| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Behavioral Ecology Vol. 14 No. 2: 274-281
© 2003 International Society for Behavioral Ecology
The cause of correlations between nightly numbers of male and female barking treefrogs (Hyla gratiosa) attending choruses
Department of Biology, James Madison University, MSC 7801, Harrisonburg, VA 22807, USA
Address correspondence to C.G. Murphy. E-mail: murphycg{at}jmu.edu.
Received 27 March 2002; revised 10 May 2002; accepted 3 August 2002.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
The number of males displaying in a lek or chorus each day is often positively correlated with the number of females visiting or mating in the aggregation. I tested hypotheses that might explain such correlations in a study of the barking treefrog (Hyla gratiosa). Experimental reduction of the number of calling males did not reduce female visitation, ruling out the hypothesis that such correlations are owing to female preference for, or passive attraction to, larger choruses. Separate regression of the numbers of males and females on 13 environmental variables explained 45–74% of the variance in the nightly numbers of males and females. Partial regression coefficients for most of the 13 variables were not significantly different between the sexes, and the relative importance of variables in explaining variation in the numbers of individuals was similar for both sexes. These results support the hypothesis that positive correlations between nightly numbers of males and females are owing to a similar response by both sexes to the same environmental variables. Thus, it appears that the intense sounds emanating from choruses of H. gratiosa do not function in long-range communication and may instead be an epiphenomenon of intense, short-range vocal competition for females.
Key words: anuran amphibians, attendance, choruses, correlations, environmental variables, leks, mating success.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
In species that breed in leks or choruses, males gather at specific sites, where they display to females (Bradbury, 1981
; Höglund and Alatalo, 1995). Females visit these aggregations and choose mates from among the assembled males. Studies of such species have commonly revealed positive correlations between the number of displaying males and the number of females visiting or mating at aggregations and, less commonly, between aggregation size and per-capita male mating success (for review, Henzi et al., 1995; Höglund and Alatalo, 1995; Shelly, 2001; Sutherland, 1996). These correlations have been found both across aggregations of different average sizes and within a single aggregation across different days of a breeding season. Such correlations have been of interest primarily because of the insight they can provide about the selective forces favoring group display by males. For example, the female preference hypothesis (for review, see Höglund and Alatalo, 1995), which states that males display in groups because females prefer to mate with such males over solitary males, predicts that the number of females visiting an aggregation should increase with group size at a rate greater than linear.
Most of the interest in correlations between group size and mating success has focused on correlations across aggregations of different average sizes, but it is equally important to understand the causes of correlations between daily numbers of males and females. First, most of the costs and benefits to males and females of attending groups of different size (for review, see Höglund and Alatalo, 1995) should apply to within-season variation in group size of a particular aggregation, as well as to among-aggregation variation in average group size. Within-season variation in groups size can be quite large, with differences between the smallest and largest groups being as much as an order of magnitude (e.g., Bradbury et al., 1989; Henzi et al., 1995; Wagner and Sullivan, 1992) and as large as differences among groups on a given day (e.g., Bradbury et al., 1989; Green, 1990; Wagner and Sullivan, 1992). Hence, selection on choice of groups size is likely to be as strong for within-season choice as it is for among-aggregation choice. Second, several of the hypotheses proposed to explain the evolution of group display make predictions about within-season correlations and hence can be tested with studies of single aggregations. For example, predictions made by the female preference hypothesis (e.g., the number of females visiting an aggregation should increase with group size at a rate greater than linear) should apply to within-season, as well as among-group, comparisons. Thus, understanding the causes of within-season correlations for a single lek or chorus can provide insight into the evolution of group display (e.g., Henzi et al., 1995; Ryan et al., 1981).
Regardless of the ultimate costs and benefits of attending groups of various sizes, there are two general, proximate mechanisms that could produce positive correlations between the daily number of males and females attending a single aggregation. First, an individual's attendance at a lek or chorus may be directly influenced by the number of individuals of the opposite sex present at the aggregation (Alexander, 1975; Bradbury, 1981; Henzi et al., 1995). This mechanism seems most plausible for females, which are likely to be able to detect over substantial distances the intense auditory and visual signals produced by displaying males (Gerhardt and Klump, 1988; Phillips, 1990). Signal output from an aggregation (e.g., intensity, proportion of time that signals are broadcast) will be roughly correlated with the number of displaying males (Bradbury, 1981) and thus could provide a rough measure of group size. Alternatively, the greater signal output of larger groups may increase the area over which the group will contact females, resulting in greater attendance by females, even if females lack preferences for larger groups (the stimulus pooling or passive attraction model; Bradbury, 1981; Höglund and Alatalo, 1995). It has been widely assumed that sounds emanating from leks and choruses, especially those of anuran amphibians, function in long-range communication and that the intensity of these sounds affects female visitation rates (e.g., Gerhardt and Klump, 1988; Henzi et al., 1995; Wells, 1977).
