Headbob Display Structure in the Naturalized Anolis Lizards of Bermuda: Sex, Context, and Population Effects

JOSEPH M. MACEDONIA1,2 AND DAVID L. CLARK3

1Department of Biology, University of Memphis, Memphis, Tennessee 38152 USA 3Department of Biology, Alma College, Alma, Michigan 48801 USA

ABSTRACT.—Many members of the Iguania group of lizards engage in stereotyped and species-specific ‘‘pushup’’ or ‘‘headbob’’ displays. Temporal attributes of displays have been quantified for a number of species in the genus Anolis, but few of these studies have examined effects on display structure of signaler sex, display context, and population. With this goal in mind we conducted a comparative study of three Anolis species permanently established (‘‘naturalized’’) on the island of Bermuda: Anolis grahami from Ja- maica, Anolis extremus from Barbados, and Anolis leachi from Antigua and Barbuda. These anoles are dis- tantly related to each other and to Anolis carolinensis—the only Anolis species for which all of the above influences on display structure have been examined in detail. Adults were field-captured and transported to the laboratory where paired interactions were videotaped and headbob displays analyzed. Results revealed one or more variables to exhibit sexual dimorphism in each species, and display context had little influence on signal structure. We then compared results from our founder populations on Bermuda with those published for two of our study species’ source populations. No significant differences in headbob display units were found between A. grahami on Bermuda and on Jamaica. In contrast, male A. extremus on Bermuda produced fewer units per display than did males from Barbados, although the nature of the published data prevented statistical comparison.

To be effective, animal signals must be both mer, 1989; Queral et al., 1995; Orrell and Jens- detectable against a background of irrelevant in- sen, 1998). Often absent from these studies, formation and discriminable from other signals however, are comparisons of display structure with which they might be confused (e.g., Guil- between the sexes (Jenssen et al., 2000), among ford and Dawkins, 1991; Endler, 1992; Fleish- display contexts (DeCourcy and Jenssen, 1994; man, 1992). By reducing within-signal variation Lovern et al., 1999), and among populations of or increasing between-signal variation, or both, the same species (e.g., Jenssen, 1971, 1981; Lov- selection tends to produce signals that are at ern et al., 1999; Macedonia and Clark, 2001). once readily perceived and recognizable. Selec- Moreover, most studies that have examined po- tion for signal stereotypy thus is anticipated to tential sex, context, or population effects on An- drive structural divergence of signal form with- olis display structure have been conducted on in and among species, particularly for signals one species: Anolis carolinensis. Therefore, it is that facilitate mate recognition and reproductive unknown how broadly the ﬁndings of these isolation (e.g., Crews and Williams, 1977; Jens- studies apply to the genus as a whole. sen and Gladson, 1984). In the present study, we examine the potential In the Iguania group of lizards, many taxa effects of display variation attributable to sex of perform stereotyped and often species-speciﬁc the signaler, display context, and population in ‘‘pushup’’ or ‘‘headbob’’ displays (e.g., Carpen- three Anolis species permanently established ter and Ferguson, 1977). These displays occur in (‘‘naturalized’’) on Bermuda: Anolis grahami multiple contexts, such as undirected signaling from Jamaica, Anolis extremus from Barbados, by territorial males, during courtship, in same- and Anolis leachi from Antigua and Barbuda. sex aggressive encounters, and in signaling to These three species are neither closely related to predators (e.g., Carpenter and Ferguson, 1977; one another nor to A. carolinensis (e.g., Jackman Martins, 1991; Leal, 1999). Quantitative descrip- et al., 1999; Creer et al., 2001; Schneider et al., tions of headbob displays have been published 2001). for a number of species in the large genus Anolis The colonization of Bermuda by A. grahami (e.g., Jenssen, 1977a, 1979a,b, 1981, 1983; Scott, has been well documented: 26 males and 45 fe- 1984; Bels, 1986; Fleishman, 1988; Font and Kra- males were deliberately introduced from King- ston, Jamaica in 1905 (Wingate, 1965; Losos, 2 Corresponding Author. Present address: Depart- 1996). In contrast, the introductions of A. extre- ment of Life Sciences, Arizona State University West, mus and A. leachi appear to have been acciden- P.O. Box 37100, Phoenix, Arizona 85069, USA; E-mail: tal, and both species are thought to have arrived [email protected] roughly between 1940 and 1945 (Wingate, 1965). HEADBOB DISPLAYS IN BERMUDIAN ANOLES 267

It is virtually certain that A. extremus and A. lea- vember 1994 and transported to animal housing chi became established on Bermuda from much at Alma College (Alma, MI). Snout–vent lengths smaller founder populations than A. grahami, (mean Ϯ 1 SE, in mm) for subjects included in and these differences seem to be reflected in the present study were A. grahami, males (65.0 their genetic variability. For example, in a study Ϯ 0.6, N ϭ 16), females (46.0 Ϯ 0.9, N ϭ 6); A. of 24 allozymes Gorman et al. (1976) found that extremus, males (65.9 Ϯ 1.9, N ϭ 9), females (52.0 Bermudian A. grahami retained about 82% of the Ϯ 0.2, N ϭ 4); A. leachi, males (93.5 Ϯ 2.1, N ϭ heterozygosity of their source population, 8), females (68.3 Ϯ 1.0, N ϭ 8). whereas in Bermudian A. leachi and A. extremus Lizards were maintained in a room kept on a heterozygosity was approximately 56% and 14:10 L:D light cycle, 60–65% relative humidity, 16% that of their respective source populations. and temperature of 26–29ЊC. Ceiling- and wall- Given that performance of stereotyped headbob mounted full spectrum fluorescent lights pro- displays appears to be under relatively tight ge- vided ambient illumination. Individual lizards netic control (e.g., Cooper, 1971; Stamps, 1978; were housed in plastic cages (inside dimensions: Roggenbuck and Jenssen, 1986), one might an- 28 ϫ 17 ϫ 19 cm). Each cage included a perch ticipate differences in genetic variation between (1.25-cm diameter wooden dowel) and 4-ply pa- the source and founder populations of our study per towel flooring that helped to maintain high species to be borne out in their display struc- relative humidity. Lizards were misted daily ture. and alternately fed crickets and wax worms In this work, we address four questions relat- dusted with phosphorous-free calcium/D3 and ed to headbob displays in our study species. reptile vitamins every three days. First, given the elaboration of male secondary Experimental Protocol.—Subjects were removed sexual traits (e.g., dewlap) and strong sexual from their home cages and placed in wooden size dimorphism (SSD) in these anoles, we ask observation chambers (180 ϫ 39 ϫ 44 cm) fash- whether their headbob displays also exhibit sex- ioned after those illustrated in DeCourcy and ual dimorphism. Specifically, we ask whether Jenssen (1994) for 1–3 days prior to trials. Two the sexes differ in the quantity of units pro- full spectrum fluorescent bulbs and four 100- duced per display by A. grahami and A. extremus watt incandescent bulbs with aluminum reflec- (or the number of displays produced per volley tor shields were suspended above each test tank by A. leachi, see below) and in the durations of to provide light and heat for basking on a 14:10 displays and display units. L:D cycle. A divider bisected each chamber so Second, although both sexes engage in head- that the lizards on either side could not see each bob displays, males (1) use them during pro- other until the divider was removed at trial tracted ritualized contests with other males, (2) commencement. Lizards in the observation produce them far more frequently than do fe- chambers were misted with water twice daily males and, unlike females, (3) broadcast them and fed every other day. During the acclimation in a nondirected fashion in the course of daily period for male-male trials (only), a conspecific activities (e.g., Jenssen et al., 2000). Because female was placed in a small plastic cage near headbob displays appear to be under strong a basking block on each side of the observation sexual selection in males, we ask whether dis- chamber to potentiate male territorial behavior. play temporal structure reflects this selection These caged females were removed from the differential by being more stereotyped in males chamber several minutes prior to raising the than in females. center divider at the initiation of a trial. Third, we wish to know whether display Data Acquisition.—Lizard interactions were re- structure is influenced by display context. Here, corded using two camcorders mounted on tri- we ask whether temporal aspects of headbob pods, with each camera following an assigned displays differ between those performed in ag- subject. A ‘‘screen-splitter’’ combined the two gressive contests as opposed to those performed video images onto one tape with a time stamp. in courtship. All lizard interactions were videotaped from be- Last, given appreciable differences in genetic hind a cloth blind, with small openings cut to heterozygosity between the founder populations accommodate the camcorder lenses. of our study species populations on Bermuda Displays were quantified by superimposing a and their source populations in the Caribbean coordinate system over the video image of a dis- (Gorman et al., 1976, see below), we ask wheth- playing lizard (see Macedonia and Clark, 2001 er differences also occur between source and founder populations in display structure and for details). The changes in coordinates were stereotypy. compiled and display-action-pattern (DAP) graphs were produced (e.g., Carpenter, 1962; MATERIALS AND METHODS Jenssen, 1978). Subjects and Housing.—Adult anoles were col- To be consistent with previously published lected on Bermuda from 1–6 June and 22–27 No- literature, displays were divided into odd- and 268 J. M. MACEDONIA AND D. L. CLARK

FIG. 1. Representative headbob displays from Anolis grahami, Anolis extremus,andAnolis leachi on Bermuda. (A) A. grahami Type A headbob display, (B) A. grahami Type B headbob display, (C) A. extremus headbob display, (D) A. leachi headbob display. Numbers in brackets above the displays indicate ‘‘odd numbered’’ display units; unlabeled ‘‘even numbered’’ units are pauses. In ‘‘d,’’ letters in brackets indicate display type. even-numbered units, where odd-numbered trunk/head motion were produced by our units define periods of motion and even-number study species. Some examples included (1) long units depict pauses. Unit lengths for each dis- series’ of nonstereotyped vertical head/trunk play were calculated as in Macedonia and Clark movements referred to as ‘‘step bobbing’’ (2001). (Rand, 1967; Jenssen, 1979b), (2) rapid, small Up to nine headbob units were measured for amplitude vertical head movements variably A. grahami displays to facilitate comparison with termed ‘‘courtship bobs,’’ ‘‘courtship-nods,’’ previously published work (Jenssen, 1981). Up ‘‘shudders,’’ or ‘‘jiggles’’ (Carpenter and Fer- to seven units were measured for A. extremus guson, 1977; Jenssen, 1977a) and, in A. extremus, displays, as males rarely produced more than high amplitude ‘‘jerk bobs’’ (Stamps and Bar- this. We arbitrarily required a minimum of sev- low, 1973). Unlike signature displays, these oth- en display units for A. grahami (Fig. 1A–B) and er motion patterns lack species specificity. five for A. extremus (Fig. 1C) for a display to be Potential effects of sex and context on tem- included in our quantitative analyses. By com- poral display structure were examined at the parison, A. leachi produced ‘‘volleys’’ (Jenssen, level of individual display units. Each subject 1977a) of displays, in which two display types contributed a single mean value for each display were performed in a variable order (Fig. 1D). unit. We used the Mann-Whitney U-test to test Because displays were brief (roughly the dura- for intraspecific sex differences in displays. The tion of one A. grahami display unit), we did not Wilcoxon Matched-Pairs Signed-Ranks test was partition them into units. used to test whether display unit durations dif- Data Analysis.—We restricted our analyses to fered between same-sex and opposite-sex con- species-specific motion patterns termed ‘‘signa- texts for male subjects that contributed displays ture displays’’ (Jenssen, 1977a), although several to both contexts. We limited these latter analy- other kinds of displays involving conspicuous ses to males because females either did not ex- HEADBOB DISPLAYS IN BERMUDIAN ANOLES 269

TABLE 1. Headbob displays of Anolis: sample sizes by display context.

Number of displays Context Display (Mean Ϯ SE) Total displays (N) A. Anolis grahami adult males (N ϭ 16) and females (N ϭ 6) Male to male Type A 5.6 Ϯ 1.5 (N ϭ 90) Males Male to male Type B 2.0 Ϯ 0.7 (N ϭ 32) 209 Male to female Type A 4.6 Ϯ 1.1 (N ϭ 74) Male to female Type B 0.8 Ϯ 0.3 (N ϭ 13) Female to male Type A 5.0 Ϯ 1.4 (N ϭ 164) Females Female to male Type B 3.0 Ϯ 0.8 (N ϭ 45) 209 B. Anolis extremus adult males (N ϭ 9) and females (N ϭ 4) Male to male 11.2 Ϯ 1.8 (N ϭ 101) Males Male to female 6.4 Ϯ 2.4 (N ϭ 58) 159 Female to female 7.3 Ϯ 5.7 (N ϭ 29) Females Female to male 9.8 Ϯ 5.9 (N ϭ 39) 68 C. Anolis leachi adult males (N ϭ 8) and females (N ϭ 8) Male to male Type A 24.3 Ϯ 10.6 (N ϭ 194) Males Male to male Type B 15.4 Ϯ 7.8 (N ϭ 123) 498 Male to female Type A 14.1 Ϯ 5.3 (N ϭ 113) Male to female Type B 8.5 Ϯ 3.8 (N ϭ 68) Female to male Type A 9.5 Ϯ 6.2 (N ϭ 76) Females Female to male Type B 9.0 Ϯ 3.2 (N ϭ 72) 663 Female to female Type A 41.3 Ϯ 19.1 (N ϭ 330) Female to female Type B 23.1 Ϯ 12.6 (N ϭ 185) perience the same-sex context (A. grahami)or ble 2; Fig. 1A–B). In practice, we used the du- the same females did not produce displays in ration of Unit 2 to classify displays as Type A both contexts (A. extremus). Two few A. leachi or B, as this unit exhibits approximately an or- subjects of either sex produced displays in both der of magnitude difference in duration be- contexts to facilitate statistical tests. We used tween the two display types (Table 2; Jenssen, nested ANOVAs to assess the proportion of var- 1981). The first nine units in Type A displays iation attributable to sex, among-subject and (Fig. 1A) summed to approximately 3.7 sec in within-subject variance for each display unit duration, whereas those in Type B displays (Fig. (e.g., DeCourcy and Jenssen, 1994; Jenssen et al., 1B) summed to about 3.0 sec (Table 2). In con- 2000). For this analysis, all unit duration data trast, A. extremus produced only one type of ste- for each subject were used rather than a mean reotyped headbob display: a series of sinusoidal value for each unit. motions whose first 7 units summed to approx- Statistical analyses were conducted with SPSS imately 0.8 sec (Table 3; Fig. 1C). for Macintosh (vers. 10) except for the nested Headbob displays of A. leachi differed from ANOVAs, which were run in a DOS environ- those of the other two species in several ways. ment with BIOM PC (F. J. Rohlf, BIOM: a pack- First, displays always were performed in volleys age of statistical programs to accompany the (i.e., series) that ranged between about 7 and 12 text BIOMETRY. W. H. Freeman, New York, sec in duration (mean ϭ 8.6 Ϯ 0.34 sec; N ϭ 16 1986). Sample sizes for displays analyzed are subjects). Volleys contained two basic display presented in Table 1. We follow Barlow’s (1968) types, which we termed Type A and Type B, criterion for behavioral stereotypy, where dis- that appeared in variable sequences and num- play units having coefficients of variation (CVs) bers of repetitions (Fig. 1D). Type A displays less than 35% are considered highly stereo- began with a rapid head drop, followed imme- typed. diately with a rebounding head upswing and a second drop. Type B displays begin with a RESULTS sharp head rise, followed by a brief pause, a Headbob displays in A. grahami comprised a sharp head drop, and a small rebounding rise. series of ‘‘plateau-shaped’’ units, and two dis- Nested ANOVAs revealed the proportions of play types were performed, Type A and B (Jens- display variance attributable to signaler sex, sen, 1977b, 1981). These two display types are among-subject variation within a sex class, and readily distinguished by differences in the du- within-subject variation (Fig. 2). In A. grahami ration of Units 2, 6, and 8: In Type B displays Type A displays, a subject’s sex accounted for the duration of these units approaches zero (Ta- smallest amount of the variation (21.7%), fol- 270 J. M. MACEDONIA AND D. L. CLARK

