Age Estimation and Dispersal in the Spotted Hyena (Crocuta Crocuta)

Journal of Mammalogy, 84(3):1019±1030, 2003

AGE ESTIMATION AND DISPERSAL IN THE SPOTTED HYENA (CROCUTA CROCUTA)

RUSSELL C. VAN HORN,* TERESA L. MCELHINNY, AND KAY E. HOLEKAMP Department of Zoology, Michigan State University, 203 Natural Science Building, East Lansing, MI 48824, USA Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 We used observations of known-age, free-ranging spotted hyenas (Crocuta crocuta) from a large social group to develop age-estimation models. A model based on tooth-eruption data estimates the ages of animals 10.0±15.5 (Ϯ1.1 SD) months old. We used tooth-wear data to estimate the ages of adult males Ϯ4.9 months and ages of females Ϯ22.6 months. Analysis of known and estimated ages shows that males usually disperse from their natal group when 24±60 months of age. Eight of 20 males whose fates were known lived in 3 groups over time, and at least 7 of 41 resident immigrant males appeared to arrive in the study group years after leaving their birthplaces. Thus, males of this species often engage in multiple dispersal events.

Key words: age estimation, Crocuta crocuta, dispersal, spotted hyenas

Age in¯uences many aspects of mam- mation of annuli may be irregular in C. cro- malian biology, including physiology cuta (Lindeque and Skinner 1984) and oth- (Blom et al. 1994), epidemiology (Mills et er carnivores inhabiting constant environ- al. 1999), and behavior (Bernstein and ments, as is often the case in equatorial re- Ehardt 1985). Because anthropogenic dis- gions. Also, this method requires tooth turbances may alter age-speci®c mortality extraction, which may reduce future feed- and population structure (e.g., Hofer et al. ing ef®ciency of the subject (Van Valken- 1993), the ability to estimate the ages of burgh 1988). animals can facilitate their conservation. In Tooth-wear data represent another poten- addition, age data can enhance understand- tial basis for age estimation in live African ing of mammalian evolution because age it- carnivores (Smuts et al. 1978; Stander self may be an important parameter in evo- 1997). Such data can be collected less in- lutionary processes such as sexual selection trusively than data on annuli, although both (Brooks and Kemp 2001). require immobilization of subjects. Kruuk Counts of dental annuli are frequently (1972), in his seminal work on spotted hy- used to estimate the ages of mammals, in- ena behavior and ecology, had no data from cluding carnivores (Driscoll et al. 1985; known-age animals but assigned individu- Fandos et al. 1993; Landon et al. 1998; als to 1 of 5 relative age classes based on Spinage 1973). Annuli were used by van wear of p3. Lindeque and Skinner (1984) Jaarsveld et al. (1987) to estimate the ages later used the surface area of p3 to distin- of spotted hyenas (Crocuta crocuta) living guish 7 relative age classes among South at relatively high latitudes in southern Af- African C. crocuta. Relationships between rica. These age estimates have been used in dental morphology and age in captive spot- analyses of growth and development (van ted hyenas were described by Binder and Jaarsveld et al. 1987, 1988). However, for- Van Valkenburgh (2000), but the diet of * Correspondent: [email protected] captive animals is softer and probably more