An alternative hypothesis is that members of both sexes may respond in a similar fashion to the same environmental variables (e.g., temperature, rainfall). This hypothesis is suggested by the common observation that the numbers of males and females present in leks or choruses are correlated with environmental variables such as rainfall or temperature (Bradbury et al., 1989; Henzi et al., 1995; Sinsch, 1988; for a review, see Brooke et al., 2000). Males and females may respond to the same environmental variables because of benefits from attending the aggregation under particular conditions. For example, display and assessment of potential mates may be less costly for endotherms at higher temperatures than at lower temperatures (Vehrencamp et al., 1989). Alternatively, one sex may respond to particular conditions because they predict the abundance of members of the opposite sex, who also respond to these factors because of the benefits of mating or displaying under those particular conditions. For example, females of a lekking bird may visit leks early in the spring because that is the optimal time to initiate a clutch, and males may, in turn, display preferentially during that period because that is the time when females are most likely to be sexually receptive.
I report on a study of chorus size in the barking treefrog (Hyla gratiosa). First, I describe the within-season relationships between the number of calling males (chorus size) and the numbers of mating females (both absolute and per male). Second, I test the two hypotheses that might explain positive correlations between the nightly numbers of males and females. The hypothesis that males respond directly to the number of females present in the chorus can be ruled out because males arrive at the chorus and begin calling before females arrive (Murphy, 1999). However, females appear able to detect and respond to chorus sounds from as far away as 320 m (Gerhardt and Klump, 1988), so such sounds could potentially influence directly their decisions to visit the chorus. I tested this hypothesis by experimentally reducing the number of calling males and recording the number of females that arrived to mate. To test the hypothesis that males and females respond to the same environmental variables, I examined the relationship between environmental variables and the number of males and females present in the chorus.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
H. gratiosa breeds from March through August in ponds that lack fish. During the day, individuals spend their time in the canopy of the forest surrounding the breeding ponds, and individuals usually do not remain in breeding ponds during the day. Shortly after sunset, males descend from the canopy of forests to begin calling in these ponds. They inflate themselves and float on the surface of the water, calling from the perimeter of ponds for an average of 2.5 h (Murphy, 1999
). Females arrive at the pond shortly after males (Murphy, 1999), approaching males and nudging them to initiate amplexus. Males do not defend oviposition sites, rarely behave as satellites, do not pursue females, and do not attempt to dislodge already mated males (Murphy, 1992). Pairs remain near the male's call site for 1–2 h before departing to oviposit underwater elsewhere in the pond.
To examine within-season relationships between the number of calling males and mating females, my assistants and I recorded the numbers of calling males and mating females that visited the chorus at two ponds during three breeding seasons (1987, 1988, 1990). Descriptions of the ponds, lengths of study seasons, and monitoring techniques are given in Murphy (1994).
Exclusion experiment
To test the hypothesis that the greater signal output of larger choruses results in greater attraction of females, I conducted an experiment in 1997 in which I reduced the number of calling males. Nights were randomly assigned to either exclusion (n = 26) or control (n = 24) treatments. On exclusion nights, males were captured in a specially designed fence (Murphy 1993) as they arrived at the pond. The few males that entered the pond by jumping from branches overhanging the fence or that had daytime retreat sites inside the fence were captured by hand as soon as they began calling. Females arriving at the pond were placed inside the fence, which prevented them from leaving the pond. Captured males were placed in 0.5-l plastic containers with a small amount of water and transported to a larger plastic container located 55 m from the pond's edge. Males were held until 1.5 h after sunset; by this time, 85% of females have arrived at the pond (Murphy, 1999). Males were then released inside the fence at the point of their capture and allowed to call; calling began within 4–7 min of the release of males. I searched the pond for mated pairs 30 min after the start of calling; females take, on average, 15.9 min to select mates in natural choruses (range, 3.5–32.6 min; Murphy and Gerhardt, 2002). Any females arriving after males were released inside the fence but before completion of pair searches were excluded from the chorus, thereby excluding from female counts those females attracted by the chorus that formed after males were released inside the fence. On control nights, individuals of both sexes were placed inside the fence as they arrived, and I searched the pond for mated pairs beginning 1.5 h after sunset. During pair searches, any females arriving at the pond were excluded to prevent their inclusion in female counts.