TABLE 2. Between-sex comparisons of headbob displays from 16 adult male and six adult female Anolis grahami on Bermuda. Each subject contributed a mean value per parameter. Unit durations in seconds. Percent duration (% Dur) ϭ a given unit’s mean duration as a proportion of the sum of all unit mean durations for that display type. Test statistic is Mann-Whitney U. * Coefﬁcient of variation (CV) unreliable when mean Ͻ0.1.

Unit statistics Males (N ϭ 16) Females (N ϭ 6) Display Unit Mean SE CV % Dur Mean SE CV % Dur UP Type A U1 1.230 0.132 43.1 32.4 0.899 0.035 9.5 25.0 16.0 .018 U2 0.423 0.062 58.6 11.2 0.295 0.005 3.7 8.2 21.5 .051 U3 0.327 0.016 19.1 8.6 0.253 0.018 17.2 7.0 15.0 .015 U4 0.459 0.035 30.7 12.1 0.495 0.027 13.6 13.7 43.5 .740 U5 0.349 0.014 16.6 9.2 0.285 0.013 11.2 7.9 14.5 .013 U6 0.219 0.018 32.4 5.8 0.290 0.017 14.0 8.0 19.5 .036 U7 0.384 0.023 23.5 10.1 0.411 0.037 22.2 11.4 43.0 .712 U8 0.114 0.025 87.4 3.0 0.175 0.022 30.5 4.9 22.0 .055 U9 0.286 0.023 32.0 7.5 0.501 0.037 18.1 13.9 00.0 .000 U1–9 3.791 0.107 76.2 100.0 3.603 0.072 54.3 100.0 48.0 1.000 Type B U1 1.158 0.066 20.7 36.9 0.905 0.051 13.8 31.2 14.0 .028 U2* 0.048 0.031 — 1.5 0.042 0.020 — 1.4 27.5 .296 U3 0.345 0.021 21.4 11.0 0.327 0.026 19.1 11.3 37.0 .861 U4 0.446 0.067 54.0 14.2 0.495 0.042 20.9 17.1 38.5 .965 U5 0.310 0.019 22.6 9.9 0.293 0.020 17.0 10.1 32.0 .539 U6* 0.098 0.039 — 3.1 0.020 0.009 — 0.7 26.0 .246 U7 0.410 0.028 24.9 13.0 0.426 0.017 10.0 14.7 31.5 .510 U8* 0.012 0.005 — 0.4 0.017 0.014 — 0.6 38.0 .919 U9 0.315 0.021 22.7 10.0 0.375 0.022 13.1 12.9 10.0 .035 U1–9 3.142 0.114 98.1 100.0 2.899 0.095 88.3 100.0 20.0 .096 lowed by among-subject variance (27.7%), and ance in A. leachi displays could be attributed to within-subject variance (51.7%). Type B dis- within-subject variation (Type A displays: plays showed the same hierarchy in levels of 98.9%; Type B displays: 99.3%). variation: sex (6.4%), among-subject (33.8%), Sex Effects on Display Temporal Structure.—In and within-subject (59.8%) variance. Although A. grahami Type A displays five of nine units sex contributed least to the variation in both dis- (77.8%) differed significantly in duration be- play types, that contribution was considerably tween the sexes (with two additional units greater in Type A than in Type B displays. reaching P Ͻ 0.06), whereas only two of nine In A. extremus displays (Fig. 2C), subject sex units (22.2%) differed significantly in Type B accounted for the least amount of variation displays (Table 2). In total display duration (i.e., (7.1%), among-subject variance accounted for an the sum of Units 1–9), neither display type dif- intermediate amount of variation (23.3%), and fered significantly between the sexes (Table 2). within-subject variance accounted for most of However, females produced significantly more the variation (69.6%). Virtually all of the vari- units per Type A display than did males (fe-

TABLE 3. Between-sex comparisons of headbob displays from nine adult male and four adult female Anolis extremus on Bermuda. Legend as in Table 2.

Unit statistics Males (N ϭ 9) Females (N ϭ 4) Unit Mean SE CV % Dur Mean SE CV % Dur UP U1 0.254 0.010 11.8 32.0 0.251 0.013 10.1 31.0 17.5 .938 U2* 0.016 0.004 — 2.0 0.047 0.011 — 5.8 2.5 .016 U3 0.177 0.014 23.0 22.3 0.147 0.011 14.4 18.1 9.0 .165 U4* 0.016 0.003 — 2.0 0.042 0.009 — 5.2 2.0 .013 U5 0.172 0.008 14.4 21.7 0.169 0.007 8.1 20.9 14.0 .537 U6* 0.054 0.018 — 6.8 0.037 0.001 — 4.5 8.5 .712 U7 0.104 0.011 18.1 13.2 0.118 0.011 18.6 14.5 4.0 .480 U1–7 0.793 0.034 80.4 100.0 0.810 0.030 69.3 100.0 6.0 .064 HEADBOB DISPLAYS IN BERMUDIAN ANOLES 271

males on all units for which this statistic could be calculated reliably (15 of 18 units total: Table 2). In A. extremus, only two display units (28.6%) differed significantly in duration between the sexes, although total display duration (sum of Units 1–7) approached significance (Table 3). Like A. grahami, female A. extremus produced significantly more units per display than males (female mean ϭ 8.24 Ϯ 0.73 sec, N ϭ 4; male mean ϭ 5.03 Ϯ 0.23 sec, N ϭ 9; U ϭ 0, P ϭ 0.005). CVs were larger for males than females on three of the four units for which they could be calculated reliably (four of seven units total: Table 3). Mean duration of A. leachi Type A displays was identical for both sexes, but differences in mean duration of Type B displays approached significance (Table 4). The quantity of Type A and B displays produced in volleys did not differ between the sexes (Type A displays: male mean ϭ 6.88 Ϯ 0.49, N ϭ 8; female mean ϭ 6.92 Ϯ 0.55, N ϭ 8, U ϭ 30.5, P ϭ 0.875; Type B displays: male mean ϭ 6.52 Ϯ 0.63, N ϭ 8; female mean ϭ 6.31 Ϯ 0.50, N ϭ 8, U ϭ 31.5, P ϭ 0.958. However, male volleys were significantly longer in duration than were female volleys (male mean ϭ 9.47 Ϯ 0.44 sec, N ϭ 8; female mean ϭ 7.73 Ϯ 0.23 sec, N ϭ 8; U ϭ 4.0, P ϭ 0.003). In Type A displays mean CVs for the two sexes were virtually identical, whereas males exhibited a slightly smaller mean CV for Type B displays (Table 4). FIG. 2. Stacked bar graphs showing the proportion Context Effects on Display Temporal Structure.— of variance in displays attributable in two-level nested In A. grahami Type A displays, unit durations ANOVAs to between-sex (black), among-subject (di- did not differ significantly between male-male agonal hatching), and within-subject (white) variation. and male-female contexts (Table 5). Too few in- (A) Anolis grahami Type A headbob display, (B) A. grahami Type B headbob display, (C) Anolis extremus dividuals produced Type B displays in both headbob display. contexts to test for context effects. The number of units produced per display also did not differ between contexts (male-male: mean number of male mean ϭ 14.27 Ϯ 1.17 sec, N ϭ 6; male units ϭ 9.69 Ϯ 0.33; male-female: mean number mean ϭ 9.65 Ϯ 0.23 sec, N ϭ 16; U ϭ 9, P ϭ of units ϭ 9.77 Ϯ 0.25; Wilcoxon Z ϭϪ0.169, P 0.004). A similar relationship was found for ϭ 0.866, N ϭ 8). In A. extremus, five display Type B displays (female mean ϭ 16.59 Ϯ 1.76 units were produced consistently in male-male sec, N ϭ 6; male mean ϭ 11.13 Ϯ 0.51 sec, N ϭ and male-female contexts, but only one of these 16; U ϭ 4, P ϭ 0.003). Taking both display types differed significantly in duration between con- together, the CV was larger for males than fe- texts (Table 6). The number of display units pro-

TABLE 4. Between-sex comparisons on headbob displays from eight adult male and eight adult female Anolis leachi on Bermuda. Legend as in Table 2.

Display statistics Males (N ϭ 8) Females (N ϭ 8) Display Mean SE CV Mean SE CV UP Type A 0.311 0.013 11.5 0.311 0.013 11.8 30.0 0.834 Type B 0.348 0.012 9.8 0.308 0.014 12.4 14.5 0.066 272 J. M. MACEDONIA AND D. L. CLARK

TABLE 5. Differences in male Anolis grahami Type TABLE 6. Differences in male Anolis extremus head- A headbob display unit durations between male-male bob display unit durations between male-male and and male-female contexts. Eight of 16 subjects con- male-female contexts. Six of nine subjects consistently tributed displays to both contexts. Test statistic is Wil- produced displays of ﬁve units in both contexts. Test coxon Z. statistic is Wilcoxon Z.

Unit ZPUnit ZP U1 Ϫ0.560 .575 U1 Ϫ0.943 .345 U2 Ϫ0.840 .401 U2 Ϫ2.023 .043 U3 Ϫ0.140 .889 U3 Ϫ0.734 .463 U4 Ϫ1.260 .208 U4 Ϫ1.782 .075 U5 Ϫ0.420 .674 U5 Ϫ0.674 .500 U6 Ϫ0.560 .575 U7 Ϫ1.820 .069 U8 Ϫ0.840 .401 U9 Ϫ1.352 .176 sexual dimorphism in display unit durations varied with display type. Few sex differences were found in unit durations of A. extremus displays and in A. grahami Type B displays, but duced did not differ between contexts (male- most units differed in duration between the sex- male: mean number of units ϭ 4.61 Ϯ 0.41; es in A. grahami Type A displays. Sex differenc- male-female: mean number of units ϭ 4.93 Ϯ es in the total duration of displays approached 0.46; Z ϭϪ0.734, P ϭ 0.463, N ϭ 6). significance for A. extremus and for A. leachi Population Effects on Display Temporal Struc- Type B (but not Type A) displays, and male vol- ture.—Neither display type of A. grahami males leys were significantly longer than female vol- differed significantly in duration between the leys in A. leachi. source population on Jamaica (data from Jens- By comparison, Jenssen et al. (2000) measured sen, 1981:table 1) and the founder population on the duration of 24 units across three displays Bermuda (Type A display: Wilcoxon Z ϭ types in A. carolinensis and found only four Ϫ1.364, P ϭ 0.173; Type B display: Z ϭϪ1.007, (about 17%) to differ significantly between the P ϭ 0.314, N ϭ 9 units). To compare other tem- sexes. Nevertheless, one of three display types poral attributes of the displays between the two (Type B) differed significantly in total duration populations we chose four basic parameters: between the sexes (Jenssen et al., 2000). Stamps units having the longest duration, shortest du- (1978), however, found no significant sex differ- ration, largest CV, and smallest CV. None of ences in the duration of headbob display units these parameters differed between the two pop- for the signature display of Anolis aeneus, a spe- ulations in Type A displays (Table 7). In Type cies closely related to A. extremus (Creer et al., B displays the two populations again exhibited 2001). As an example of other Iguania, Martins the same longest- and shortest-duration units, (1991) did not find sex differences in the dura- but they differed in which units exhibited the tion of Sceloporus graciosus pushup displays, al- largest and smallest CVs (Table 7). though males did produce significantly more For A. extremus, the range and mean number headbobs per display than females. In sum, and of display units produced was smaller in the despite among-species variation in which dis- Bermuda population (mean ϭ 5, range ϭ 3–7; play parameters differed significantly between N ϭ 159 displays) than in the source population the sexes, most Anolis species whose displays on Barbados (mean ϭ 7, range ϭ 3–9, N ϭ 32 have been scrutinized for sexual dimorphism displays: data extrapolated from ‘‘peaks per have provided some level of support for it. bobbing sequence’’ in Gorman, 1968). Statistical Do Sex Differences Exist in Anolis Display Ste- comparisons were not possible due to the na- reotypy?—We predicted that stereotypy in head- ture of these published data. No display data on bob unit durations would be greater in males Caribbean A. leachi were available for compari- than in females, given the far greater impor- son to our study population on Bermuda. tance of display to males (e.g., Jenssen et al., 2000). Our findings did not support this predic- DISCUSSION tion. In virtually every case, unit durations were Are Anolis Headbob Displays Sexually Dimor- more variable in male than in female A. grahami phic?—Our results suggest that, although some and A. extremus displays. One possible expla- aspects of display temporal structure are sexu- nation for this result is that increased variation ally dimorphic, others are not. For example, fe- in display unit durations could facilitate indi- male A. grahami and A. extremus both produced vidual recognition among territory holding significantly more units per display than did males. This seems unlikely, however, given that conspecific males. In contrast, the presence of our nested ANOVAs revealed far greater within- HEADBOB DISPLAYS IN BERMUDIAN ANOLES 273

TABLE 7. Comparison of mean unit durations and coefﬁcients of variation (CV) between Anolis grahami males from a founder population on Bermuda and from the source population on Jamaica. Founder population: 164 Type A and 45 Type B displays from 16 males (this study); source population: 84 Type A and 52 Type B displays from 16 males (from Jenssen, 1981). Highest values for each population in bold type with asterisks, and lowest values in italics with asterisks. Dashes indicate coefﬁcient of variation (CV) unreliable when mean Ͻ0.1.