1019 1020 JOURNAL OF MAMMALOGY Vol. 84, No. 3 nutritious than that of free-living C. crocuta Our 1st goal was to develop and validate (Berger et al. 1992), so these relationships age-estimation models for C. crocuta using may differ in the wild. Tooth-wear criteria data from known-age wild individuals. Our may overestimate the ages of young ani- 2nd goal was to test the hypothesis that sec- mals and underestimate the ages of older ondary dispersal occurs in spotted hyenas animals (Spinage 1973), so model valida- and thus that not all interclan transfer is pri- tion is particularly important. Known-age mary dispersal. To accomplish this, we data from free-ranging C. crocuta have not compared the ages of immigrating and em- been used to build or validate any age-es- igrating males to determine what proportion timation models, although such validation might be secondary dispersers. Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 is desirable (e.g., Gipson et al. 2000; Harris et al. 1992; Oosthuizen and Bester 1997). MATERIALS AND METHODS The ability to estimate absolute age of spot- We focused on 1 clan of C. crocuta inhabiting ted hyenas would facilitate the investigation the Talek region (1Њ40ЈS, 35Њ50ЈE) of the Masai of biological phenomena dependent on mat- Mara National Reserve, Kenya, in open grass- uration, such as dispersal. land (Frank 1986). From June 1988 through Natal or primary dispersal is the com- June 2001 this clan was observed for 23±31 plete and permanent departure of an indi- days per month, except for April 1991 when hy- vidual from its birthplace (Greenwood enas were monitored for only 14 days. All in- 1980). Secondary dispersal is any dispersal dividuals in the study clan could be recognized movement occurring after natal dispersal by their unique spot patterns. Spotted hyenas are strongly monomorphic in size and appearance (Pusey and Packer 1987) and is relatively (Hamilton et al. 1986), but we were able to sex common among small mammals (reviewed animals in this study by the dimorphic mor- as `transfer' dispersal in Cockburn 1992) phology of their erect phalluses (Frank et al. and some primates (reviewed in van Noord- 1990). Ages of cubs when 1st observed were wijk and van Schaik 2001). However, it is estimated to Ϯ7 days based on pelage, size, and considered uncommon among large mam- behavior (Holekamp and Smale 1998). mals (Sinclair 1992), and the extent to Hyenas were anesthetized with Telazol (W. A. which it occurs among carnivores is cur- Butler Company, Brighton, Michigan; 2.5 mg/kg) rently unknown. administered in a dart using a CO2-powered ri¯e Female spotted hyenas rarely disperse (Telinject Inc., Saugus, California). Most hyenas from the clan, which includes adult natal (76.8%, n ϭ 151) were darted only once; but to accomplish other research objectives, some hye- females and their offspring and 1 to several nas were darted repeatedly. Cubs typically were adult immigrant males (Holekamp et al. darted when 8±12 months of age (n ϭ 63). Natal 1993). However, nearly every male even- males also were darted when Ն24 months of age tually emigrates and assumes a low rank in (n ϭ 14), and natal females were darted again another clan (East and Hofer 2001; Hen- when we believed them to be pregnant (n ϭ 36). schel and Skinner 1987; Holekamp and Immigrant males were considered resident if they Smale 1998; Smale et al. 1997) after pu- stayed in the clan Ն6 months (n ϭ 58), and most berty at about 24 months of age (Matthews resident males (70.7%) were darted. Nonresident 1939). Natal dispersal appears necessary for immigrant males (i.e., those that stayed in the males to achieve reproductive success clan Յ6 months) were excluded from all analy- (Engh et al. 2002). Before both primary and ses. Eighteen resident immigrant males and 13 natal males were radiocollared. Those natal males secondary dispersal, males use the range of known from radiotelemetry data to establish res- their current clan as a secure base from idency in 1 clan (i.e., remained in a non-Talek which to explore (Smale et al. 1997; E. E. clan for Ն6 months) before moving to a 2nd clan Boydston, in litt.), then transfer directly into were classed as secondary dispersers, as were ra- a new clan; they do not become nomads (E. diocollared resident Talek immigrant males that E. Boydston, in litt.). subsequently established residency in another August 2003 VAN HORN ET AL.ÐAGE ESTIMATION AND DISPERSAL IN CROCUTA 1021 clan. Smale et al. (1997) used radiotelemetry to appreciable effect on model performance and so demonstrate that dispersing male hyenas usually were retained for model validation. take exploratory forays outside their natal home Multicollinearity among predictor variables range before emigrating. Therefore, adult natal and low ratios of sample size to predictor vari- males without radiocollars were considered dis- ables can produce spurious regressions (Neter et persers only if they had engaged in such excur- al. 1996). We conducted principal components sions before their disappearance from the clan, if analyses (PCA) on the correlation matrix of each they were in good health when last seen, and if model-building data set (Morrison 1990) to re- they were last seen in the Talek area when Ն24 duce iteratively the collinearity and dimension- months of age. If these conditions were met, the ality of the data while retaining the maximum estimated age of primary dispersal (n ϭ 26) was amount of information. Multiple linear regres- Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 the age of the male when he was last seen in the sion models were built with the remaining pre- Talek territory before his 1st absence of Ն6 dictor variables (Appendix I). Various transfor- months. mations (e.g., log, ln) of predictor and response Thirty-one morphological measurements (15 data (i.e., age in months) were examined. No dental, 4 cranial, and 12 postcranial; Appendix I) variable-selection method will automatically were collected from immobilized hyenas (n ϭ ®nd the most appropriate predictor variables 228 dartings of 165 animals) and those found (Neter et al. 1996). Therefore, we chose the ®nal dead (i.e., n ϭ 33 necropsies, including 25 ani- predictors based on congruence of forward, re- mals not darted). Incomplete observations were verse, and stepwise variable selection proce- discarded, and measures from paired ipsilateral dures while minimizing the residual mean- teeth (e.g., height of left and right p3) were av- square error, maximizing the coef®cient of mul- eraged. The presence of deciduous teeth was not- tiple determination (R2), and minimizing the ed, as was the sequence and timing of tooth erup- number of predictors. tion and replacement. Maternal rank at parturition We assessed a model by examining the ac- was tested as a predictor of the age at initiation curacy (i.e., average discrepancy between esti- of tooth replacement among offspring in post hoc mated and known age) and precision (i.e., SD of quantitative analyses. We chose maternal rank be- estimation error) of the age estimates it produced cause it affects growth rates of cubs and ages at from the validation data set. We deemed a model weaning (Hofer and East 1996; Holekamp et al. acceptably accurate if the average estimation er- 1996) and thus indicates cub nutrition. ror was Յ6 months and acceptably precise if the Darting data were sorted based on whether the SD of the error was Յ12 months. Our initial goal permanent dentition was complete (i.e., subadult in model building was to achieve these levels of or adult). To avoid circularity and nonindepen- accuracy and precision, but because none of the dence during model validation, data were sub- 1st multiple linear regression models met these divided further into model-building and model- criteria for estimating ages of animals with fully testing, or validation, data sets. Observations erupted permanent dentition, a 2nd approach from animals without fully erupted permanent was used. We examined sexual dimorphism in dentition were divided randomly into 2 samples dental morphology by combining all model- of equal size for model construction (i.e., sub- building and validation data from observations adult model building) and evaluation (i.e., sub- of complete permanent dentition and then com- adult validation). Data from individuals with ful- paring known-age data from males and females ly erupted permanent dentition were collected for variables previously used in estimating the from August 1989 through December 1999. If ages of C. crocuta (Kruuk 1972) and likely to an animal was darted repeatedly, 1 darting was vary with age (i.e., height and occlusal length of chosen at random to represent that individual in p3, height of upper and lower canines). We com- the adult model-building data set. The remaining pared males and females with Mann±Whitney adult records were combined with observations tests in age classes where suf®cient samples ex- collected from January to June 2001 to form the isted for comparison. We set ␣ϭ0.05 for all adult validation data set. These data include a signi®cance tests and used sequential Bonferroni few nonindependent points (e.g., repeated dart- adjustments to control for the effects of multiple ings from 12 animals were included in both comparisons (Rice 1989). Because hyenas are building and validation data sets), which had no monomorphic (this study; Hamilton et al. 1986) 1022 JOURNAL OF MAMMALOGY Vol. 84, No. 3 and known-age data for females were more ex- TABLE 1.ÐTooth eruption in 144 known-age, tensive and variable than those for males, a free-living Crocuta crocuta (from 119 dartings, model to estimate ages of males was built using 25 necropsies). data from females and tested with data from males. Correlations between estimated and Age range Dentition type (months) n known ages, and known ages and estimation error, were compared between models. Statistical No adult teeth 0.5±7.5 25 analyses were conducted with statistical analysis Only adult I1 present 6.5±9.0 16 software (SAS Institute Inc. 1999). Adult I1 and I2 present 7.8±12.0 43 We were able to create and test an age esti- Some adult cheek teeth present 10.0±15.0 38

