variance in Brodman, 1996; Jaskula and Brodman, 2000). SVL than control groups after seven days (F-test, Greater mortality of intraguild predators in the F1,79 5 14.0, P 5 0.001). presence of supplemental prey also occurs in Mesocosm Experiment.—Treatment larvae that some invertebrate taxa (Moran and Hurd, 1997). were fed A. maculatum hatchlings had signifi- Intraguild predation is a common interaction cantly greater initial variance in SVL than in invertebrates and vertebrates with disparate control larvae (ANOVA, F1,24 5 85.6, P , sizes and is important in influencing the 0.001). After 60 days in mesocosms there was abundance and distribution of species (Wis- a significant difference in response variables singer et al., 1996; Holt and Polis, 1997; Lucas between control and treatment (MANOVA, F1,24 et al., 1998; Schellhorn et al., 1999). Interactions 5 70.9, P , 0.001), among species (MANOVA, such as competition and predation between size F1,24 5 42.2, P , 0.001), and the interaction classes affect size-specific rates of survival and between treatment and species (MANOVA, F1,24 growth (Polis and McCormick, 1987; Persson, 5 5.1, P 5 0.014). There were significant within- 1988; Wissinger, 1989; Smith, 1990; Lucas et al., subject effects between treatment and the mul- 1998). This leads to a distribution of body sizes tiple variables (Wilks’, F4,24 5 86.7, P , 0.001). in a population that affects future population Treatment larvae that were fed A. maculatum growth (Caswell, 1989; Lomnicki, 1988; Smith, hatchlings had significantly greater rates of 1990), community structure (Polis and McCor- aggression (ANOVA, F1,24 5 61.6, P , 0.001), mick, 1987; Cortwright, 1988; Holt and Polis, significantly greater mortality (ANOVA, F1,24 5 1997), and evolution (Polis and McCormick, 101.9, P , 0.001), and SVL (ANOVA, F1,24 5 1987; Kirkpatrick, 1984; Lomnicki, 1988; Sem- 42.6, P , 0.001) after 60 days than control larvae. litsch et al., 1988). In populations where intra- These differences were also significant for each guild predation or cannibalism occurs, the of the three species (Table 1). The consumption ecological effects of size structure are more of a conspecific was directly observed in all three pronounced because the ability to be an intra- species during the observation periods (twice guild predator is often limited by the size INTRAGUILD PREDATION ON CONGENER SALAMANDERS 25 difference between the predator and prey (Fox, CRUMP, M. L. 1992. Cannibalism in amphibians. In M. 1975; Polis and McCormick, 1987; Polis et al., A. Elgar and B. J. Crespi (eds.), Cannibalism: 1989; Ziemba and Collins, 1999). Because in- Ecology and Evolution among Diverse Taxa, pp. traguild predation and cannibalism increase 256–276. Oxford Univ. Press, Oxford. DONG, Q., AND G. A. POLIS, 1992. The dynamics of with density in Ambystoma salamander larvae, cannibalistic populations: a foraging perspective. In these may be important components of density- M. A. Elgar and B. J. Crespi (eds.), Cannibalism: dependent population regulation (Lannoo and Ecology and Evolution among Diverse Taxa, pp. Bachmann, 1984; Crump, 1992; Maret and 13–37. Oxford Univ. Press, Oxford. Collins, 1994; Ziemba and Collins, 1999). FINKE,D.L.,AND R. F. DENNO. 