Type A displays Type B displays Mean unit duration (s) Unit CV Mean unit duration (s) Unit CV Unit Founder Source Founder Source Founder Source Founder Source 1 1.230* 0.905* 43% 19% 1.158* 1.000* 21%* 17% 2 0.423 0.585 59% 19% 0.048 0.000* — — 3 0.327 0.384 19% 20% 0.345 0.342 21%* 30%* 4 0.459 0.738 30% 24% 0.446 0.683 54%* 26% 5 0.349 0.298 17%* 16%* 0.310 0.316 23% 16%* 6 0.219 0.381 32% 21% 0.098 0.000* — — 7 0.384 0.433 24% 18% 0.410 0.415 25% 16%* 8 0.114* 0.166* 87%* 44%* 0.012* 0.000* — — 9 0.286 0.369 32% 24% 0.315 0.295 23% 25% Total 3.791 4.259 3.142 3.051

subject variation than between-subject variation context (male-alone and male-male), among- in display unit durations. Moreover, a much subject, and within-subject variation, DeCourcy simpler explanation exists. Our study included and Jenssen (1994) found context to account for more than twice as many male A. grahami and less than 4% of the variance in display structure. A. extremus than females, and the larger coeffi- Similarly, in an examination of male display cients of variation associated with male displays structure across three populations, Lovern et al. may stem from this sample size discrepancy. (1999) included factors for population, context Support for this possibility comes from two (male-alone, male-male, and male-female), sources. among-subject variation, and within-subject First, our analysis of displays of A. leachi in- variation. Context was found to account for an cluded eight adults of each sex, and for this spe- average of about 7% of the variance in display cies, the CVs of females were slightly higher structure. These results are consistent with our than those of males (Table 4). Second, in a study findings that display context had only a subtle of headbob displays of A. carolinensis that in- influence on the structure of displays of Anolis. cluded 22 females and eight males, Jenssen et al. Does Anolis Display Structure Vary at the Popu- (2000) found female CVs to be larger than those lation Level?—We tested whether mean unit du- of males in 20 of 24 units measured across three rations differed between the headbob displays display types. In sum, if the coefficient of vari- of A. grahami source (Kingston, Jamaica) and ation is as tightly linked to sample size as these founder (Bermuda) populations but were unable comparisons suggest, its value as a measure of to detect a significant difference. We also ex- stereotypy clearly diminishes as sample size sex amined four parameters characterizing the ma- bias increases. jor temporal attributes of the displays: the lon- Does Context Influence Anolis Display Temporal gest and shortest units, and the units with the Structure?—We examined the possibility that largest and smallest CVs. Except for the largest context influenced display structure by compar- and smallest CV values in Type B displays, the ing displays produced by the same males in two populations did not differ on the parame- male-male and male-female encounters. Our ters examined. analysis revealed a context effect on display unit Gorman et al. (1976) considered the roughly duration for only one unit of 14 examined (Ta- 18% reduction in heterozygosity in A. grahami bles 5–6), and failed to show any effect of con- on Bermuda (compared to their source popula- text on the mean number of units produced in tion) to be relatively minor, and attributed this displays. fact to the unusually large founder population. By comparison, the potential effect of context Likewise, our findings for A. grahami suggest on display temporal structure was investigated that the temporal patterning and stereotypy in in two laboratory studies of headbob displays headbob displays have changed little in nearly of A. carolinensis using nested ANOVAs. Includ- a century of separation from the source popu- ing in their analysis factors for display type, lation. 274 J. M. MACEDONIA AND D. L. CLARK

This stability in display structure likewise has and Lamont, 1988). By comparison, the same been noted in another study of anole introduc- display units in Cyclura nubila lewisi from Grand tions. Lovern et al. (1999) compared the head- Cayman differed from those of C. n. nubila on bob displays of A. carolinensis introduced to Ha- Cuba by only about 20%. Although the size of waii around 1950 with those of endemic A. car- the founding population of C. n. lewisi on Grand olinensis in Florida and Georgia. Although some Cayman cannot be estimated, the rapid change significant differences in display unit durations in display structure in C. n. nubila on Puerto were found across the three populations, the Rico illustrates the potential of small founding Hawaiian founder population deviated no more population size as a catalyst to display evolu- from the two endemic populations in unit du- tion. rations or CVs than the two endemic popula- As a final example, Macedonia and Clark tions deviated from each other. In fact, the in- (2001) documented differences in headbob dis- troduced Hawaiian population was more simi- plays among three ‘‘color’’ populations (green, lar in display structure to the population from brown, and blue) of the Grand Cayman anole Georgia than the latter was to conspecifics in the (Anolis conspersus), which shares a most recent adjacent state of Florida (Lovern et al., 1999). common ancestor with the ‘‘Kingston’’ popula- In striking contrast, the first four of nine tion of A. grahami (Jackman et al., 2002). The headbob display units from two A. nebulosus distribution of the blue morph is contiguous populations, located approximately 250 km with those of the green and brown morphs, apart in Mexico, exhibited virtually no overlap whereas the latter two populations are allopat- in duration (Jenssen, 1971). Similarly large pop- ric. Each color population differed significantly ulation-level differences have been described for from one or more of the others in display unit other members of the Iguania. For example, Fer- durations, and the brown form was the most guson (1970) and McKinney (1971) documented distinctive. Although significant, the differences considerable population differences in headbob in unit durations among these color populations display structure of Uta stansburiana involving were subtle. In that sense, they were more sim- unit durations and the total number of units ilar in magnitude to intraspecifc sex differences produced per display. Likewise, Martins et al. shown in the present study than to some of the (1998) found significant population differences comparisons described above for conspecific in the number of units produced in displays by populations separated by large geographic dis- Sceloporus graciosus from Utah, Oregon, and tances. southern California. Generally speaking, these Given that Gorman et al. (1976) found genetic studies compared populations separated by heterozygosity in A. extremus on Bermuda to be considerable (but contiguous) geographic dis- only about 16% that of the source population on tances and most likely by substantial expanses Barbados, we anticipated that we would docu- of time. Display structure differences of equiv- ment differences in display structure between alent magnitude, for example, have been shown the two populations of this species. Unfortu- for A. grahami populations separated by millions nately, the published display data available for of years on Jamaica (Jenssen, 1981; Macedonia the Barbados population constrained our com- and Clark, 2001; Jackman et al., 2002). Accord- parison to the mean number and range of units ingly, comparisons of populations separated in produced in displays, which did appear to dif- time by thousands or millions of years might fer between the two populations. Acquiring de- reasonably be expected to exhibit more dramat- tailed display data from A. extremus on Barba- ic differences than those anticipated between dos in the future would permit a closer exami- populations isolated by only 50–100 years (e.g., nation of display structure in two conspecific this study; Lovern et al., 1999). populations that differ dramatically in genetic Nevertheless, results from a comparative variability. study of rock iguanas (Cyclura) suggest that large differences in display structure can arise Acknowledgments.—We thank N. Burnie, in a short time. Martins and Lamont (1988) com- M.R.C.V.S., Government Veterinary Officer, for pared headbob displays from the source popu- his enthusiastic hospitality and facilitation of lation of Cyclura nubila nubila on Cuba with collection permits on Bermuda, as well as G. Ro- those from descendents of several individuals senthal for assistance in anole collection in 1995. that escaped from a zoo on Puerto Rico in the We are grateful to numerous undergraduate stu- early 1960s. The mean duration of display units dents at Alma College and at the University of involving motion (cf. pauses) in the founder Memphis who assisted us in one more of the population was observed to be about 350% following activities: construction of lizard ob- greater than in the source population. Moreover, servation chambers, field collection of lizards, this difference in display temporal structure videotaping Anolis displays, transcribing dis- evolved in only about six generations (Martins play motion coordinates from the videotapes, HEADBOB DISPLAYS IN BERMUDIAN ANOLES 275 and entering those data into spreadsheets. This po of early diversification in Anolis lizards. Sys- manuscript was improved by constructive com- tematic Biology 48:254–285. ments from K. Orrel, G. Smith, and an anony- JACKMAN,T.R.,D.J.IRSCHICK,K.DE QUEIROZ,J.B.LO- mous reviewer. SOS, AND A. LARSON. 2002. Molecular phylogenetic perspective on evolution of lizards of the Anolis grahami series. Journal of Experimental Zoology (Molec- LITERATURE CITED ular and Developmental Evolution) 294:1–16. BARLOW, G. W. 1968. Ethological units of behaviour. JENSSEN, T. A. 1971. Display analysis of Anolis nebu- In D. Ingle (ed.), Central Nervous System and Fish losus (Sauria, Iguanidae). Copeia 1971:197–209. Behaviour, pp. 217–232. Univ. of Chicago Press, . 1977a. Evolution of anoline lizard display be- Chicago. havior. American Zoologist 17:203–215. BELS, V. L. 1986. Analysis of the display-action-pat- . 1977b. Morphological, behavioral and electro- tern of Anolis chlorocyanus (Sauria: Iguanidae). Cop- phoretic evidence of hybridization between the liz- eia 1986:963–970 ards, Anolis grahami and Anolis lineatopus neckeri,on CARPENTER, C. C. 1962. Patterns of behavior in two Jamaica. Copeia 1977:270–276. Oklahoma lizards. American Midland Naturalist . 1978. Display diversity in anoline lizards and 67:132–151. problems of interpretation. In N. Greenberg and P. D. CARPENTER,C.C.,AND G. W. FERGUSON. 1977. Vari- Maclean (eds.), Behavior and Neurology of Lizards: ation and evolution of stereotyped behavior in rep- An Interdisciplinary Conference, pp. 268–285. Nation- tiles. In C. Ganz and D. W. Tinkle (eds.), Biology al Institutes of Mental Health, Rockville, MD. of the Reptilia. Vol. 7. Ecology and Behaviour A, . 1979a. Display behaviour of male Anolis opal- pp. 335–554. Academic Press, New York. inus (Sauria, Iguanidae): a case of weak display ste- COOPER, W. E. 1971. Display behavior of hatchling reotypy. Animal Behaviour 27:173–184. Anolis carolinensis. Herpetologica 27:498–500. . 1979b. Display modifiers of Anolis opalinus CREER,D.A.,K.DE QUEIROZ,T.R.JACKMAN,J.B.LO- (Lacertilia: Iguanidae). Herpetologica 35:21–30. SOS, AND A. LARSON. 2001. Systematics of the An- . 1981. Unusual display behavior by Anolis gra- olis roquet series of the southern Lesser Antilles. hami from western Jamaica. Copeia 1981:728–733. Journal of Herpetology 35:428–441. . 1983. Display behavior of two Hatian lizards, CREWS,D.,AND E. E. WILLIAMS. 1977. Hormones, re- Anolis cybotes and Anolis distichus. In A. G. J. Rhod- productive behavior, and speciation. American Zo- in and K. Miyata (eds.), Advances in Herpetology ologist 17:271–286. and Evolutionary Biology: Essays in Honor of Er- DECOURCY,K.R.,AND T. A. JENSSEN. 1994. Structure nest E. Williams, pp. 552–569. Museum of Com- and use of male territorial headbob signals by the parative Zoology, Cambridge, MA. lizard Anolis carolinensis. Animal Behaviour 47: JENSSEN,T.A.,AND N. L. GLADSON. 1984. A compar- 251–262. ative display analysis of the Anolis brevirostris com- ENDLER, J. A. 1992. Signals, signal conditions, and the plex in Haiti. Journal of Herpetology 18:217–230. direction of evolution. American Naturalist 139: JENSSEN, T. A., K. S. ORRELL, AND M. B. LOVERN. 2000. S125–153. Sexual dimorphisms in aggressive signal structure FERGUSON, G. W. 1970. Variation and evolution of the and use by a polygynous lizard, Anolis carolinensis. pushup displays of the side-blotched lizard genus Copeia 2000:140–149. Uta (Iguanidae). Systematic Zoology 20:79–101. LEAL, M. 1999. Honest signalling during prey-pred- FLEISHMAN, L. J. 1988. Sensory influences on physical ator interactions in the lizard, Anolis cristatellus. design of a visual display. Animal Behaviour 36: Animal Behaviour 58:521–526. 1420–1424. LOSOS, J. B. 1996. Dynamics of range expansion by . 1992. The influence of the sensory system three introduced species of Anolis lizards on Ber- and the environment on motion patterns in the vi- muda. Journal of Herpetology 30:204–210. sual displays of anoline lizards and other verte- LOVERN, M. B., T. A. JENSSEN,K.S.ORRELL, AND T. brates. American Naturalist 139:S36–61. TUCHAK. 1999. Comparisons of temporal display FONT,E.,AND M. KRAMER. 1989. A mulitvariate clus- structure across contexts and populations in male tering approach to display repertoire analysis: Anolis carolinensis: signal stability or lability? Her- headbobbing in Anolis equestris. Amphibia-Reptilia petologica 55:222–234. 10:331–344. MARTINS, E. P. 1991. A field study of individual and GORMAN, G. C. 1968. The relationships of Anolis of sex differences in the push-up display by the sage- the roquet species group (Sauria: Iguanidae)—III. brush lizard, Sceloporus graciosus. Animal Behavior Comparative study of display behavior. Breviora 45:25–36. 284:1–31. MARTINS,E.P.,AND J. LAMONT. 1998. Estimating an- GORMAN,G.C.,Y.J.KIM, AND S. Y. YANG. 1976. The cestral states of a communicative display: a com- genetics of colonization: Loss of variability among parative study of Cyclura rock iguanas. Animal Be- introduced populations of Anolis lizards (Reptilia, haviour 55:1685–1706 Lacertilia, Iguanidae). Journal of Herpetology 12: MARTINS,E.P.,A.BISSELL, AND K. MORGAN. 1998. 47–51. Population differences in a lizard communicative GUILFORD,T.,AND M. S. DAWKINS. 1991. Receiver display: evidence for rapid change in structure and psychology and the evolution of animal signals. function. Animal Behaviour 56:1113–1119. Animal Behaviour 42:1–14. MACEDONIA,J.M.,AND D. L. CLARK. 2001. Headbob JACKMAN,T.R.,A.LARSON,K.DE QUEIROZ, AND J. B. display analysis of the Grand Cayman anole, An- LOSOS. 1999. Phylogenetic relationships and tem- olis conspersus. Journal of Herpetology 35:300–310. 276 J. M. MACEDONIA AND D. L. CLARK