All adult teeth present 13.0±18.0 22 Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 mator based on the relationship Binder (1998) observed between height of C1 and age in young (Յ40 months old) captive hyenas. We also assessed the estimator developed by van Jaarsveld Maternal rank at cub's birth, a proxy for et al. (1987) that predicted the height (in mm) nutritional state, was not associated signif- of p3 from the number of visible dentine lines icantly with the age at which the 1st decid- in cross sections of p3, with the latter as a proxy uous teeth were replaced (r ϭϪ0.074, n ϭ for age. 53, P ϭ 0.596). Replacement of deciduous Using the best age-estimation model for adult teeth begins with the incisors (1st, 2nd, and males, we estimated age at darting of resident 3rd upper incisors and 1st, 2nd, and 3rd immigrant males and accounted for time elapsed lower incisors in series), is followed by the since arrival in Talek (Holekamp and Smale cheek teeth (1st, 4th, 3rd, and 2nd upper 1998) to derive an estimated age at immigration. premolars and 1st lower molar, 2nd, 4th, Estimated ages at immigration and known ages of emigration were compared with a 2-tailed chi- and 3rd lower premolar in series), and ends square test for homogeneity (Daniel 1990). Res- with the canines. The last deciduous teeth ident immigrant males were considered second- were replaced by 13±18 months of age. ary dispersers into the Talek clan if their esti- We found no differences between males mated age at arrival was greater than the sum of and females of the same age in height (in the maximum age of natal emigrants and the mm) or occlusal length (in mm) of p3, or maximum model overestimation error; other- height (in mm) of upper or lower canines wise they were considered primary dispersers. (all P Ͼ 0.05; Fig. 1). Therefore, data from This produces a conservative estimate of the fre- both sexes were pooled. quency of secondary dispersal. Values are pre- The subadult model-building data set in- sented as mean Ϯ SD. cluded 19 (13 female, 6 male) known-age RESULTS observations (12.7 Ϯ 1.8 months) and was We collected 191 complete known-age validated with 19 (7 female, 12 male) 1.8 observations of 151 spotted hyenas (80 fe- known-age observations (12.7 Ϯ months). Selection procedures for variables males, 71 males), ranging in age from 0.2 in multiple linear regression suggested that to 198 months (females aged XÅ ϭ 28.3 Ϯ 2 variables be used: height (in mm) of the 46.1 months, males aged 17.1 Ϯ 14.1 months). Permanent dentition of C. crocuta 1st lower canine (c1) and height (in mm) includes i 3/3, c 1/1, p 4/3, m 0±1/1, total of the 1st upper canine (C1; Appendix I). 32±34. Deciduous dentition consists of i 3/ The best model for animals with incomplete 32.48, d.f. 2, 3, c 1/1, p 3/3, total 28. Substantial indi- permanent dentition (F ϭ ϭ 16, P 0.0001) predicted age as vidual variation in timing of tooth eruption Ͻ and replacement (Table 1) precluded a log (age in months) quantitative age estimator for cubs with 10 only deciduous teeth. Deciduous incisors ϭ 0.647 ϩ 0.158[log10 (C1)] and canines had erupted by birth, and deciduous cheek teeth erupted by 2 months. ϩ 0.238[log10 (c1)] (1) August 2003 VAN HORN ET AL.ÐAGE ESTIMATION AND DISPERSAL IN CROCUTA 1023 Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021

FIG. 1.ÐTooth measurements of known-age free-living Crocuta crocuta. White bars are values for males, gray bars for females, shown as means Ϯ SD. a) Height of p3, b) occlusal length of p3, c) height of c1, and d) height of C1. Variables are de®ned in Appendix I. Sample sizes are noted on bars. Two males were darted when Ͼ48 months old (i.e., 60 months, 88.5 months); measurements from these males lie within the range of values from females of similar age.

We found a positive correlation between set (18 female, 9 male; 58.8 Ϯ 39.5 known ages and estimates from this model months). Concordance among selection (Fig. 2). We also found a negative correla- methods for variables in multiple linear re- tion between known ages and this model's gression suggested that 2 predictor vari- estimation errors (Table 2), as is often seen ables be used: occlusal length (in mm) of in models based on tooth wear. Neverthe- p3 (OCCp3; Appendix I), and height (in less, this model performs acceptably. mm) of c1 (c1). The best model for mixed- The adult model-building data set of 60 sexed observations (F ϭ 168.58, d.f. ϭ 2, (38 female, 22 male) known-age observa- 57, P Ͻ 0.0001) predicted age as tions (56.6 Ϯ 40.7 months) was used to construct mixed-sex models to produce age age in months ϭ 34.281 ϩ 15.423(OCCp3) estimates for either females or males, which were tested with the adult validation data Ϫ 2.473(c1) (2) 1024 JOURNAL OF MAMMALOGY Vol. 84, No. 3