Near the end of the study, the pond had filled so much that the inner barrier of the fence had to be removed, making it possible for individuals to leave the pond. Thereafter, I used a modified procedure on exclusion nights (n = 5); no change was needed on control nights. Instead of releasing arriving females inside the fence, we placed them in 0.5-l plastic containers with a small amount of water and held them just inside the fence at their point of capture. We released females 5 min after males started calling and searched the pond for mated pairs 30 min after the release of females.
Some males called when held in the large plastic container on exclusion nights, but these calls were not audible to humans at the pond and were thus below the threshold of hearing of female H. gratiosa. Over the range of frequencies encompassed by the calls of H. gratiosa, human thresholds of hearing are approximately 0–10 dB (Gelfand, 1997), whereas most female H. gratiosa do not respond phonotactically to chorus sounds with sound pressure levels of 32 dB (Gerhardt and Klump, 1988). Neural thresholds for male calls, as measured in the midbrains of female H. gratiosa, are greater than 27 dB (Schwartz JJ, unpublished data).
In all analyses of the exclusion experiment, I used the number of females found mated, rather than the number of females captured in the fence on arrival, because some females visit the pond to obtain water rather than to mate (Murphy, 1992). Neither the results nor the conclusions of this study are altered by substituting the number of arriving females or the operational sex ratio (OSR = females/males) for the number found mated.
Environmental variables and numbers of individuals
To determine the influence of environmental variables on the number of calling males and mating females present in the chorus each night, I constructed regression models for each sex separately by using 13 independent variables that could plausibly be assessed directly by frogs at or near their retreat sites (Table 1). Temperature and humidity were measured at 2000 h because this was the time closest to the initiation of calling. Quadratic terms were used for temperature variables to investigate the possibility that there may be temperature optima (indicated by a significant negative regression coefficient). Quadratic terms were expressed as deviations from the mean to reduce correlations between quadratic and linear terms (Neter et al., 1990). To normalize residuals, the number of males and females were square-root transformed. Outlier analysis failed to detect any influential outliers. Durbin-Watson tests for serial correlation of residuals (Draper and Smith, 1998), which might occur if the number of individuals on a day were correlated with the numbers the next day, revealed a correlation for only one analysis (males in 1988) that was marginally significant (DW = 1.01, critical value
=0.01 = 1.07, sequential Bonferonni for rejection of H0 = 0.05/6 = 0.008).
|
Measurements of rainfall during 1988 and 1990 and maximum and minimum temperatures during 1990 were made at the study pond; all other measurements were made by the U.S. Weather Service at the Tallahassee Airport, located 4 km from the study site. Maximum and minimum temperatures recorded at the study site in 1990 were strongly correlated with those recorded at the airport (Pearson product moment correlation for maximum temperature: r =.94, n = 70, p <.001; for minimum temperature: r =.92, n = 70, p <.001). The correlation between rainfall at the two sites was slightly lower (Spearman rank correlation for 1988: rs =.80, n = 72, p =.0001; for 1990: rs =.76, n = 79, p =.0001). When data normally collected at the pond were not available, I substituted the appropriate data from the U.S Weather Service.
Power analyses
Power analyses are presented when interpretations of nonsignificant results are important to the conclusions of the study. Results of power analyses are presented as the minimum effect detectable with 80% power (Cohen's [1988] recommendation for a desirable power). The power of correlation tests was calculated with standard techniques (Zar, 1996), and the power of Mann-Whitney tests was approximated by using simulation. In each run of the simulation, values were randomly generated for each group (control and experimental) from a Poisson distribution using the means of the observed data. A Mann-Whitney test was calculated, and the simulation repeated 5000 times. Power was calculated as the proportion of the 5000 simulations in which the calculated test statistic exceeded the critical value for = 0.05. To determine the minimum effect detectable with 80% power, I decreased the average for the experimental group until the point at which 80% of the 5000 runs resulted in a significant Mann-Whitney test (i.e., power was 80%).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Numbers of males and females visiting the chorus
Breeding by H. gratiosa was relatively continuous during all three study seasons, with females mating at the chorus on the vast majority of nights when censuses of mated pairs were conducted (Table 2). Large numbers of individuals of both sexes visited the chorus, but only a small proportion of these individuals visited the chorus on any given night, and there was considerable variation among nights in the number of calling males and mating females (Table 2). Males outnumbered females on nearly all nights, and OSRs were low (Table 2). The numbers of calling males present in the chorus each night was positively correlated with the number of mating females during all three years (Spearman rank correlations, 1987: rs =.69, n = 92, p =.0001; 1988: rs =.68, n = 71, p =.0001; and 1990: rs =.89, n = 76, p =.0001; Figure 1) and with the OSR in two of the three years (Spearman rank correlations, 1987: rs =.31, n = 92, p =.003; 1988: rs =.06, n = 60, p =.65; and 1990: rs =.42, n = 73, p =.0003; Figure 1).