MCKINNEY, C. O. 1971. Individual and intrapopula- ontogeny of display behaviour in Sceloporus undu- tional variation in the push-up display of Uta stans- latus (Sauria: Iguanidae). Ethology 71:153–165. buriana. Copeia 1971:159–160. SCHNEIDER, C. J., J. B. LOSOS, AND K. DE QUEIROZ. ORRELL,K.S.,AND T. A. JENSSEN. 1998. Display be- 2001. Evolutionary relationships of the Anolis bi- havior of Anolis bahorucoensis: an anole with a di- maculatus group from the northern Lesser Antilles. Journal of Herpetology 35:1–12. minutive dewlap. Caribbean Journal of Science 34: SCOTT, M. P. 1984. Agonistic and courtship displays 113–125. of male Anolis sagrei. Breviora 479:1–22. QUERAL, A., R. GARCIA,R.POWELL,J.S.PARMERLEE STAMPS, J. A. 1978. A ﬁeld study of the ontogeny of JR., D. D. SMITH, AND A. LATHROP. 1995. Agonistic social behavior in the lizard Anolis aeneus. Behav- responses by a grass anole, Anolis olssoni,fromthe iour 46:1–31. Dominican Republic, to male conspeciﬁcs. Am- STAMPS,J.A.,AND G. W. BARLOW. 1973. Variation and phibia-Reptilia 16:313–321. stereotypy in the displays of Anolis aeneus (Sauria: RAND, A. S. 1967. Ecology and social organization in Iguanidae). Behaviour 47:67–94. the iguanid lizard Anolis lineatopus. Proceedings of WINGATE, D. B. 1965. Terrestrial fauna of Bermuda. the United States National Museum 122:1–79. Herpetologica 21:202–218. ROGGENBUCK,M.E.,AND T. A. JENSSEN. 1986. The Accepted: 14 October 2002.

Sharing Amazonian Rain-Forest Trees: Ecology of Anolis punctatus and Anolis transversalis (Squamata: Polychrotidae)

LAURIE J. VITT,1,2 TERESA CRISTINA S. AVILA-PIRES,3 MARIA CRISTINA ESPO´ SITO,3 SHAWN S. SARTORIUS,4 AND PETER A. ZANI5,6

1Sam Noble Oklahoma Museum of Natural History and Department of Zoology, University of Oklahoma, Norman, Oklahoma 73072-7029, USA; E-mail: [email protected] 3Departamento de Zoologia, Museu Paraense Emı´lio Goeldi, Ministe´rio da Cieˆncia e Tecnologia, Caixa Postal 399, 66017-970 Bele´m, Para´, Brasil 4Department of Zoology and Sam Noble Oklahoma Museum of Natural History, University of Oklahoma, Norman, Oklahoma 73019-0606, USA 5Department of Biology, University of Oregon, Eugene, Oregon 97403 USA

ABSTRACT.—We studied the ecology of Anolis punctatus and Anolis transversalis at six localities in the Amazon region of Ecuador and Brazil from 1994–1999. Both lizards are arboreal, about the same size (A. punctatus slightly larger) but differ in some body proportions. Anolis transversalis is restricted to undisturbed primary forest more so than A. punctatus, but both use similar microhabitats. During midday, A. transversalis was not observed, suggesting that individuals may not be active near the ground at that time. Both species are thermal conformers although each was able to maintain Tb slightly higher than that of its perch. Most lizards contained prey, and, overall, a positive relationship existed between lizard body size (SVL) and mean prey volume. Volumetrically, A. punctatus ate mostly ants and orthopterans, whereas A. transversalis ate mostly roaches, beetles, and ants. Dietary overlaps were low (0.56). These two species can be considered ‘‘crown giants’’ among Amazonian anoles. Similar to many Amazonian rain-forest lizards, ecological traits of A. punctatus and A. transversalis render them likely candidates for local extinction when deforestation occurs.

Amazonian anoles occur in a variety of mi- coauratus, and possibly A. ortoni). Four of the 10 crohabitats, from leaf litter on the forest floor Amazonian anoles (Anolis philopunctatus, Anolis (Anolis nitens) to the tops of the largest trees phyllorhinus, A. punctatus,andA. transversalis) (Anolis punctatus, Anolis transversalis, Anolis fus- have been grouped in the genus Dactyloa (Guyer and Savage, 1992, 1995). This genus appears to be monophyletic but is not used here because it 2 Corresponding Author. renders remaining anoles paraphyletic (Jackman 6 Present address: Department of Kinesiology and et al., 1999). The status of A. philopunctatus has Applied Physiology, 334 University of Colorado, Boul- been questioned (Avila-Pires, 1995) because the der, Colorado 80309-0354. ECOLOGY OF AMAZONIAN ANOLES 277 only character separating it from A. punctatus is son); (5) approximately 40 km east of Guajara´- presence of an orange dewlap with large black Mirim, Rondoˆnia (hereafter ‘‘Rondoˆnia’’), Brazil spots (Rodrigues, 1988), a character that varies (10Њ19ЈS ϫ 64Њ34ЈW) in tropical lowland forest in A. punctatus. Anolis phyllorhinus is known during January to March 1999 (rainy season); and from only two specimens, each from a different (6) south of the Amazon River (hereafter ‘‘Rio So- locality (Williams, 1965, 1979). Anolis punctatus limo˜es’’) and nearly due south of Manaus, Ama- and A. transversalis are not restricted to crowns zonas (3Њ20ЈS ϫ 59Њ4ЈW) in moderately disturbed of large trees; indeed, both are frequently found rain forest during December to January 1998–1999 on small trees under the canopy (e.g., Hoog- (rainy season). moed, 1973; Duellman, 1978, 1987; Vitt and Two kinds of data were collected: spot obser- Zani, 1996). Nevertheless, these and their close vational samples that consisted of a set of vari- relatives are the most arboreal among Amazo- ables recorded for each individual lizard ob- nian anoles. Relative to other Amazonian ano- served in the field and specimen-based samples. les, they can be considered crown giants. Not From two to four investigators spent nearly ev- only are they considerably larger than other ery day conducting haphazard searches for liz- Amazonian anoles, their morphology is consis- ards and recording observations at each field tently different from other species (e.g., Irschick site. Because we lived in camps or structures et al., 1997). within or at the edge of forest at all sites, no We describe in some detail the ecology of A. temporal bias in our sampling exists; one or punctatus and A. transversalis based on our stud- more of us were in the forest during the entire ies conducted during the wet seasons of 1994– activity cycle of the lizards. 1999. Neither is easy to collect or observe in We recorded the following data for each in- large numbers so we have pooled data from all dividual lizard observed in the field: time of sites to provide a general overview of their ecol- day, habitat type, microhabitat the lizard was ogy. We consider descriptive ecological studies observed in, height and diameter of the perch of lizards important because (1) basic ecological (if on a perch), whether it was sunny or cloudy, data are lacking for most species, (2) ecological lizard exposure to light (direct sun, filtered sun, data provide the basis for understanding the or shade), and behavior (if any). Lizards were role of each species in complex ecosystems, (3) found in a wide variety of habitat types, each of ecological data can be used in phylogenetic which was recorded. To facilitate data summary, analyses that synthesize data from many stud- we condensed our habitat list to include only the ies and trace the evolutionary history of such following broad categories: primary terra firme traits, and (4) ecological data provide an arsenal forest (never flooded), disturbed terra firme for- for defending the value of natural ecosystems est, forest edge, low primary forest (flooded when drastic habitat alteration is considered. during rains), river forest (riparian forest along Rapid deforestation associated with develop- rivers, usually hardwood-palm mix), stream ment has occurred in Amazonia in the absence bed, and treefall. Similarly, lizards occupied a of data on ecology of most of its plant and an- diversity of microhabitats, most of them off of imal species (e.g., Fearnside and Ferraz, 1995). the ground. We condensed our observations into the following broad categories: branch, leaf, leaf MATERIALS AND METHODS litter, log, palm frond, palm leaf, tree trunk, and Lizards were studied at one locality in Ecuador vine. and five localities in Brazil: (1) Amazonian region When lizards were collected while active, we of northeastern Ecuador in Sucumbıós Province recorded body (Tb), substrate (Tss), and air (Ta) (0Њ0Ј latitude ϫ 76Њ10Ј west longitude) near the temperatures with Miller-Weber rapid register- Rio Cuyabeno (hereafter ‘‘Cuyabeno’’) during Feb- ing thermometers. Lizards were typically held ruary to April 1994 (rainy season); (2) terra firme by the head or a front limb, whereas tempera- Amazonian lowland rain forest near the Rio Cu- tures were measured to minimize heat gain rua´-Una at Agropecuaria Treviso LTDA, approxi- from the collector’s hand. All body tempera- mately 101 km south and 18 km east of Santare´m, tures were taken within 30 sec of capture. We Para´, Brazil (hereafter ‘‘Curua´-Una’’; 3Њ9ЈS ϫ also collected both species while they were 54Њ50ЈW) during February to April 1995 (rainy asleep at night. For these, we recorded the exact season); (3) approximately 5 km north of Porto position where they were sleeping and the time Walter, Acre (hereafter ‘‘Rio Jurua´’’), (8Њ15ЈS ϫ that we discovered them. Collected lizards were 72Њ46ЈW) in undisturbed terra firme rainforest of euthanized following approved procedures the Jurua´ River Basin during February to April (Anonymous, 1987) within 2 h of capture and 1996 (rainy season); (4) the Rio Ituxi (hereafter sexed. The following morphological measure- ‘‘Rio Ituxi’’) in the southwestern portion of Ama- ments were taken prior to preservation: snout– zonas (8Њ20ЈS ϫ 65Њ43ЈW) in moderately disturbed vent length, tail length (unregenerated portion), rain forest during January to April 1997 (rainy sea- regenerated tail length, if present (only the re- 278 L. J. VITT ET AL. generated portion) to 0.1 mm with a linear rule, also produces summary datasets that can be total mass to 0.1 g with Acculab௡ field balances, used in standard statistical analyses including a head width, length, and height, body width and matrix containing mean prey length, width, and height, hind-leg length, and foreleg length to 0.1 volume for each lizard as well as total number mm with MaxCal௡ digital calipers. We com- of prey and total stomach contents volume. pared SVL of males and females to determine These data were merged with corresponding whether sexual dimorphism in size existed. We data on lizard morphology for further analyses. also calculated residuals from the regressions of All diet data were log10-transformed prior to sta- morphological variables on SVL (all log10-trans- tistical analyses as were relevant morphological formed) and conducted a Principal Components variables from lizards to normalize distribu- Analysis on the residuals. This adjusts for size tions. We used regression analyses to determine variation among lizards (Miles, 1994). Lizards which morphological variables were most close- were tagged for individual identification, iden- ly associated with dietary variables and regres- tification numbers were recorded, and lizards sion analyses were used to determine the nature were preserved in 10% formalin. Later, all liz- of relationships between mean prey size and lizards were dissected, stomachs were removed ard size. and placed in 70% alcohol for later examination, To compare diets between species, we calcu- and reproductive organs examined to confirm lated a matrix of proportional utilization coef- sex/age class categorization. Sexual maturity of ficients (pi; see Pianka, 1973, 1986) and normal- males was based on the presence of enlarged ized electivities (ei) using the combined diets of testes and convoluted epididymides indicative A. punctatus and A. transversalis from all sites as of breeding activity. In all cases, males consid- a measure of the total spectrum of prey eaten ered mature on the basis of reproductive organs by the two species. We then calculated geomet- also had dewlap coloration typical of sexually ric means of pi and ei (described in detail by mature males. Females that contained oviductal Winemiller and Pianka, 1990). Geometric means eggs or enlarged vitellogenic follicles were con- (gi) balance bias associated with use of either pi sidered sexually mature. or ei in further dietary analyses. We then cal- To determine diets of each species, we opened culated overlaps with the formula stomachs, removed contents, and carefully nn spread them on a Petri dish. Prey removed from gg ͸͸ij ik stomachs were compressed with legs and/or iϭ1 iϭ1 ␾ϭjk wings (when present) pressed against their bod- nn gg22 ies. We counted prey, identified them to the low- ͸͸ij ik est taxonomic category possible (usually fami- Ίiϭ1 iϭ1 ly), and measured their length and width to 0.01 where symbols are the same as above and j and mm with MaxCal௡ digital calipers. In measur- k representing the two species. We use this anal- ing width of prey, we estimated average width ysis only as an assessment of the similarity of to minimize measurement error. These data diets between the two species and make no as- were entered into the diet program BugRun 1.7, sumptions about resource availability. which estimates individual prey volumes based on the formula for a prolate spheroid, Volume RESULTS ϭ (4/3)␲ (1/2 length · 1/2 width2). The pro- Morphological Comparisons.—Anolis punctatus gram calculates proportional use coefficients for of both sexes are green with almost no mark- numerical and volumetric diet data on a lizard ings on the body (Fig. 1A–B). Occasionally, fe- species by species basis by dividing values for males have faint transverse banding consisting each prey category by the totals. It also counts of lighter areas. Male A. transversalis are green the number of lizards containing each prey type with transverse bands comprised of fine dark (frequency), calculates niche breadth based on colored dots (Fig. 1C). Females, however, have numerical and volumetric data and produces a broad transverse bands, usually brown in color summary dietary table. We use the inverse of (Fig. 1D). The species can be easily distin- Simpson’s (1949) diversity measure to estimate guished in the field because A. transversalis has niche breadth: aqua-blue eyes, A. punctatus does not. The largest A. punctatus was a male 90 mm SVL weigh- n ing 10.7 g. The largest A. transversalis wasafe- B ϭ 1 p 2 ͸ i male 87 mm SVL weighing 10.1 g. Most indi- ΋iϭ1 viduals we collected were adults (Fig. 2). where p is the proportional use of prey item i To compare morphological features between and n is the number of prey categories. Values sexes within each species, we restricted our vary from 1 (exclusive use of a single prey type) analyses to lizards Ն 60 mm SVL. Sexual di- to n (even use of all prey types). BugRun 1.7 morphism in size was apparent in A. punctatus ECOLOGY OF AMAZONIAN ANOLES 279