(Fig. 3). The association between estimation error and known age was weak and nonsigni®cant for both models (Table 2). The only trend in error was that both models were less precise for animals Ն72 months old (Fig. 3). No model built with data from males to predict ages of females performed as well as models 2 or 3, but this was expected because fewer males were

sampled at many ages. Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 The best model (F ϭ 217.59, d.f. ϭ 3, 56, P Ͻ 0.0001) for estimating ages of adult males (n ϭ 28, 30.4 Ϯ 15.4 months) was built with data on known-age females (n ϭ 56, 73.8 Ϯ 41.5 months) and estimated age FIG. 2.ÐAssociation of known and estimated ages (n ϭ 19) of male and female Crocuta cro- of males as cuta with erupting permanent dentition from regression (equation 1) of log10(age in months) on log10 (age in months) log10(height of C1) and log10(height of c1). Solid line indicates theoretically perfect association. ϭ 1.237 Ϫ 0.070(OCCp3) ϩ 0.040(OCCp3)23Ϫ 0.002(OCCp3) However, spotted hyenas often break c1 (4) (Van Valkenburgh 1988). If c1 was excluded from the analysis, the resulting models This model produced a positive correlation were less accurate and precise. The best of between estimated and known ages (Fig. 4) these models (F ϭ 412.27, d.f. ϭ 1, 58, P and a negative correlation between known Ͻ 0.0001) was ages and estimation errors (Table 2). The model met our overall accuracy and preci- log (age in months) 10 sion criteria (Table 2) and did not decline ϭ 1.004 ϩ 0.116(OCCp3) (3) in precision among the oldest males, which are much younger than the oldest females Both mixed-sex models have acceptable ac- (Fig. 4). curacy (Table 2) and produce estimates that Model 4 was used to estimate ages at ar- are correlated positively with known ages rival of resident Talek immigrants (45.4 Ϯ

TABLE 2.ÐValidation of models to estimate age in free-living Crocuta crocuta from tooth measurements, including number of observations used in model derivation (nd) and validation (nv), with the residual mean-square error and coef®cient of multiple determination (R2) from each equation. A negative error indicates that the regression underestimated the known age. See text for equations describing models.

Association Model Residual Estimation error (months) of age and error estimating mean square 2 Å age nd nv error R X SD Range rP (1) 19 19 0.911 0.805 0.11 1.13 Ϫ2.11±1.62 Ϫ0.616 0.005 (2) 60 27 15.762 0.855 1.07 22.63 Ϫ63.51±55.98 Ϫ0.356 0.069 (3) 60 27 0.104 0.877 Ϫ2.65 26.56 Ϫ82.34±65.31 Ϫ0.126 0.530 (4) 56 28 0.085 0.920 Ϫ0.54 4.94 Ϫ12.96±10.91 Ϫ0.411 0.029 August 2003 VAN HORN ET AL.ÐAGE ESTIMATION AND DISPERSAL IN CROCUTA 1025 Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021

FIG. 4.ÐAssociation of known and estimated ages (n ϭ 28) of male Crocuta crocuta from

regression (equation 4) of log10(age in months) on occlusal length of p3. Solid line indicates theoretically perfect association.

months) than the next oldest disperser, and missed our absence cutoff by only 1.1% (i.e., 2 days Ͻ 6 months), we consider his dispersal age an outlier. Based on the other observed ages at dispersal and model estimation error, resident immigrants who were estimated to be Յ70.1 months old at arrival were considered primary dispersers. Most resident immigrants appeared to be primary dispersers (n ϭ 34, 37.6 Ϯ 12.4 months), but 17% (n ϭ 7, 83.0 Ϯ 11.9 months) were

FIG. 3.ÐAssociation of known and estimated ages (n ϭ 27) of male and female Crocuta crocuta a) from regression (equation 2) of age (in months) on height of p3 and height of c1, and b) from regression (equation 3) of log10(age in months) on height of p3. Solid line indicates theoretically perfect association.