|
|
Exclusion experiment
The exclusion experiment substantially reduced signal output from the chorus. On exclusion nights, only a median of two males (range, 0–5; n = 26) called, compared with a median of 14 males on control nights (range, 0–26; n = 24, p =.0001, Mann-Whitney U test). The few males that called on exclusion nights produced, in total, a median of only 14 calls and a maximum (65) only slightly greater than the average produced by a male in only 1 min (55; Murphy and Gerhardt, 1996
). However, excluding males from the pond did not affect the number of males arriving at the chorus (control nights: median, 14; range, 0–26; experimental nights: median, 16; range, 0–30; p =.20, Mann-Whitney U test) or the proportion of females arriving at the chorus that were later found mated (control nights: median, 0.8; range, 0.5–1; experimental nights: median, 0.75; range, 0–1; p =.48, Mann-Whitney U test). Excluding males did not reduce the number of females mating in the chorus either on the night of the manipulation or on the following night. On exclusion nights, a median of 2.5 (interquartile range [IQR], 1–4; range, 0–9) females mated after the release of males; in fact, as many as five females (median, 2) arrived at the chorus to mate on the five nights when no calls were produced during the exclusion period. On control nights, a median of three females (IQR, 1.5–4.5; range, 0–13) mated; this median was not significantly greater than that for exclusion nights (p =.65, Mann-Whitney test). This Mann-Whitney test could have detected with 80% power a median for experimental nights of 2.13 or less (a 29% reduction in the control median). Similarly, there was no significant difference between the number of females mating on nights following exclusion nights (median, 3; n = 24) and on nights after control nights (median, 2; n = 20, p =.62, Mann-Whitney test).
To examine the interaction between treatment on a night and treatment on the previous night, I conducted a two x two ANOVA by using these variables as the two factors and transforming (square-root [x + 0.5]) the number of mating females to meet the assumptions of the test. I excluded nights for which the chorus had not been monitored on the previous night, and therefore, sample sizes for this test were smaller than for the Mann-Whitney tests described above. There was no effect of either factor on the number of mating females (overall model: F = 1.67, df = 3,39, p =.19; night of manipulation: F = 1.53, df = 1,39, p = 0.28; previous night: F = 0.00, df = 1,39, p =.99; interaction term: F = 3.80, df = 1,39, p =.06).
Because the above tests did not have power to detect small effect sizes, the lack of significant differences between treatment groups could be owing to lack of either an effect of exclusion or insufficient statistical power. Therefore, I conducted an analysis in which a lack of effect of exclusion would cause rejection of a statistical null hypothesis. To do this, I compared the number of females mating on exclusion nights with that expected based on the number of males that actually called during the exclusion period. If the exclusion experiment failed to reduce the number of mating females, then the observed number of females should exceed the expected number. I calculated the expected number by substituting the number of calling males on exclusion nights into the regression equation (y = 0.27x - 0.27; r2 =.38) describing the relationship between the numbers of females and males on control nights. Each male that produced at least one call during the exclusion period was conservatively assumed to be detectable to females as a unique male. The dependent variable was not transformed because the equation was used to predict the number of females, rather than assess the statistical significance of the relationship; the Spearman rank correlation between the numbers of males and females on control nights was highly significant (rs =.70, p =.0008). On 23 of the 26 exclusion nights, more females mated than predicted by the number of males calling during the exclusion period (p =.0001, Wilcoxon signed-ranks test).
The above analyses examined short-term effects of the exclusion. To determine whether exclusion produced a longer-term effect, I examined how the number of females attending the chorus changed over the course of the study season (3 June through 31 July) and compared this change to that of other seasons over the same period. Because transformations of the number of females failed to normalize residuals and remove heterogeneity of variances among seasons, it was not possible to compare among seasons the slope of the regression of females on day of the season. Instead, I measured seasonal change in the numbers of females as the Spearman rank correlation between day of the study season (day 1 = 3 June) and the number of females present in the chorus, and I compared correlations using the z-transformation method (Sokal and Rohlf, 1981).