FIG. 1. Photographs of (A) male Anolis punctatus showing elongated and extended snout; (B) female A. punctatus showing lack of extended snout; (C) male Anolis transversalis with large dorsal crest on neck, smaller dorsal crest on back, and transverse markings of tiny dots; and (D) female A. transversalis lacking all dorsal crests and having broad, dark-colored, transverse bands.

with males larger in SVL (Mann-Whitney U- test, Z ϭϪ3.79, P Ͻ 0.0001). Males averaged 78.2 Ϯ 1.3 mm SVL and weighed 7.8 Ϯ 0.4 g (N ϭ 22), whereas females averaged 72.5 Ϯ 0.7 mm SVL and weighed 7.3 Ϯ 0.2 g (N ϭ 40). Sexual dimorphism in SVL was not detectable in A. transversalis (Mann-Whitney U-test, Z ϭϪ0.346, P ϭ 0.7296). Males averaged 78.6 Ϯ 1.0 mm SVL and weighed 7.1 Ϯ 0.3g(N ϭ 19) whereas females averaged 77.8 Ϯ 1.4 mm SVL and weighed 7.5 Ϯ 0.5g(N ϭ 12). A two factor AN- COVA with log SVL as the covariate revealed no differences between species or sexes in log mass (all P-values Ͼ 0.32); at any given SVL, all lizards were similar in mass. Head width, head length, and head height were similar between sexes in A. punctatus (ANCOVA with log SVL as the covariate, all P-values Ͼ 0.07) even though heads differ in ornamentation (Fig. 1A–B). Sim- ilar results were evident in comparisons between male and female A. transversalis (all P- values Ͼ 0.23). Sexual dimorphism in head size independent of body size does not exist in these lizards. Finally, similar results were also ob- FIG. 2. Snout–vent length distributions for males tained in sexual comparisons of size-adjusted and females of Anolis punctatus and Anolis transversalis. hind- and foreleg lengths. Arrows indicate SVL at sexual maturity. Differences were apparent in overall body 280 L. J. VITT ET AL.

TABLE 1. Results of a principal components analysis on size-adjusted morphological variables for two Am- azonian Anolis. Factor scores result from analyses based on residuals from the log-log regressions of each morphological variable on SVL. The oblique solution reference structure is shown.

Variable Factor I Factor II Factor III Factor IV Mass 0.746 0.108 0.034 0.081 Head width 0.269 0.783 Ϫ0.025 Ϫ0.140 Head length 0.662 Ϫ0.110 0.094 Ϫ0.250 Head height Ϫ0.424 0.788 0.040 0.186 Body width 0.817 Ϫ0.130 Ϫ0.178 0.208 Body height Ϫ0.004 0.019 0.000 0.905 Hind-leg length Ϫ0.195 Ϫ0.039 0.835 0.114 Foreleg length 0.065 0.020 0.727 Ϫ0.127 Eigen value 3.007 1.635 1.113 0.838 Percent of variation 37.6 20.4 13.9 10.5 shape between the two species and between the interaction term was signiﬁcant (P-values Ͼ sexes. Slightly more than 80% of size-adjusted 0.32). No effect of species was evident for Factor morphological variation was explained in the III (P ϭ 0.26), but sexes (F1,101 ϭ 18.1, P Ͻ 0.0001)

first four axes of a principal components anal- and the species-sex interaction (F1,101 ϭ 21.6, P ysis (Table 1). Factor I described a gradient Ͻ 0.0001) were significant. based on body width, mass, and head length. Habitat and Microhabitat Use, Activity, and Ther- Factor II described a gradient based on head mal Ecology.—Both species were most frequently width and height. Factor III described a gradient found in forested habitats, but A. transversalis based on hind-leg and foreleg length. Factor IV avoided habitats where direct sun was available, described a gradient based on body height. Al- such as forest edge, streambeds, and treefalls though species and sexes were not easily distin- (Fig. 4). Nevertheless, frequencies of use of dif- guishable on a bivariate plot of factors I and II ferent habitats were not significant (Wilcoxon (Fig. 3), significant differences exist. A two-fac- Signed Rank Test on proportions, Z ϭϪ0.507, tor (species and sex) ANOVA on Factor I indi- P ϭ 0.61). No significant differences existed be- cated a species (F1,101 ϭ 153.8, P Ͻ 0.0001) and sex (F1,101 ϭ 44.5, P Ͻ 0.0001) difference as well as a significant interaction (F1,101 ϭ 9.7, P ϭ 0.0024). A species effect was evident for Factor

II (F1,101 ϭ 18.5, P Ͻ 0.0001) but neither sex nor

FIG. 3. Plot of factor scores for the ﬁrst axes from a Principal Components Analysis on size adjusted morphological variables for Anolis punctatus and Anolis FIG. 4. Habitat and microhabitat use by Anolis transversalis. punctatus and Anolis transversalis. ECOLOGY OF AMAZONIAN ANOLES 281

FIG. 5. Daily activity of Anolis punctatus and Anolis transversalis based on ﬁeld observations. tween species in frequency of use of different microhabitats habitats (Wilcoxon Signed Rank FIG. 6. Relationship between Tb and Tss for Anolis Test on proportions, Z ϭ 0, P Ͼ 0.99), both were punctatus and Anolis transversalis with inserts showing found primarily on tree trunks (Fig. 4). Anolis distributions of Tb for each species. punctatus were found active throughout the day with slightly more observed during midday. Ac- tivity data are limited for A. transversalis; nev- (17.1%) in river forest, and one (2.9%) in a ertheless, these lizards were more often ob- stream bed. Of these, 21 (60%) were asleep on served in morning and late afternoon disap- branches, eight (22.9%) were on leaves, one peared during midday (Fig. 5). (2.9%) was on a palm leaf, four (11.4%) were on Body temperatures of 32 active A. punctatus tree trunks, and one (2.9%) was on a vine. Perch averaged 29.2 Ϯ 0.3ЊC(x¯ Ϯ SE). Corresponding heights for 14 sleeping A. punctatus averaged

Tss and Ta were 28.0 Ϯ 0.3 and 27.9 Ϯ 0.3ЊC, 1.86 Ϯ 0.23 m off the ground. Diameters of six respectively. Body temperatures of 9 active A. perches averaged 0.17 Ϯ 0.02 cm. Three A. punc- transversalis averaged 27.6 Ϯ 0.6ЊC. Correspond- tatus were asleep on leaves averaging 21.6 Ϯ 9.3 ing Tss and Ta were 26.0 Ϯ 0.3 and 26.3 Ϯ 0.4ЊC, cm in length by 9.3 Ϯ 4.3 cm in width. Among respectively. Differences between species in Tb, sleeping A. transversalis, seven (33.3%) were

Tss,andTa were significant (ANOVA, all P-val- found in disturbed terra firme forest, one (4.8%) ues Ͻ 0.004). A highly significant relationship in low primary forest, eight (38.1%) in primary 2 existed between lizard Tb and Tss (R ϭ 0.755, terra firme forest, and five (23.8%) in river for-

F1,37 ϭ 118.4, P Ͻ 0.0001; Fig. 6). An ANCOVA est. Of these, 19 (90.5%) were asleep on branch- on Tb with Tss as the covariate failed to detect a es, one (4.8%) was on a leaf, and one (4.8%) was species effect on Tb (F1,35 ϭ 1.9, P ϭ 0.18). The on a vine. Perch heights for 21 sleeping A. trans- interaction term (species, Tss) was also nonsig- versalis averaged 2.81 Ϯ 0.26 m off ground. Di- niﬁcant (F1,35 ϭ 1.9, P ϭ 0.18). Differences be- ameters of 21 perches averaged 0.23 Ϯ 0.05 cm. tween species in Tb resulted from the effect of In both species, individuals asleep on branches

Tss on Tb. Anolis punctatus had higher Tb (29.8 Ϯ were positioned with the head toward the end 0.3ЊC) while active when sun was available than of the branch. those active under clouds (28.4 Ϯ 0.5ЊC; F1,30 ϭ Diets.—The diet of A. punctatus was dominat- 8.0, P ϭ 0.008). Anolis transversalis active when ed numerically by ants and beetles and volu- sun was available did not have higher Tb than metrically by ants and orthopterans. Ants and those active under clouds (F1,7 ϭ 0.3, P ϭ 0.60). beetles also dominated the diet of A. transver- Among 57 A. punctatus, 39 were first observed salis numerically, whereas roaches, beetles, and in shade, five were observed in filtered sun, and ants dominated the diet volumetrically (Table 2). 13 were observed in sun. Among 11 active A. Dietary overlap between the two species was transversalis, 8 were first observed in shade, only 0.56 indicating that each species uses a three were observed in filtered sun, and zero large portion of resources different from the were observed in sun. Exposure did not affect other. Although dietary niche breadth values

Tb in A. punctatus (F2,29 ϭ 0.9, P ϭ 0.40). Samples based on numerical data were similar (5.00 for were not large enough to make similar compar- A. punctatus, 4.45 for A. transversalis), niche isons for A. transversalis. breadth values based on volumetric data were Sleeping Lizards.—Among sleeping A. puncta- quite different (7.13 for A. punctatus, 4.87 A. tus, seven (20%) were found in disturbed terra transversalis) indicating that A. punctatus is less ﬁrme forest, one (2.9%) in low primary forest, specialized with respect to prey. Differences be- 20 (57.1%) in primary terra ﬁrme forest, six tween species in prey size, numbers of prey, and 282 L. J. VITT ET AL.

TABLE 2. Diet Summary for Anolis punctatus and Anolis transversalis from the Amazon. No. is the number of a particular prey type, Vol. is estimated total volume for that prey type, and Freq. is the number of lizards eating that particular prey type.

Anolis punctatus Anolis transversalis Prey type No. %No. Vol. (mm3) %Vol. Freq. No. %No. Vol. (mm3) %Vol. Freq. Springtails — — — — — 2 1.69 0.14 0 2 Orthopterans 20 6.60 1779.97 15.77 19 4 3.39 434.47 7.69 4 Blattarians 2 0.66 454.95 4.03 2 5 4.24 2282.35 40.41 4 Mantids 4 1.32 73.03 0.65 4 2 1.69 270.03 4.78 2 Homopterans 8 2.64 381.66 3.38 8 1 0.85 155.63 2.76 1 Hemipterans 11 3.63 1024.38 9.08 9 4 3.39 430.98 7.63 3 Coleopterans 39 12.87 921.92 8.17 22 16 13.56 610.20 10.80 11 Lepidopterans 1 0.33 49.06 0.43 1 — — — — — Psocopterans 2 0.66 1.89 0.02 1 — — — — — Dermapterans 2 0.66 8.37 0.07 1 — — — — — Ants 120 39.60 3186.21 28.23 23 51 43.22 552.18 9.78 12 Hymenopterans 20 6.60 967.73 8.58 17 3 2.54 57.55 1.02 3 Dipterans 2 0.66 31.03 0.27 2 4 3.39 34.19 0.61 3 Insect larvae 20 6.60 626.89 5.56 16 4 3.39 266.80 4.72 4 Spiders 22 7.26 634.67 5.62 19 8 6.78 269.50 4.77 6 Mites 1 0.33 0.01 0 1 9 7.63 0.22 0 2 Millipedes 23 7.59 686.59 6.08 11 4 3.39 133.98 2.37 4 Mollusks 4 1.32 234.13 2.07 4 — — — — — Frogs 1 0.33 48.92 0.43 1 — — — — — Lizards 1 0.33 173.22 1.54 1 1 0.85 150.20 2.66 1 SUMS 303 100 11284.63 100 — 118 100 5648.44 100 — Niche breadths 5.0 7.13 4.45 4.87 numbers of prey types were minimal (Table 3) ard SVL and total prey volume for both species but the differences in prey size and numbers combined (Fig. 7), and an ANCOVA with log disappeared when the effect of lizard body size SVL as the covariate detected no difference in was considered. An ANCOVA on log10 mean stomach content volume between species (F1,89 prey volume per lizard with log10 SVL as the ϭ 2.7, P ϭ 0.10) or sexes (F1,89 ϭ 0.003, P ϭ 0.96). covariate revealed no effect of species (F1,89 ϭ Lizards had equally full stomachs regardless of

1.6, P ϭ 0.21) or sex (F1,89 ϭ 0.03, P ϭ 0.87) on species or sex. Nevertheless, few lizards had full prey size. Likewise, no effect of species (F1,89 ϭ stomachs and likewise few had empty stomachs

0.2, P ϭ 0.64) or sex (F1,89 ϭ 0.02, P ϭ 0.88) was (Fig. 7). evident on number of prey eaten when the effect of lizard body size was taken out. Similar re- DISCUSSION sults were obtained using various measures of Anolis punctatus and A. transversalis are strictly head size as the covariate. Species and sexes ate arboreal anoles that live in Amazonian trees of the same numbers and sizes of prey at any giv- various sizes. Although A. punctatus is sometimes en body size. found at forest edges or edges of clearings where A signiﬁcant relationship existed between liz- sun is available, these habitats are rarely used by

TABLE 3. Summary statistics on individual prey items for the Amazonian anoles Anolis punctatus and Anolis transversalis. Prey size statistics indicate means Ϯ 1 SE (min–max). Means of log10-transformed prey length, width, and volume were signiﬁcantly different (ANOVA, all P-values Ͻ0.0001). Means of number of prey/lizard were also signiﬁcantly different (ANOVA on log10-transformed data, F1,95 ϭ 7.6, P ϭ 0.0069).