21.1 months). These estimated ages at im- FIG. 5.ÐComparison of age distributions of migration were different (Fig. 5) from Talek emigrants (black bars, n ϭ 40) and resi- known ages at emigration from Talek (n ϭ dent immigrant (white bars, n ϭ 41) male Cro- 40, 42.1 Ϯ 10.5 months). Because the old- cuta crocuta at dispersal. Resident immigrants est (i.e., 75.9 months) disperser from Talek were those that remained in the Talek clan for was more than 1 SD older (i.e., 15.3 Ն6 months. 1026 JOURNAL OF MAMMALOGY Vol. 84, No. 3 evidently engaging in secondary dispersal the age at which tooth replacement was ini- when they joined the Talek clan. Fates of tiated. Nonetheless, maternal rank may play 20 males were known using radiotelemetry, a greater role in dental development in cubs which revealed that secondary dispersal oc- when females are under greater nutritional curred among a substantial minority (40%) stress than in this study. of both natal (2 of 9) and resident immi- We tested an estimator derived from grant (6 of 11) males. Small sample sizes Binder (1998) with data from free-ranging and methodological differences precluded animals similar in age to those she studied comparison of the rates of secondary dis- (n ϭ 12, 26.7 Ϯ 8.2 months). This model persal between natal and resident immigrant produced estimates, from the height of per- Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 males. manent C1, not correlated with known ages (r ϭ 0.276, P ϭ 0.386), and it consistently DISCUSSION underestimated known ages (error ϭϪ23.0 The sequence of tooth replacement ob- Ϯ 7.9 months). The mixed-sex model of served in this study matched that seen in C. van Jaarsveld et al. (1987) produced esti- crocuta by Slaughter et al. (1974), includ- mates that were positively correlated with ing variability in replacement of P2 and p2. known ages (r ϭ 0.756, P Ͻ 0.0001), but As we expected based on earlier studies it overestimated the age of all but 1 indi- (Frank et al. 1991; Pournelle 1965; Schnei- vidual (n ϭ 27, error ϭ 73.2 Ϯ 34.8 der 1926), the deciduous incisors and ca- months). This model yielded no association nines were erupted at birth. Pournelle between estimation error and known age (r (1965) noted the eruption of deciduous ϭ 0.02, P ϭ 0.92). cheek teeth at 31 days of age, consistent van Jaarsveld et al. (1988) described with our observation of the eruption of strong relationships between ages estimated those teeth by 2 months of age. Binder and from dental annuli and morphometric data Van Valkenburgh (2000) noted that adult (e.g., shoulder height, body weight, total dentition of captive spotted hyenas might body length) in South African spotted hy- be fully erupted at younger ages (i.e., 12± enas, but these morphological measures 14 months) than in wild conspeci®cs. They were not the best predictors of age in Talek observed ages that agree with those of hyenas. Rather, age of Talek hyenas was Kruuk (1972) from wild C. crocuta but best estimated by dental measurements. which are younger than those observed in Spotted hyenas crush and consume large this study or by Mills (1990). The faster amounts of bone (Kruuk 1972) using p3 eruption observed in captive animals might (Biknevicius 1996; Van Valkenburgh result from improved nutrition. The softer, 1996), so it is not surprising that wear of yet more nutritious diet of young (i.e., Յ40 this tooth re¯ects age. Another advantage months old), captive C. crocuta (Berger et to using p3 in an age estimator for C. cro- al. 1992) also might have caused the re- cuta is that breakage of this tooth is less gression derived from Binder (1998) to uncommon than that of other teeth (Van Val- derestimate the ages of young free-ranging kenburgh 1988). We recommend that re- C. crocuta. This suggests that variation in searchers calculate the mean anterior±pos- nutritional status among wild C. crocuta terior occlusal length (in mm) of both right may in¯uence the rate of tooth eruption and and left p3 to reduce estimation error due replacement and explain some of the esti- to breakage and asymmetrical tooth wear. mation error of our model for animals with The latter appears common in wild C. cro- subadult dentition. We could not directly cuta (Lindeque and Skinner 1984). measure the nutritional status of wild C. Overestimation of ages with the equation crocuta cubs, but a surrogate measure of of van Jaarsveld et al. (1987) is consistent this, maternal rank, was not associated with with the observation that the maximum es- August 2003 VAN HORN ET AL.ÐAGE ESTIMATION AND DISPERSAL IN CROCUTA 1027 timated age for C. crocuta in that study was thors (Hamilton et al. 1986; van Jaarsveld 7 years greater than the maximum known et al. 1988) who have suggested that sexual age among Talek hyenas (i.e., about 24 dimorphism in C. crocuta is very slight. In- years versus 17 years). This overestimation dividual variation in tooth wear has been might be due to irregular formation of den- observed in roe deer (Hewison et al. 1999) tine lines or to differences in tooth wear and red fox (Harris 1978) and linked to var- between spotted hyenas in Kenya and South iation in enamel mineralization within red Africa. Such interpopulation variation has deer (Cervus elaphus) populations (Kier- been observed in roe deer (Capreolus ca- dorf and Becher 1997). We propose that in- preolusÐHewison et al. 1999) and Spanish dividual variation is more important than Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 ibex (Capra pyrenaicaÐFandos et al. sexual dimorphism in shaping tooth wear 1993) but has not been seen in carnivores among spotted hyenas. Social rank may in- (Gipson et al. 2000; Harris 1978; Smuts et ¯uence diet composition and cumulative al. 1978). differences in tooth wear, but we cannot ad- Spinage (1973) predicted that tooth wear dress this with our current data. would overestimate the ages of young ani- Natal Talek males disperse at approxi- mals and underestimate the ages of old an- mately the same ages as males in the ad- imals. This has been observed in roe deer jacent Serengeti ecosystem (East and Hofer (Hewison et al. 1999) and with our best es- 2001). Talek males do not become nomadic timator of male age (i.e., equation 4). None- (E. E. Boydston, in litt.), so older immi- theless, we ®nd the error of our model ac- grants in this area must be secondary dis- ceptable, and this is the model we recom- persers. Secondary dispersal occurs fre- mend for estimating the ages of adult male quently among male C. crocuta in our study C. crocuta. Previous researchers experi- area, so rank of immigrant males is not as enced dif®culty in estimating the ages of closely linked to age as in the Serengeti older gray wolves (Canis lupusÐLandon et (East and Hofer 1991), where secondary al. 1998), red deer (Cervus elaphusÐ dispersal appears to occur rarely (Hofer et Brown and Chapman 1991), and fallow al. 1993). Secondary dispersal in the Ser- deer (Dama damaÐBrown and Chapman engeti may be constrained by more limited 1990), and our best estimator of female opportunities for dispersal (E. E. Boydston, ages (i.e., equation 2) also