2002. Intraguild I found that the experimental treatment of predation diminished in complex-structured vege- feeding a single congener to a potential intra- tation: implications for prey suppression. Ecology guild predator generated size variation that 83:643–652. resulted in a greater number of aggressive FOX, L. R. 1975. Cannibalism in natural populations. behaviors toward conspecifics in three species. Annual Review of Ecology and Systematics 6:87– These results are consistent with studies on 106. cannibalism that show that size differences GORBMAN, A. 1964. Endocrinology of the Amphibia. In J. A. Moore (ed.), Physiology of the Amphibia, pp among members of a population determine the 371–425. Academic Press, New York. rate of cannibalism (Polis, 1981; Polis et al., 1989; HOLT, R. D., AND G. A. POLIS, 1997. A theoretical Dong and Polis, 1992; Larson et al., 1999; Ziemba framework for intraguild predation. American and Collins, 1999). Naturalist 149:745–764. I hypothesize that intraguild predation causes JASKULA, J., AND R. BRODMAN. 2000. Density-dependent increased size variation, which leads to in- effects on microhabitat selection and activity of two creased aggression and mortality caused by species of larval Ambystoma salamanders. In H. cannibalism. Because intraguild predation de- Kaiser, G. S. Casper, and N. Bernstein (eds.), creases in complexly structured habitat (Brod- Investigating Amphibian Declines: Proceedings of man, 1996; Finke and Denno, 2002), habitat the 1998 Midwest Declining Amphibian Confer- structure may also affect cannibalism rates in ence. Journal of the Iowa Academy of Science. populations of Ambystoma with large size vari- Special Issue 107:146–150. ation of larvae. Future studies on intraguild KEEN, W. H., J. TRAVIS, AND J. JUILIANA. 1984. Larval predation need to document rates of cannibalism growth in three sympatric Ambystoma salamander species: species differences and the effect of as habitat complexity and food levels vary and temperature. Canadian Journal of Zoology should focus on behavior throughout develop- 62:1043–1047. ment. KIRKPATRICK, M. 1984. Demographic models based on size, not age, for organisms with indeterminant Acknowledgments.—I thank J. Jaskula for help- growth. Ecology 65:1874–1884. ing to collect eggs and maintain larvae in the lab LANNOO, M. J., AND M. D. BACHMANN. 1984. Aspects of and K. Tiede and H. Hall for assistance in cannibalistic morphs in a population of Ambystoma measuring SVLs. I thank C. Wong and M. Seely t. tigrinum larvae. American Midland Naturalist for help with statistical analyses. I also thank C. 112:103–109. Wistrom for insightful comments on an earlier LANNOO, M. J., L. LOWCOCK, AND J. P. BOGART. 1989. Sibling cannibalism in noncannibal morph Ambys- draft of this manuscript. toma tigrinum larvae and its correlation with high growth rates and early metamorphosis. Canadian Journal of Zoology 67:1911–1914. LITERATURE CITED LARSON, K. L., W. DUFFY,E.JOHNSON,M.F.DONOVAN, BRODMAN, R. 1995. Annual variation in breeding AND M. J. LANNOO. 1999. ‘‘Paedocannibal’’ morph success of two syntopic species of Ambystoma barred tiger salamanders (Ambystoma tigrinum salamanders. Journal of Herpetology 29:111–113. mavortium) from eastern South Dakota. American ———. 1996. Effects of intraguild interactions on Midland Naturalist 141:124–139. fitness and microhabitat use of larval Ambystoma LOMNICKI, A. 1988. Population Ecology of Individuals. Princeton Univ. Press. Princeton, NJ. salamanders. Copeia 1996:372–378. LUCAS, E., D. CODERRE, AND J. BRODEUR. 1998. Intraguild ———. 1999. Food and space dependent effects during predation among aphid predators: characterization the interactions of two species of larval salaman- and influence of extraguild prey density. Ecology ders. Journal of Freshwater Ecology 14:431–437. 79:1084–1092. CASWELL, H. 1989. Matrix population models: construc- MARET,T.J.,AND J. P. COLLINS. 1994. Individual tion, analysis, and interpretation. Sinauer Asso., responses to population size structure: the role of Inc., Sutherland, MA. size variation in controlling expression of a trophic CORTWRIGHT, S. A. 1988. Intraguild predation and polyphenism. Oecologia 100:279–285. competition: an analysis of net growth shifts in MINTON, S. A. 1972. Amphibians and Reptiles of larval amphibian prey. Canadian Journal of Zool- Indiana. Indiana Academy of Science Monograph, ogy 66:1813–1821. no 3. Indiana Academy of Science, Indianapolis. 26 ROBERT BRODMAN