Variable A. punctatus A. transversalis Number of lizards 63 34 Number of prey types 19 15 Number of prey 303 118 Number of prey/lizard 4.8 Ϯ 0.59 3.5 Ϯ 0.97 Prey length (mm) 8.46 Ϯ 0.25 (0.38–25.06) 6.94 Ϯ 0.51 (0.22–23.84) Prey width (mm) 2.15 Ϯ 0.07 (0.24–6.76) 1.97 Ϯ 0.16 (0.17–8.02) Prey volume (mm3) 37.24 Ϯ 3.69 (0.01–346.50) 47.87 Ϯ 10.5 (0–728.12) ECOLOGY OF AMAZONIAN ANOLES 283

sic differences in how the lizards respond to temperature in their natural environment. Anolis transversalis selects cooler microhabitats than does A. punctatus. These differences mirror observations on thermoregulatory behavior in the field. Anolis transversalis usually is observed in shaded forest, well away from treefalls and forest edges, whereas A. punctatus frequently is found near microhabitats receiving sunlight. Anolis transversalis often is found in shade, and when approached, typically moves to the opposite side of branches or trunks. When on branches, individuals adhere closely to the branch rendering them extremely cryptic. Anolis punctatus, however, tends to climb higher up trunks when disturbed (e.g., Vanzolini, 1972) but remains on the same side of the trunk as the FIG. 7. Relationship between lizard body size observer. When on smaller branches or shrubs, (SVL) and total stomach volume indicating that most they do not adhere closely to the branch, and lizards had food in their stomachs when captured. they usually do not move to the other side of the branch to escape detection. Taken together, A. transversalis. Tree trunks were the most com- these data and observations suggest that A. mon microhabitats in which we observed both punctatus and A. transversalis use different parts species while active. However, these are among of the arboreal habitat. the easiest microhabitats on which to observe liz- Similarities in prey size between A. punctatus ards. Many individuals in other microhabitats and A. transversalis reflect similarities in overall (e.g., on leaves) may not have been detected. We lizard body size and its effect on prey size. Dif- also were limited in our ability to observe both ferences in prey types may reflect differences in species because we worked from the ground. Liz- parts of the arboreal habitat used by the two ards high in trees probably were undetected by species, but more data will be necessary to con- us and may have been common in the canopy. firm this. Our observation that A. punctatus and Anolis transversalis has been observed in the can- A. transversalis had substantial numbers of prey opy of Amazonian rain forest (Dixon and Soini, in their stomachs indicates that these lizards, 1975, 1986), and we have observed several A. like many others, are able to maintain positive punctatus climb tree trunks (primarily palms) and energy balance while active (see Huey et al., disappear into the canopy. 2001). Apparent differences in activity periods are Anolis punctatus and A. transversalis could be difficult to assess. We suspect that both species considered the ‘‘crown giants’’ among Amazo- are active throughout the day even though we nian anoles. Crown giants, described by Wil- did not see A. transversalis at midday. Anolis liams (1972), are anoles with body sizes sub- transversalis likely ascends trees and disappears stantially larger than those of congeners occur- into the canopy. In addition, it is an extremely ring on tree trunks and on the ground. Like oth- shy and cryptic lizard. Ground-level sampling er Anolis ecomorphs, crown giants have arisen of canopy species is no doubt biased. Canopy independently several times in other anole fau- sampling could provide additional insight into nas (e.g., Losos, 1992, 1994). Although Anolis activity patterns of both species (e.g., Reagan, ecomorphs clearly exist in Amazonian rain for- 1966). Behavioral differences between the two est, the crown giant ‘‘ecomorph’’ in the Amazon species may also contribute to our inability to is not morphologically most similar to crown gi- detect A. transversalis throughout the day. Body ants in Caribbean island anole faunas (Irschick tempeatures in both species varies with Tss, but et al., 1997). neither is a strict thermal conformer. Body tem- Finally, similar to studies of other Amazonian peratures in both species are almost always rain forest lizards (e.g., Sartorius et al., 1999; Vitt slightly higher than ambient temperatures (see et al., 1998), this study suggests that ecological also Rand and Humphrey, 1968). Slight differ- attributes of A. punctatus and A. transversalis are ences in active Tb between species reflect differ- such that deforestation, and in some cases, min- ences in behavior with respect to the sun. Body imal disturbance to the forest may adversely af- temperatures of A. transversalis were lower than fect these lizards. Relatively low Tb, preference those of A. punctatus, but the differences in Tb for forested habitats (especially A. transversalis), resulted from the effect of differences in ambi- and use of arboreal perches restrict them to for- ent temperatures on lizard Tb rather than intrin- est. Forest ‘‘development’’ or modification, par- 284 L. J. VITT ET AL. ticularly if it included cutting of large trees FEARNSIDE,P.M.AND J. FERRAZ. 1995. A conservation would likely cause local extinction. Not only do gap analysis of Brazil’s Amazonian Vegetation. large trees provide the arboreal habitat used by Conservation Biology 9:1134–1147. these lizards, but they also buffer the thermal GUYER,C.,AND J. M. SAVAGE. 1986. Cladistic rela- impact of the tropical sun, creating an environ- tionships among anoles (Sauria: Iguanidae). Sys- ment conducive to reptiles that cannot tolerate tematic Zoology 35:509–531. . 1992. Anole systematics revisited. Systematic thermal extreme. Biology 41:89–110. Acknowledgments.—A research conventio be- HUEY, R. B., E. R. PIANKA, AND L. J. VITT. 2001. How tween the Sam Noble Oklahoma Museum of often do lizards ‘‘run on empty’’? Ecology 82:1–7. Natural History and the Museu Paraense E. HOOGMOED, M. S. 1973. Notes on the herpetofauna Goeldi made the Brazilian portion of research of Surinam. IV. The lizards and amphisbaenians of possible. We thank W. E. Magnusson, A. Lima, Surinam. Biogeographica 4:1–419. M. C. Arau´ jo, J. P. Caldwell, V. Oliveira, M. IRSCHICK,D.J.,L.J.VITT,P.A.ZANI, AND J. B. LOSOS. Scheffer, and A. Leite for logistic help or for 1997. A comparison of evolutionary radiations in mainland and Caribbean Anolis lizards. Ecology helping to find lizards. SOS Amazoˆnia and the 78:2191–2203. Acre Union of Seringueiros in Rio Branco, the JACKMAN,T.R.,A.LARSON,K.D.QUEIROZ, AND J. B. CEMEX Timber Company in Para´, the Instituto LOSOS. 1999. Phylogenetic relationships and tem- Nacional de Pesquisas da Amazonica (INPA), po of early diversification in Anolis lizards. Sys- Conselho Nacional de Desenvolvimento Cientı´- tematic Biology. 48:254–285. fico e Tecnolo´gico (CNPq, Portaria MCT 170, de LOSOS, J. B. 1992. The evolution of convergent struc- 28/09/94), the Instituto Brasileiro do Meio Am- ture in Caribbean Anolis communities. Systematic biente e dos Recursos Naturais Renova´veis Biology 41:403–420. (IBAMA, permit 073/94-DIFAS), and the Museu . 1994. Historical contingency and lizard com- Paraense E. Goeldi in Bele´m contributed to lo- munity ecology. In L. J. Vitt and E. R. Pianka gistics or research and collecting permits. L. (eds.), Lizard Ecology: Historical and Experimen- Coloma coordinated our research in Ecuador tal Perspectives, pp. 319–333. Princeton Univ. and both the QCAZ in Quito and the Estacioń Press, Princeton, NJ. Biolo´gica de Cuyabeno contributed logistic sup- MILES, D. B. 1994. Covariation between morphology port. Permits were issued by Ministry of Agri- and locomotory performance in Sceloporine lizards. In L. J. Vitt and E. R. Pianka (eds.), Lizard culture and Livestock of the Republic of Ecua- Ecology: Historical and Experimental Perspectives, dor. All animals were treated in accordance pp. 207–235. Princeton Univ. Press, Princeton, NJ. with federal, state, and university regulations PIANKA, E. R. 1973. The structure of lizard commu- (Animal Care Assurance 73-R-100, approved 8 nities. Annual Reviews of Ecology and Systematics November 1994). We thank American and Varig 4:53–74. Airlines for allowing excess baggage. National . 1986. Ecology and Natural History of Desert Science Foundation grants DEB-9200779 and Lizards. Princeton Univ. Press, Princeton, NJ. DEB-9505518 to LJV and J. P. Caldwell support- RAND,A.S.,AND S. S. HUMPHREY. 1968. Interspecific ed part of the work. competition in the tropical rain forest: ecological distribution among lizards at Bele´m, Para´. Pro- LITERATURE CITED ceedings of the U.S. National Museum 125:1–17. ANONYMOUS. 1987. Guidelines for the use of live am- REAGAN, D. P. 1996. Anoline lizards. In D. P. Reagan phibians and reptiles in field research. Publ. ASIH, and R. B. Waide (eds.), The Food Web of a Tropical HL, and SSAR:1–14. Rain Forest, pp. 321–346 Univ. of Chicago Press, AVILA-PIRES, T. C. S. 1995. Lizards of Brazilian Ama- Chicago. zonia (Reptilia: Squamata). Zoologische Verhan- RODRIGUES, M. T. 1988. A new anole of the punctatus delingen Leiden 299:1–706. group from central Amazonia (Sauria, Iguanidae). DIXON,J.R.,AND P. S OINI. 1975. The reptiles of the Papeís Avulsos Zoologia (Saõ Paulo) 36:333–336. upper Amazon basin, Iquitos region, Peru. Part I. SARTORIUS, S. S., L. J. VITT, AND G. R. COLLI. 1999. Lizards and Amphisbaenians. Milwaukee Public Use of naturally and anthropogenically disturbed Museum Contributions in Biology and Geology 4: habitats in Amazonian rainforest by the teiid liz- 1–58. ard Ameiva ameiva. Biological Conservation 90:91– . 1986. The reptiles of the upper Amazon basin, Iquitos region, Peru. Milwaukee Public Muse- 101. um Contributions in Biology and Geology, Mil- SIMPSON, E. H. 1949. Measurement of diversity. Na- waukee, WI. ture 163:688. DUELLMAN, W. E. 1978. The biology of an equatorial VANZOLINI, P. E. 1972. Miscellaneous notes on the herpetofauna in Amazonian Ecuador. Miscella- ecology of some brasilian lizards (Sauria). Papeís neous Publications of the Museum of Natural His- Avulsos Zoologia (Saõ Paulo) 26:83–115. tory, Univ. of Kansas 65:1–352. VITT,L.J.,AND P. A. ZANI. 1996. Organization of a . 1987. Lizards in an Amazonian rain forest taxonomically diverse lizard assemblage in Ama- community: resource utilization and abundance. zonian Ecuador. Canadian Journal of Zoology 74: National Geographic Society Research 3:489–500. 1313–1335. ECOLOGY OF AMAZONIAN ANOLES 285

VITT, L. J., T. C. S. AVILA-PIRES,J.P.CALDWELL, AND congeners in a complex island fauna: a trial anal- V. R. L. OLIVEIRA. 1998. The impact of individual ysis. Evolutionary Biology 6:47–89. tree harvesting on thermal environments of lizards . 1979. South American anoles, the species in Amazonian rain forest. Conservation Biology groups. 2. the proboscis anoles (Anolis laveis 12:654–664. group). Breviora 449:1–19. INEMILLER AND IANKA WILLIAMS, E. E. 1965. South American Anolis (Sauria, W ,K.O., E. R. P . 1990. Organi- Iguanidae): two new species of the punctatus group. zation in natural assemblages of desert lizards and tropical ﬁshes. Ecological Monographs 60:27–55. Breviora 233:1–15. . 1972. The origin of faunas. Evolution of lizard Accepted: 22 November 2002.