had decreased in litt.). Genetic analyses have shown that precision among older individuals. Esti- primary dispersal by males is a near-uni- mates from this model are less precise after versal prerequisite for reproductive success an age (i.e., 72 months) older than most in Talek (Engh et al. 2002), so we suggest (i.e., 76%) of the animals whose measure- that males also undergo secondary dispersal ments were used to build and test the to improve their reproductive success. mixed-sex models. Therefore, we conclude However, we do not know whether second- that our model provides realistic estimates ary dispersers accrue increased reproduc- across most of the lifespan of wild C. cro- tive bene®ts in their new clans. cuta. We recommend this equation to estimate the ages of female spotted hyenas or ACKNOWLEDGMENTS hyenas of unknown sex. We thank the Of®ce of the President of Kenya Sex affects age estimation in white-tailed for permission to conduct this research. We also deer (Odocoileus virginianusÐVan Deelen thank the Kenya Wildlife Service, the National et al. 2000). Because we found no sexual Museums of Kenya, the Narok County Council, dimorphism in those variables that best re- and the Senior Warden of the Masai Mara Na- ¯ected and predicted age and because our tional Reserve for their cooperation. C. W. best estimator of male age was built with Ramm was instrumental in the early stages of data from females, we agree with those au- analyses, but his untimely death prevented his 1028 JOURNAL OF MAMMALOGY Vol. 84, No. 3 participation in the ®nal analyses. We are grate- DANIEL, W. W. 1990. Applied nonparametric statistics. ful to S. Winterstein for statistical advice. We PWS-Kent Publishing Company, Boston, Massachu- setts. appreciate the assistance and comments of M. DRISCOLL, K. M., G. S. JONES, AND F. N ICHY. 1985. An Szykman, A. Engh, B. Lundrigan, E. Boydston, ef®cient method by which to determine age of car- S. Wahaj, and 2 anonymous reviewers. The fol- nivores, using dentine rings. Journal of Zoology lowing individuals assisted with data collection: (London) 205:309±313. EAST,M.L.,AND H. HOFER. 1991. Loud-calling in a N. Berry, S. Cooper, S. Dloniak, M. Durham, J. female-dominated mammalian society. II. Behav- Friedman, P. Garrett, I. Graham, K. Greene, T. ioural contexts and functions of whooping of spotted Harty, C. Katona, J. Kolowski, T. Morelli, K. hyaenas, Crocuta crocuta. Animal Behaviour 42: Nutt, G. Ording, J. Tanner, H. Watts, K. Weibel, 651±669. and B. White. This work was supported by Na- EAST,M.L.,AND H. HOFER. 2001. Male spotted hyenas Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 (Crocuta crocuta) queue for status in social groups tional Science Foundation grants IBN9309805, dominated by females. Behavioral Ecology 12:558± IBN9630667, IBN9906445, and IBN0113170. 568. ENGH,A.L.,ET AL. 2002. Reproductive skew among LITERATURE CITED males in a female-dominated mammalian society. Behavioral Ecology 13:193±200. BERGER, D. M. P., L. G. FRANK, AND S. E. GLICKMAN. FANDOS, P., J. F. ORUETA, AND Y. A RANDA. 1993. Tooth 1992. Unraveling ancient mysteries: biology, behav- wear and its relation to kind of food: the repercus- ior, and captive management of the spotted hyena, sion on age criteria in Capra pyrenaica. Acta The- Crocuta crocuta. Pp. 139±147 in Proceedings of the riologica 38:93±102. joint meeting of the American Association of Zoo FRANK, L. G. 1986. Social organization of the spotted Veterinarians and the American Association of hyaena (Crocuta crocuta). I. Demography. Animal Wildlife Veterinarians in Oakland, California. Behaviour 34:1500±1509. BERNSTEIN, I. S., AND C. EHARDT. 1985. Age-sex dif- FRANK, L. G., S. E. GLICKMAN, AND P. L ICHT. 1991. ferences in the expression of agonistic behavior in Fatal sibling aggression, precocial development, and rhesus monkey (Macaca mulatta) groups. Journal of androgens in neonatal spotted hyaenas. Science 252: Comparative Psychology 99:115±132. 702±704. BIKNEVICIUS, A. R. 1996. Functional discrimination in FRANK, L. G., S. E. GLICKMAN, AND I. POWCH. 1990. the masticatory apparatus of juvenile and adult cou- Sexual dimorphism in the spotted hyaena (Crocuta gars (Puma concolor) and spotted hyenas (Crocuta crocuta). Journal of Zoology (London) 221:308± crocuta). Canadian Journal of Zoology 74:1934± 313. 1942. GIPSON, P. S., W. B. BALLARD,R.M.NOWAK, AND L. BINDER, W. J. 1998. Functional aspects of tooth and D. MECH. 2000. Accuracy and precision of estimat- jaw development in large carnivores. Ph.D. disser- ing age of gray wolves by tooth wear. Journal of tation, University of California, Los Angeles. Wildlife Management 64:752±758. BINDER,W.J.,AND B. VAN VALKENBURGH. 2000. De- GREENWOOD, P. J. 1980. Mating systems, philopatry velopment of bite strength and feeding behaviour in and dispersal in birds and mammals. Animal Behav- juvenile spotted hyenas (Crocuta crocuta). Journal iour 28:1140±1162. of Zoology (London) 252:273±283. HAMILTON, W. J. I., R. T. TILSON, AND L. G. FRANK. BLOM, J. M. C., J. M. GERBER, AND R. J. NELSON. 1994. 1986. Sexual monomorphism in spotted hyenas, Day length affects immune cell numbers in deer Crocuta crocuta. Ethology 71:63±73. mice: interactions with age, sex, and prenatal pho- HARRIS, S. 1978. Age determination in the red fox toperiod. American Journal of Physiology 267: (Vulpes vulpes)Ðan evaluation of technique ef®- R596±R601. ciency as applied to a sample of suburban foxes. BROOKS, R., AND D. J. KEMP. 2001. Can older males Journal of Zoology (London) 184:91±117. deliver the good genes? Trends in Ecology and Evo- HARRIS, S., W. J. CRESSWELL, AND C. L. CHEESEMAN. lution 16:308±313. 1992. Age determination of badgers (Meles meles) BROWN,W.A.B.,AND N. G. CHAPMAN. 1990. The from tooth wear: the need for a pragmatic approach. dentition of fallow deer (Dama dama): a scoring Journal of Zoology (London) 228:679±684. scheme to assess age from wear of the permanent HENSCHEL,J.R.,AND J. D. SKINNER. 1987. Social re- molariform teeth. Journal of Zoology (London) 221: lationships and dispersal patterns in a clan of spotted 659±682. hyaenas Crocuta crocuta in the Kruger National BROWN,W.A.B.,AND N. G. CHAPMAN. 1991. The Park. South African Journal of Zoology 22:18±24. dentition of red deer (Cervus elaphus): a scoring HEWISON, A. J. M., J. P. VINCENT,J.M.ANGIBAULT,D. scheme to assess age from wear of the permanent DELORME,G.VAN LAERE, AND J. M. GAILLARD. molariform teeth. Journal of Zoology (London) 224: 1999. Tests of estimation of age from tooth wear on 519±536. roe deer of known age: variation within and among COCKBURN, A. 1992. Habitat heterogeneity and dis- populations. Canadian Journal of Zoology 77:58±67. persal: environmental and genetic patchiness. Pp. HOFER, H., AND M. L. EAST. 1996. The components of 65±95 in Animal dispersal: small mammals as mod- parental care and their ®tness consequences: a life els (N. C. Stenseth and W. Z. Lidicker, JR., eds.). history perspective. Verhandlungen der Deutschen Chapman & Hall, New York. Zoologischen Gesellschaft 89:149±164. August 2003 VAN HORN ET AL.ÐAGE ESTIMATION AND DISPERSAL IN CROCUTA 1029