MORAN, M. D., AND L. E. HURD. 1997. Relieving food field and laboratory experiments. Journal of Her- limitation reduces survivorship of a generalist petology 17:210–220. predator. Ecology 78:1266–1270. TALENTINO, K., AND E. LANDRE. 1991. Comparative de- PECKARSKY, B. L., B. W. TAYLOR,A.RMCINTOSH,M.A. velopment of two species of sympatric Ambystoma MCPEEK, AND D. A. LYLE. 2001. Variation in mayfly salamanders. Journal of Freshwater Ecology size at metamorphosis as a developmental response 6:395–401. to risk of predation. Ecology 82:740–757. VAN BUSKIRK,J.,AND D. C. SMITH. 1991. Density- PERSSON, L. 1988. Assymmetrics in competitive and dependent population regulation in a salamander. predatory interactions in fish populations. In B. Ecology 72:1747–1756. Ebenman and L. Persson (eds.), Size Structured WALLS, S. C. 1998. Density dependence in a larval Populations, pp. 203–218. Springer, New York. salamander: the effects of interference and food POLIS, G. A. 1981. The evolution and dynamics of intraspecific predation. Annual Review of Ecology limitation. Copeia 1998:926–935. and Systematics 12:225–251. WILBUR, H. M. 1972. Competition, predation, and the POLIS, G. A., AND S. J. MCCORMICK. 1987. Intraguild structure of the Ambystoma-Rana sylvatica commu- predation and competition among desert scorpions. nity. Ecology 53:3–21. Ecology 68:332–343. WILBUR, H. M., AND J. P. COLLINS. 1973. Ecological POLIS, G. A., C. A. MYERS, AND R. D. HOLT. 1989. The aspects of amphibian metamorphosis. Science ecology and evolution of intraguild predation: 182:1305–1314. Potential competitors that eat each other. Annual WILDY, E. L., D. P. CHIVERS,J.M.KIESECKER, AND A. R. Review of Ecology and Systematics 20:297–330. BLAUSTEIN. 1998. Cannibalism enhances growth in SCHELLHORN, N. A., AND D. A. ANDOW. 1999. Cannibal- larval long-toed salamanders (Ambystoma macro- ism and interspecific predation: role of oviposition dactylum). Journal of Herpetology 32:286–289. behavior. Ecological Applications 9:418–428. WISSINGER,S.A.1989.Seasonalvariationinthe SCOTT, D. E. 1990. Effects of larval density in Ambystoma intensity of competition and predation among opacum: an experiment in large-scale field enclo- dragonfly larvae. Ecology 70:1017–1027. sures. Ecology 71:296–306. ———. 1992. Niche overlap and the potential for SEMLITSCH, R. D., D. E. SCOTT, AND J. H. K. PECHMANN. competition and intraguild predation between size- 1988. Time and size at metamorphosis related to structured populations. Ecology 73:1431–1444. adult fitness in Ambystoma talpoideum. Ecology WISSINGER, S. A., G. B. SPARKS,G.L.ROUSE,W.S.BROWN, 69:184–192. AND H. STELTZER. 1996. Intraguild predation and SMITH, C. K. 1990. Effects of variation in body size on cannibalism among larvae of detritivorous caddis- intraspecific competition among larval salaman- flies and subalpine wetlands. Ecology 77:2421– ders. Ecology 71:1777–1788. 2430. STANGEL, P. W. 1988. Premetamorphic survival of the ZIEMBA, R. E., AND J. P. COLLINS. 1999. Development of salamander Ambystoma maculatum in eastern Mas- size structure in tiger salamanders: the role of sachusetts. Journal of Herpetology 22:345–347. intraspecific interference. Oecologia 120:524–529. STENHOUSE, S. L., N. G. HAIRSTON, AND A. E. COBEY. 1983. Predation and competition in Ambystoma larvae: Accepted: 4 October 2003.