Intraspeciﬁc Variation in the Advertisement Call of the Sunset Frog Spicospina ﬂammocaerulea (Anura: Myobatrachidae): A Frog with a Limited Geographic Distribution

M. J. SMITH,1,2 J. D. ROBERTS,3 T. J. H AMMOND,3 AND R. A. DAVIS3

1Biological, Earth and Environmental Sciences, University of New South Wales, Sydney, New South Wales, 2052, Australia 3Department of Zoology, University of Western Australia, 35 Stirling Highway, Crawley, Western Australia 6009, Australia

ABSTRACT.—Intraspecific variation in acoustic signals may reflect local variation in the intensity of natural and sexual selection and random drift. We examined intraspecific variation in the advertisement call of Spicospina flammocaerulea, a southwestern Australian frog species with a limited distribution, fragmented range, small population sizes, and specific breeding habitat requirements. Of the six populations examined, one in particular differed significantly in dominant frequency and body size. There was no relationship between among-population geographic distance and among-population divergence in call structure suggesting that the divergence is not caused by random drift. A correlation analysis detected a positive relationship between the size of males and females found in amplexus. However, there was no evidence of mated males differing in size from unmated males indicating that the differences in dominant frequency are unrelated sexual selection. Call structure variation may reflect differences in recruitment and resultant age structure of local populations.

Examination of intraspecific variation in ani- compatibility of the male (Littlejohn et al., 1960), mal communication systems has not only pro- and can also vary with size and condition (Ryan vided valuable insights into the processes be- and Wilczynski, 1991). Accordingly, variation in hind speciation and the potential for future spe- call structure within and among populations ciation (Ryan and Wilczynski, 1991; Ryan et al., may reflect a range of natural and sexual selec- 1996) but has also led to the discovery of new tion pressures or random genetic drift. species (e.g., Lee and Main, 1954; Littlejohn, Roberts et al. (1997) described a new species 1957; Watson et al., 1971). Amphibian commu- of myobatrachid frog from southwestern Aus- nication systems provide excellent models for tralia, Spicospina flammocaerulea. They described such studies because the male advertisement the advertisement call of S. flammocaerulea from call is used to attract mates in most species and recordings of seven individuals from the type to resolve intrasexual conflicts in many species locality, Mountain Road (Roberts et al., 1997). (Searcy and Andersson, 1986). The call can in- Although no frogs have been heard calling at dicate the genetic quality (Welch et al., 2001) or this site since 1998 (J. D. Roberts, unpubl. data), the number of known populations has increased 2 Corresponding Author. Present address: 215 Tuck- significantly during the same period (Burbidge er Hall, Division of Biological Sciences, University of and Roberts, 2002). Spicospina flammocaerulea has Missouri, Columbia, Missouri 65211, USA. E-mail: a highly fragmented and restricted geographic [email protected] range of about 300 km2 (Burbidge and Roberts, 286 M. J. SMITH ET AL.

2002) and appears to occur primarily in peat swamps at the upper reaches of drainage systems (Roberts et al., 1997). Spicospina flammocaerulea breeds from late spring to early summer (Roberts et al., 1999) when swamps are drying and consequently there is little chance of downstream dispersal of tadpoles. Nothing is known about nonlarval migration in this species. The limited information available on the life history of S. flammocaerulea suggests that population sizes are small, distri- FIG. 1. The advertisement call of Spicospina flammocaerulea. butions are fragmented, and this species re- quires very specific habitats. These attributes have been associated with considerable genetic site consisted of a series of ponds with indeter- subdivision among populations in other am- minate drainage in a peat swamp. phibian species (e.g., Driscoll, 1998; Shaffer et The advertisement call of S. flammocaerulea al., 2000). usually consists of two pulsed notes (Fig. 1), Given the importance of the advertisement however, individuals occasionally omit the first call in anuran reproductive biology, and the note from the call (Roberts et al., 1997). Males possibility that geographic divergence in adver- call in bouts (Roberts et al., 1997) with peak call tisement call structure can be associated with rates of about 1 to 2 calls secϪ1 (Smith unpubl. genetic subdivision (Ryan and Wilczynski, 1991; data). With the exception of Mountain Road, Ryan et al., 1996; Castellano et al., 1998), we ex- frogs were recorded from 8 to 14 November amined variation in the advertisement call of S. 2000. Frogs at Mountain Road were recorded 1 flammocaerulea among six populations. Other November 1994 (Roberts et al., 1997). Record- studies have found considerable intraspecific ings were made on a Sony Professional Walk- variation in the advertisement call of frog spe- mans (WM-D6C) or a Sony TCD5-Pro II cassette cies that inhabit broad geographic areas with a recorder with Beyer Dynamic M88N (C) micro- range of environmental and climatic conditions phones. The snout–vent lengths of all frogs were (Nevo and Capranica, 1985; Ryan and Wilczyn- measured with dial calipers (Ϯ 1mm) and the ski, 1991; Ryan et al., 1996; Hasegawa et al., water temperature at the call site of each frog 1999). Here, we examined geographic variation was measured directly after his call was record- in a species that has a limited distribution and ed (Ϯ 1ЊC). All frogs were recorded while call- in which the populations probably experience ing in water. similar climatic and environmental conditions. Calls were digitized at a sampling rate of 44.1 We also looked for a relationship between geo- kHz and analyzed automatically with an algo- graphic distance and the level of among-popu- rithm using Signal software (version 3.12; En- lation divergence in advertisement call structure gineering Design). Frequency spectra were cal- based upon the premise that local differentia- culated using a Hamming window with a fre- tion will be greater with increased isolation and quency resolution of 1.3 Hz and FFT size of decreased migration (Falconer and Mackay, 32768 points. Fifteen randomly chosen calls also 1996). Finally, we examined the relationship be- were analyzed manually using Cool Edit Pro tween mating success and male size, as females (vers. 1.2) to assess the accuracy of the algo- in other anuran species have been shown to pre- rithm. All measured values from the two anal- fer larger males (Sullivan et al., 1995) or males ysis systems were within 3% of each other. Calls of a particular size (Robertson, 1990). Addition- reported by Roberts et al. (1997) were reana- ally, male size may determine the outcomes of lyzed using the automated system, with calls male-male competition (Halliday and Tejedo, from one frog removed because of background 1995). If male size influences reproductive suc- interference. cess in this species, either through male-male The number of calls recorded from each male competition or female mate choice, then geo- ranged from 4 to 10 (x¯ ϭ 9.5 Ϯ 1.3 SD) and graphic variation in size-related call properties average values for each male were used in all may be caused by geographic variation in sex- analyses. Nine call properties were measured: ual selection. note duration (ms), pulse number, dominant frequency (kHz) and pulse rate (pulses secϪ1) MATERIALS AND METHODS from both notes and internote duration (ms). The six study sites, Trent Road 1, Trent Road Each call property was regressed on call site 2, Middle Road 3, Boronia Road 2, Bow River, temperature using a pooled regression combin- and Mountain Road were located in southwest- ing data from all frogs at all sites. Residuals ern Australia, northeast of Walpole. Each study from these regressions were used in all statis- CALL VARIATION IN THE SUNSET FROG 287

TABLE 1. Principal component analysis of the advertisement call properties of male Spicospina ﬂammocaerulea. The correlation between the component and the original variable are presented. Values equal to and greater than 0.70 are underlined.

Component Variable 1234 First note duration 0.91 0.12 0.19 0.12 First note pulse number 0.88 0.10 0.16 Ϫ0.35 First note dom. freq. 0.07 0.99 Ϫ0.01 Ϫ0.04 First note pulse rate 0.70 0.08 0.07 Ϫ0.52 Internote duration Ϫ0.86 Ϫ0.04 0.17 0.14 Second note duration 0.05 Ϫ0.02 0.99 0.06 Second note pulse number 0.14 Ϫ0.03 0.77 Ϫ0.60 Second note dom. freq. 0.11 0.99 Ϫ0.03 0.01 Second note pulse rate 0.15 0.03 0.03 Ϫ0.96

tical analyses to correct for temperature effects males were located in the chorus at the same on call properties. time (N ϭ 68). Advertisement Call Variation.—We used a ca- Significance levels of 0.05 were used in all nonical discriminant function analysis (DFA; analyses. Statistica version 5.0 was used for all Diekhoff, 1992) to determine whether the over- analyses except the Mantel tests which were all structure of the advertisement call varied performed with TFPGA version 1.3. Means are among populations. Wilk’s lambda was used to presented with Ϯ 1SD. determine whether differences among populations were statistically significant. The discrim- RESULTS ination power of the analyses was assessed by Advertisement Call and Body Size.—Univariate reclassifying individuals with the canonical analyses demonstrated that there was consid- discriminant functions (Diekhoff, 1992). To erable variation in call structure among popu- avoid problems of correlation between vari- lations (Table 2). Multiple comparisons using ables (Table 1), we used a principal component the Tukey-Kramer procedure (P Ͻ 0.05) re- analysis (PCA) to break the variables into axes vealed that Middle Rd 3 had significantly lower that were uncorrelated and independent in the dominant frequencies in the first and second DFA. notes than all sites but Bow River. The males at Mantel tests were used to evaluate the extent Bow River had significantly lower dominant fre- of covariation between the geographic distance quencies in the first and second notes than the among populations (in kilometers) and the ex- males at Boronia Rd 2. tent of among-population variation in the A significant amount of the variation in dom- 2 structure of the advertisement call. A dissimi- inant frequency of the first (r ϭ 0.23; F1,41 ϭ larity matrix was constructed from each of the 11.88; ␤ϭϪ0.06; P Ͻ 0.01) and second (r2 ϭ four principal components using Euclidean 0.22; F1,41 ϭ 11.41; ␤ϭϪ0.06; P Ͻ 0.01) note distances between average values for each pop- was explained by SVL. The SVLs of recorded ulation. Euclidean distances were also used to males differed significantly among populations construct the dissimilarity matrix for geo- (ANOVA: F5,37 ϭ 11.10; P Ͻ 0.01; Fig. 2). Multi- graphic distance. ple comparisons using the Tukey-Kramer pro- Body Sizes and Mating Success.—The snout– cedure (P Ͻ 0.05) revealed that, with the excep- vent length (SVL) of males and females from tion of Bow River, males at Middle Rd 3 were 27 pairs collected from Trent Rd 2 were mea- significantly larger than males at all other pop- sured. The relationship between male and fe- ulations. Males at Mountain Rd were signifi- male size was assessed with a correlation anal- cantly larger than those at Trent Rd 2. ysis. To further investigate the relationship be- Advertisement Call Cariation.—A PCA of adver- tween body size and male mating success at tisement call properties resulted in four factors Trent Rd 2, we determined whether the body with eigenvalues larger than 1 (Table 1), which size of mated males differed from unmated accounted for more than 92% of the total vari- males using an analysis of variance (ANOVA). ation in advertisement call structure. All vari- Mated males were those found in amplexing ables had a factor loading greater than 70%. The pairs (N ϭ 30, three additional males were components were weighted by component 1: found in amplexus, but the female was lost be- first note temporal structure (first note duration, fore she could be measured) and unmated first note pulse number, first note pulse rate, in- 288 M. J. SMITH ET AL. 0.01* 0.17 0.08 0.52 0.36 0.01* 0.63 0.08 0.08 -value Ͻ Ͻ P 5,50 2.10 2.14 9.90 1.61 2.14 0.86 1.25 0.70 F 10.16 7.29 4.18 6.84 1.69 0.16 21.95 1.31 0.16 13.41 6) Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ ϭ FIG. 2. Distribution of the snout–vent lengths of N ( male Spicospina flammocaerulea at six study sites. 1 ϭ 8.29 1.99 1.98 SD. Results of univariate among-population 45.25 32.06 Mountain Rd 18.38 36.82 Trent Rd 1, 2 ϭ Middle Rd 3, 3 ϭ Trent Rd 2, 4 ϭ 123.41 166.45 Ϯ Boronia Rd 2, 5 ϭ Bow River, 6 ϭ Mountain Rd. Solid square ϭ mean, Box ϭϮ1 SD, whiskers ϭ range. 0.66 2.84 0.21 28.70 3.42 3.37 1.09 0.22 18.33 10) Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ

ϭ ternote duration); component 2: dominant fre- N

( quencies (ﬁrst note dominant frequency, second Bow River 1.78 7.33 1.78 25.35 35.36 41.17 13.15 note dominant frequency); component 3: second 102.84 155.91

number of individuals. note length (second note duration, second note ϭ pulse number); component 4: second note pulse N rate (second note pulse rate). 3.43 0.96 0.14 26.26 4.38 7.26 1.41 0.14 14.92 9) A canonical discriminant analysis using the Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ ϭ PCA component score generated signiﬁcant var- N (

2.05 7.89 2.05 iation among populations (Wilk’s lambda ϭ at each study site. Values are mean Boronia Rd 2 20.65 29.71 42.23 26.38

108.57 163.47 0.38; approximate F20,156 ϭ 2.69; P Ͻ 0.01; Table 3). Middle Road 3 was well separated from the other populations, with lower scores along ca-

7.41 1.11 0.20 35.02 9.19 4.85 0.54 0.19 11.35 nonical discriminant axis 1 (Fig. 3). Most of the 10)

Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ variation in canonical discriminant axis 1 was ϭ

N explained by component 2 (Table 1), which was ( 1.55 7.71 1.52 Middle Rd 3 interpreted as a frequency and body size com- 18.92 35.74 43.24 27.35 81.56 158.39 ponent (see above). The highest numbers of in- 0.05) are marked with asterisks.

Spicospina ﬂammocaerulea dividuals correctly reclassiﬁed were at Middle Ͻ

P Road 3 (60%) and Bow River (70%). Reclassiﬁ-

5.08 1.80 0.21 34.23 8.64 5.22 0.89 0.24 20.90 cation of individuals to their correct site at the 11) call of Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ

ϭ remaining populations was poor (Trent Road 2: N

( 45%; Mountain Road: 33%; Trent Road 1: 30%; Trend Rd 2 1.94 7.20 1.93 ement 18.53 38.59 40.42 26.70 and Boronia Road 2: 11%). 103.86 156.70 Mantel tests comparing dissimilarity matrices based upon geographic distance among study populations and average among-population var- 3.60 0.81 0.19 29.72 4.08 5.18 1.08 0.16 17.09 10) Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ Ϯ ϭ N (

Trent Rd 1 ABLE 2.02 7.90 2.00 T 3. Results of a canonical discriminant anal- 25.80 33.07 42.74 13.94

105.09 165.59 ysis of variation in call structure among populations of Spicospina ﬂammocaerulea. The relationship between the canonical discriminant axes and the component variables from a PCA are expressed as the canonical structure (CS).