HOFER, H., M. L. EAST, AND K. L. I. CAMPBELL. 1993. SCHNEIDER, K. M. 1926. UÈ ber Hyanenzucht. Die Peltz- Snares, commuting hyaenas and migratory herbi- tierzucht 2:1±14. vores: humans as predators in the Serengeti. Sym- SINCLAIR, A. R. E. 1992. Do large mammals disperse posia of the Zoological Society of London 65:347± like small mammals? Pp. 234±242 in Animal dis- 366. persal: small mammals as models (N. C. Stenseth HOLEKAMP, K. E., J. O. OGUTU,L.G.FRANK,H.T. and W. Z. Lidicker, Jr., eds.). Chapman & Hall, New DUBLIN, AND L. SMALE. 1993. Fission of a spotted York. hyena clan: consequences of female absenteeism and SLAUGHTER, B. H., R. H. PINE, AND N. E. PINE. 1974. causes of female emigration. Ethology 93:285±299. Eruption of cheek teeth in Insectivora and Carniv- HOLEKAMP,K.E.,AND L. SMALE. 1998. Dispersal sta- ora. Journal of Mammalogy 55:115±125. tus in¯uences hormones and behavior in the male SMALE, L., S. NUNES, AND K. E. HOLEKAMP. 1997. Sex- spotted hyena. Hormones and Behavior 33:205±216. ually dimorphic dispersal in mammals: patterns,

HOLEKAMP, K. E., L. SMALE, AND M. SZYKMAN. 1996. causes, and consequences. Advances in the Study of Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 Rank and reproduction in the female spotted hyaena. Behavior 26:181±250. Journal of Reproduction and Fertility 108:229±237. SMUTS, G. G., J. L. ANDERSON, AND J. C. AUSTIN. 1978. KIERDORF, U., AND J. BECHER. 1997. Mineralization Age determination of the African lion (Panthera and wear of mandibular ®rst molars in red deer (Cer- leo). Journal of Zoology (London) 185:115±146. vus elaphus) of known age. Journal of Zoology SPINAGE, C. A. 1973. A review of the age determina- (London) 241:135±143. tion of mammals by means of teeth, with especial KRUUK, H. 1972. The spotted hyaena: a study of pre- reference to Africa. East African Wildlife Journal dation and social behavior. University of Chicago 11:165±187. Press, Chicago, Illinois. STANDER, P. E. 1997. Field age determination of leop- LANDON, D. B., C. A. WAITE,R.O.PETERSON, AND L. ards by tooth wear. African Journal of Ecology 35: D. MECH. 1998. Evaluation of age determination 156±161. techniques for gray wolves. Journal of Wildlife VAN DEELEN, T. R., K. M. HOLLIS,C.ANCHOR, AND D. Management 62:674±682. R. ETTER. 2000. Sex affects age determination and LINDEQUE, M., AND J. D. SKINNER. 1984. Size frequen- wear of molariform teeth in white-tailed deer. Jour- cy analysis of tooth wear in spotted hyaenas Crocuta nal of Wildlife Management 64:1076±1083. crocuta. South African Journal of Zoology 19:291± VAN JAARSVELD, A. S., J. R. HENSCHEL, AND J. D. SKIN- 294. NER. 1987. Improved age estimation in spotted hy- MATTHEWS, L. H. 1939. Reproduction in the spotted aenas (Crocuta crocuta). Journal of Zoology (Lon- hyaena, Crocuta crocuta (Erxleben). Philosophical don) 213:758±762. Transactions of the Royal Society of London, Series VAN JAARSVELD, A. S., J. D. SKINNER, AND M. LIN- B, Biological Sciences 230:1±78. DEQUE. 1988. Growth, development and parental in- MILLS, J. N., T. G. KSIAZEK,C.J.PETERS, AND J. E. vestment in the spotted hyaena (Crocuta crocuta). CHILDS. 1999. Long-term studies of Hantavirus res- Journal of Zoology (London) 216:45±53. ervoir populations in the southwestern United VAN NOORDWIJK,M.A.,AND C. P. VAN SCHAIK. 2001. States: a synthesis. Emerging Infectious Diseases 5: Career moves: transfer and rank challenge decisions 135±142. by male long-tailed macaques. Behaviour 138:359± MILLS, M. G. L. 1990. Kalahari hyaenas: comparative 395. behavioral ecology of two species. Unwin Hyman, VAN VALKENBURGH, B. 1988. Incidence of tooth break- London, United Kingdom. age among large, predatory mammals. American MORRISON, D. F. 1990. Multivariate statistical methods. Naturalist 131:291±302. 3rd ed. McGraw-Hill Publishing Company, New VAN VALKENBURGH, B. 1996. Feeding behavior in free- York. ranging large African carnivores. Journal of Mam- NETER, J., M. H. KUTNER,C.J.NACHTSHEIM, AND W. malogy 77:240±254. WASSERMAN. 1996. Applied linear statistical methods. 4th ed. McGraw-Hill Publishing Company, Submitted 30 January 2002. Accepted 4 November New York. 2002. OOSTHUIZEN, W. H., AND M. N. BESTER. 1997. Com- parison of age determination techniques for known- Associate Editor was Ronald E. Barry. age Cape fur seals. South African Journal of Zool- ogy 32:106±111. APPENDIX I POURNELLE, G. H. 1965. Observations on birth and early development of the spotted hyena. Journal of The following measurements were taken from Mammalogy 46:503. immobilized or necropsied Crocuta crocuta. PUSEY, A. E., AND C. PACKER. 1987. Dispersal and philopatry. Pp. 250±266 in Primate societies (B. B. Numbers in parentheses indicate that measure- Smuts, D. L. Cheney, R. M. Seyfarth, R. W. Wran- ments were retained for entry into multiple lin- gham, and T. T. Struhsaker, eds.). University of Chi- ear regression analyses, as the intermediate stage cago Press, Chicago, Illinois. of model construction, with numbers indicating RICE, W. R. 1989. Analyzing tables of statistical tests. Evolution 43:223±225. subsequent ®nal model equation numbers (see SAS INSTITUTE INC. 1999. SAS/STAT guide. Version text). Dental measurements were taken on the 8.1. SAS Institute Inc., Cary, North Carolina. buccal tooth surface with calipers to within 0.1 1030 JOURNAL OF MAMMALOGY Vol. 84, No. 3 mm. Cranial and postcranial measurements were of external occipital protuberance, measured measured with cloth tape measures to within 0.1 along sagittal midline (equation 1). cm. Total mass was measured on a digital scale Top of sagittal crest to zygomatic arch: Per- to within 0.1 kg. Broken or decayed teeth were pendicular distance from sagittal crest to widest not measured, and crown heights were not re- point of zygomatic arch (equation 1). corded if gum erosion precluded accurate mea- Back of sagittal crest to zygomatic arch: Max- surement. Unless otherwise noted, dental and imum distance from posterior edge of external limb measurements were taken on only 1 side occipital protuberance to widest point of zygo- of the body. matic arch. Dental measurements.ÐHeight of left p3 and Ear: Maximum distance from notch at poste- height of right p3: Minimum crown height of rior edge of tragus to apex of pinna (equation Downloaded from https://academic.oup.com/jmammal/article/84/3/1019/903732 by guest on 30 September 2021 left and right p3, respectively, measured from 1). occlusal surface to base of enamel, as in Kruuk Postcranial measurements.ÐBody: Distance (1972:33, ®gure 7). from posterior edge of external occipital protu- p3: Mean of height of left and right p3 (equa- berance to base of tail, measured along spine. tions 2±4). Tail: Distance from base of tail to posterior Occlusal length of left and right p3: Maxi- tip of most caudal vertebra, measured along dor- mum anterior±posterior length of occlusal sur- sal midline. face of left and right p3, respectively. Head circumference: Maximum circumfer- Occlusal length of p3: Mean of occlusal ence of head, measured at widest point of zy- length of left and right p3 (equations 2±4). gomatic arches, perpendicular to sagittal crest. C1: Minimum crown height of C1, measured Neck circumference: Circumference of neck, from apex to base of enamel (equations 1±3). measured midway between shoulders and head (equation 1). c1: Minimum crown height of c1, measured Shoulder height: Maximum distance from from apex to base of enamel (equations 1±4). bottom of plantar pad of forepaw to cranial an- Incisors: Maximum width of medial four up- gle of scapula, measured with foreleg extended per incisors (equations 1±3). perpendicular to spine. P2: Maximum anterior±posterior length of P2, Scapula length: Maximum distance from cra- measured at base of enamel (equations 2, 3). nial angle of scapula to anterior tip of acromion P3: Maximum anterior±posterior length of P3, (equation 1). measured at base of enamel. Length of upper leg: Maximum distance from P4: Maximum anterior±posterior length of P4, anterior tip of greater tuberosity to posterior tip measured at base of enamel. of lateral epicondyle of humerus. p2: Maximum anterior±posterior length of p2, Length of lower leg: Maximum distance from measured at base of enamel. tip of olecranon of ulna to tip of accessory car- p3: Maximum anterior±posterior length of p3, pal. measured at base of enamel. Length of front foot: Maximum distance from p4: Maximum anterior±posterior length of p4, posterior tip of accessory carpal to anterior edge measured at base of enamel. of 3rd digital pad, measured along palmar mid- Upper toothrow: Maximum anterior±posterior line (equation 1). length of upper toothrow, measured from most Length of hind foot: Maximum distance from anterior edge of C1 to posterior edge of P4 posterior tip of tuber calcanei to anterior edge (equations 2, 3). of 3rd digital pad, measured along palmar mid- Lower toothrow: Maximum anterior±posterior line. length of lower toothrow, measured from ante- Girth: Minimum circumference of torso, mea- rior edge of c1 to posterior edge of p3. sured immediately posterior to forelegs, with Cranial measurements.ÐHead: Maximum forelegs perpendicular to body. distance from tip of rhinarium to posterior edge Mass: Total mass of individual.