CS 1 CS 2 CS 3 CS 4

2. Descriptive statistics for the advertis Component 1 0.18 Ϫ0.70 0.66 Ϫ0.18 Note duration Pulse number Dom. freq. Pulse rate Internote duration Note duration Pulse number Dom. freq. Pulse rate Component 2 0.89 0.28 Ϫ0.06 0.36

ABLE Component 3 0.01 Ϫ0.63 Ϫ0.68 0.38 1 1 1 1 2 2 2 2 T

Note Call property Component 4 0.16 Ϫ0.06 Ϫ0.41 Ϫ0.90 analysis of variation are also presented. Signiﬁcant values ( CALL VARIATION IN THE SUNSET FROG 289

FIG. 3. Distribution of the principle components FIG. 4. Relationship between the snout–vent along the first two canonical axes. The principle com- length (SVL) of male and female Spicospina flammo- ponents were derived from a PCA of the advertise- caerulea from amplecting pairs at Trent Rd 2. Solid ment call properties of Spicospina flammocaerulea (Ta- line ϭ line of best fit, dashed line ϭ 95% confidence ble 3). intervals. iation in each principal component failed to find pletion (Robertson, 1990; Byrne and Roberts, any significant relationships (Mantel test: com- 1999). This pattern of sexual selection is, how- ponent 1: r ϭϪ0.07, P ϭ 0.53; component 2: r ever, not inherently directional and will only re- ϭϪ0.13, P ϭ 0.67; component 3: r ϭϪ0.07, P sult in among-population divergence in call ϭ 0.48; and component 4: r ϭϪ0.10, P ϭ 0.58). structure if females differ in average size among Mating Size and Body Size.—Female S. flam- populations. Directional sexual selection on mocaerulea at Trent Rd 2 were on average larger body size has been reported in several frog spe- (x¯ Ϯ SD ϭ 35.84 Ϯ 2.21, N ϭ 27) than males cies (e.g., Ryan, 1983) and may reflect differenc- (30.62 Ϯ 1.83, N ϭ 27; ANOVA: F1,52 ϭ 89.52; P es in male quality (Halliday and Tejedo, 1995). Ͻ 0.01). The body sizes of males and females in Because we found no evidence that mated and amplectant pairs at Trent Rd 2 correlated sig- unmated male S. flammocaerulea differed in size nificantly (r ϭ 0.55; P Ͻ 0.01; Fig. 4), but the at Trent Rd 2, we suggest that size differences body size of mated males did not differ signif- between populations are not the outcome of lo- icantly from unmated males (ANOVA: F1,96 ϭ cal differences in the direction of sexual selec- 0.99; P ϭ 0.33). tion on body size. Accordingly, other explana- tions must be sought. DISCUSSION Apart from sexual selection, a number of fac- Unlike other studies of geographic variation tors have been invoked to explain geographic in anuran advertisement call structure (e.g., variation in frog calls including reinforcement Ryan and Wilczynski, 1991; Ryan et al., 1996), (Butlin, 1987; Loftus-Hills and Littlejohn, 1992; our data for S. flammocaerulea show little evi- Howard and Gregory, 1993), changes in the dence of variation among populations. This was acoustic environment (Ryan, 1988), or diver- evident in both univariate and multivariate gence associated with morphological change analyses. However, dominant frequencies were over the geographic range of the species (Nevo, lower at Middle Road, reflecting larger body 1973). Spicospina flammocaerulea has a restricted size at that site. This population was not the distribution with little climatic variation. Con- most geographically isolated of the six study sequently, the geographic variation in dominant populations, suggesting that the variation was frequency is unlikely to be the result of changes not caused by random drift. We also detected in climate or acoustic environments. Spicospina evidence of assortative mating by size at one site flammocaerulea is not closely related to any other because there was a positive correlation be- frog species in Australia (Roberts et al., 1997) tween the size of males and females in mated and is, therefore, unlikely to hybridize with any pairs. However, we found no evidence of an other frog species, eliminating reinforcement as overall size advantage. a cause of call structure variation. Size-related call components such as frequen- Middle Road 3 could be a population that cy are used by females to assess male size in a consists mostly of older frogs if size and age number of frog species (Gerhardt, 1994; Halli- are positively correlated as in other frog spe- day and Tejedo, 1995). Females may choose cies (cf. Halliday and Verrell, 1988). This could males of a particular size to increase fertilization account for the variation among populations in success (Davies and Halliday, 1977; Krupa, 1988; body-size and dominant frequency. There is Byrne and Roberts, 1999) or to avoid sperm de- some information available that suggests that 290 M. J. SMITH ET AL. there are differences in recruitment between LITERATURE CITED populations and other studies that have re- BURBIDGE,A.A.,AND J. D. ROBERTS. 2002. Sunset ported similar size-related variation in fre- Frog Recovery Plan. Western Australian Depart- quency have attributed it to among-population ment of Conservation and Land Management, differences in recruitment patterns (e.g., Geo- Perth, Western Australia, Australia. crinia rosea and Geocrinia lutea; Roberts and BUTLIN, R. 1987. Speciation by reinforcement. Trends Wardell-Johnson, 1995). Even though the life in Ecology and Evolution 2:8–13. history of S. flammocaerulea is poorly under- BYRNE,P.G.,AND J. D. ROBERTS. 1999. Simultaneous mating with multiple males reduces fertilization stood, intervals of several years have been ob- success in the Myobatrachid Frog Crinia georgiana. served between major breeding events at some Proceedings of the Royal Society of London B Bi- populations but not others (J. D. Roberts, un- ological Sciences 266:717–721. publ. data). The two Trent Road populations CASTELLANO,S.,T.D.GIACOMA,G.ODIERNA, AND E. and Boronia Road 2 have had calling males ev- BALLETTO. 1998. Morphometric and acoustical ery year since discovery in 1997 (Roberts et al., comparison between diploid and tetraploid green 1999), indicating the potential for breeding. No toads. Biological Journal of the Linnaean Society frogs have been heard calling at Mountain 63:257–281. DAVIES,N.B.,AND T. R. HALLIDAY. 1977. Optimal Road since 1998 (Roberts et al., 1999), but at mate selection in the toad Bufo bufo. Nature 269: the time of recording the population was com- 56–58. paratively large (at least 120 frogs; Roberts et DIEKHOFF, G. 1992. Statistics for the Social and Be- al., 1999). Bow River was only discovered in havioral Sciences: Univariate, Bivariate, Multivari- 2000 and was a large population (Ͼ 100 frogs; ate, Wm. C. Brown Publishers, Dubuque, IA. J. D. Roberts, unpubl. data). Before 2000, frogs DRISCOLL, D. A. 1998. Genetic structure, metapopu- at Middle Road 3 had not been heard calling lation processes and evolution influence the conservation strategies for two endangered frog spe- since 1997 (Roberts et al., 1999), implying that cies. Biological Conservation 83:43–54. there has been no recruitment through breed- FALCONER,D.S.,AND T. F. C. MACKAY. 1996. Intro- ing for several years. Anecdotal evidence sug- duction to quantitative genetics. Longman, Essex, gests that breeding in S. flammocaerulea can be U.K. stimulated by the removal of dense vegetation GERHARDT, H. C. 1994. The evolution of vocalization through disturbances such as fire or grazing in frogs and toads. Annual Review of Ecology and (Roberts et al., 1999). There has been no graz- Systematics 25:293–324. ing at Middle Road 3, but fuel-reduction burn- HALLIDAY,T.,AND M. TEJEDO. 1995. Intrasexual selection and alternative mating behaviour. In H. ing was conducted at the site in 1985 and in Heatwole and B. K. Sullivan (eds.), Amphibian Bi- 2000. The latest fuel-reduction burn may have ology, pp. 419–468. Surrey Beatty and Sons, Chip- stimulated the breeding event in 2000. Thus, ping Norton, New South Wales, Australia. variation in age structure among populations HALLIDAY,T.R.,AND P. A. VERRELL. 1988. Body size could occur if there are differences in recruit- and age in amphibians and reptiles. Journal of ment, and, if so, this would account for some Herpetology 22:253–265. of the among-population variation in adult HASEGAWA, Y., H. UEDA, AND M. SUMIDA. 1999. Clin- al geographic variation in the advertisement call of body size and dominant frequency in S. flam- the Wrinkled Frog, Rana rugosa. Herpetologica 55: mocaerulea. Accordingly, frogs at Middle Road 318–324. 3 might have been older and larger than at oth- HOWARD,D.J.,AND P. G. G REGORY. 1993. Post-insem- er sites. ination signalling systems and reinforcement. Phil- Contrary to expectation there was little or no osophical Transactions of the Royal Society of Lon- systematic variation in male call structure that don B Biological Sciences 340:231–236. could not be explained by variation in body size. KRUPA, J. J. 1988. Fertilization efficiency in the Great The variation that was detected was likely to re- Plains Toad (Bufo cognatus). Copeia 1988:800–802. LEE,A.K.,AND A. R. MAIN. 1954. Two new species flect differences in age structure and recruit- of burrowing frogs of the genus Helioporus Gray ment patterns. This suggests that either our pre- from south-western Australia. Western Australian dictions about gene flow are wrong or that there Naturalist 4:156–158. is strong stabilizing selection on male call struc- LITTLEJOHN, M. J. 1957. A new species of frog of the ture in S. flammocaerulea. genus Crinia. Western Australian Naturalist 6:18– 23. Acknowledgments.—This research was sup- LITTLEJOHN, M. J., M. J. FOUQUETTE, AND C. JOHNSON. ported by the National Heritage Trust and con- 1960. Call discrimination by female frogs of the ducted under the ethical guidelines of the Uni- Hyla versicolor complex. Copeia 1960:47–49. LOFTUS-HILLS,J.J.,AND M. J. LITTLEJOHN. 1992. Re- versity of Western Australia and with the per- inforcement and reproductive character displace- mission of the Western Australian Department ment in Gastrophryne carolinensis and G. olivacea of Conservation and Land Management. We (Anura: Microhylidae): a reexamination. Evolution thank M. Drew for assistance in the field. 46:896–906. CALL VARIATION IN THE SUNSET FROG 291

NEVO, E. 1973. Adaptive variation in size of cricket RYAN,M.J.,AND W. W ILCZYNSKI. 1991. Evolution of frogs. Ecology 54:1271–1281. intraspeciﬁc variation in the advertisement call of NEVO,E.,AND R. R. CAPRANICA. 1985. Evolutionary a cricket frog (Acris crepitans, Hylidae). Biological origin of ethological reproductive isolation in Journal of the Linnaean Society 44:249–271. cricket frogs, Acris. Evolutionary Biology 19:147– RYAN, M. J., A. S. RAND, AND L. A. WEIGT. 1996. Al- 214. lozyme and advertisement call variation in the ROBERTS,J.D.,AND G. WARDELL-JOHNSON. 1995. Call Tungara Frog, Physalaemus pustulosus. Evolution differences between peripheral isolates of the Geo- 50:2435–2453. crinia rosea complex (Anura: Myobatrachidae) in SEARCY,W.A.,AND M. ANDERSSON. 1986. Sexual se- southwestern Australia. Copeia 1995:899–906. lection and the evolution of song. Annual Review ROBERTS,J.D.,P.HORWITZ,G.WARDELL-JOHNSON,L. of Ecology and Systematics 17:507–533. R. MAXSON, AND M. J. MAHONY. 1997. Taxonomy, SHAFFER,H.B.,G.M.FELLERS,A.MAGEE, AND S. R. relationships and conservation of a new genus and VOSS. 2000. The genetics of amphibian declines: species of Myobatrachid frog from the high rainfall population substructure and molecular differen- region of southwestern Australia. Copeia 1997: tiation in the Yosemite Toad, Bufo canorus (Anura, 373–381. Bufonidae) based on single-strand conformation ROBERTS, J. D., S. CONROY, AND K. WILLIAMS. 1999. polymorphism analysis (SSCP) and mitochondri- Conservation status of frogs in Western Australia. al DNA sequence data. Molecular Ecology 9:245– In A. Campbell (ed.), Declines and Disappearances 257. of Australian Frogs, pp. 177–184. Natural Heritage SULLIVAN,B.K.,M.J.RYAN, AND P. A . V ERRELL. Trust, Canberra, Australian Capital Territory, Aus- 1995. Female choice and mating system struc- tralia. ture. In H. Heatwole and B. K. Sullivan (eds.), ROBERTSON, J. G. M. 1990. Female choice increases fer- Amphibian Biology, pp. 469–517. Surrey Beatty tilization success in the Australian frog, Uperoleia and Sons, Chipping Norton, New South Wales, laevigata. Animal Behaviour 39:639–645. Australia. RYAN, M. J. 1983. Frequency modulated calls and WATSON,G.F.,J.J.LOFTUS-HILLS, AND M. J. LITTLE- species recognition in a Neotropical frog. Journal JOHN. 1971. The Litoria ewingi complex (Anura: of Comparitive Physiology 150:217–221. Hylidae) in south-eastern Australia I. a new spe- . 1988. Constraints and patterns in the evolu- cies from Victoria. Australian Journal of Zoology 19:401–416. tion of anuran acoustic communication. In B. WELCH, A. M., R. D. SEMLITSCH, AND H. C. GERHARDT. Fritzsch, M. Ryan, W. Wilczynski, W. Walkowiak, 2001. Call duration as an indicator of genetic qual- and T. Hetherington (eds.), The Evolution of the ity in male gray tree frogs. Science 280:1928–1930. Amphibian Auditory System, pp. 637–677. John Wiley and Sons, New York. Accepted: 22 November 2002.