The number of mating females did not decrease more rapidly over the course of the experiment (Spearman rank correlation, rs = -.31, n = 50) than over the same period in three other seasons (1987: rs = -.42, n = 48; 1988: rs = -.33, n = 58; and 1990: rs = -.21, n = 58; z-transformation test for a difference among the four correlation coefficients: 
2 = 1.38, df = 3, p =.85). A pairwise z-transformation test could have detected with 80% power a correlation of -0.59 that was significantly different ( = 0.05) from the average correlation coefficient (-0.32) and average sample size (54) for the three other seasons.
Environmental variables and the numbers of individuals
Regression models incorporating 13 environmental variables explained 45–70% of the variance in the nightly numbers of males and females present at the chorus (Table 3). Three variables were significantly related to the numbers of both sexes in all three years. More males and females were present earlier in the season, on warmer nights, and on nights with more rain during the previous 24–48 h. An additional variable, barometric pressure, was positively related to the numbers of both sexes in all years, except for females in 1990. Two variables (rainfall during that day, quadratic term for temperature at 2000 h) were significantly related to the number of both sexes in different years, with more individuals present in 1990 on nights with more rain during the previous 24 h, and in 1987 on nights with intermediate temperatures at 2000 h. The number of females was significantly and positively related to relative humidity at 2000 h in two years (1987 and 1988), and the number of males was significantly and negatively related to the quadratic term for maximum temperature in one year (1987).
|
If positive correlations between the numbers of males and females are owing to similar responses by the two sexes to the same environmental variables, then the partial regression coefficients describing the relationship between each variable and the numbers of individuals present should be similar for both sexes. To test this prediction, I first fitted a common regression equation for both sexes and then fitted separate regression equations for each sex by using a model containing an indicator for sex and an interaction term between the indicator and each environmental variable (Neter et al., 1990
). Statistically significant improvement in the fit by the model with separate regression equations over the model with a common equation would indicate that the regression equations for the two sexes differed in their intercepts, slopes, or both. Significant differences between the sexes in the slope coefficients for individual variables would be indicated by significant indicator-by-variable interaction terms. Coefficients did not differ significantly between the sexes in 1987 (F = 1.38, df = 13,156, p =.24) or 1988 (F = 0.85, df = 13,114, p =.61), but three differed in 1990 (day: p = 0.028; rainfall: p =.007; rainprev: p = 0.009; overall test: F = 24.49, df = 13,124,p =.0001). In all three cases, the signs of the regression coefficients were the same for both sexes (Table 3).
The hypothesis that the two sexes are responding to the same environmental variables also predicts that the relative importance of each variable in explaining variation in the number of individuals should be similar for both sexes. To test this prediction, I calculated, for each year of the study, the Spearman rank correlation between the amount of variation explained by each variable for both sexes. I used type II squared semipartial correlation coefficients from the above regression analyses as an estimate of the importance of each variable; these coefficients give the percentage of the total variance in the dependent variable that is owing to the unique contribution of each independent variable (Tabachnick and Fidell, 1983). In all three years, there was a positive correlation between the sexes in the amount of variance explained by each variable (Spearman rank correlations between male and female type II squared semipartial correlation coefficients, 1987: rs =.71, n = 13, p =.014; 1988: rs =.62, n = 13, p =.030; and 1990: rs =.95, n = 13, p =.001).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Hypotheses to explain positive correlations between numbers of males and females
In all three years, the nightly number of male H. gratiosa calling in the chorus was correlated with the number of females mating in the chorus. These correlations did not appear to result from direct assessment by either sex of the numbers of individuals of the opposite sex. Males cannot directly assess the number of females before departing for the chorus from their daytime retreat sites because males arrive at the chorus before females (Murphy, 1999
). Females appear not to assess the number of calling males, because excluding males from the chorus did not significantly reduce the number of females arriving at the chorus. It should be noted that the difference in the median number of females between control and exclusion nights was in the predicted direction, and that the Mann-Whitney test used to compare these medians did not have sufficient statistical power to detect what might be considered realistic effect sizes. However, two results suggest that the failure of the exclusion experiment to reveal a significant reduction in female visitation is owing to a lack of an effect rather than to low statistical power. First, females arrived to mate even on nights when no calls were produced at the pond. It is difficult to imagine how variation in signal output from the chorus could affect female visitation if females arrive to mate at the pond in the absence of any signals. Second, significantly more females arrived to mate on experimental nights than were predicted from the number of males calling during the exclusion period. This analysis tested for a lack of effect of exclusion with the rejection, rather than acceptance, of the null hypothesis, thereby avoiding the power issue associated with the comparison of medians. Taken together, the results of the exclusion experiment support the conclusion that excluding males did not affect female visitation to the chorus. Because males were eventually allowed to call on exclusion nights, the exclusion experiment cannot rule out the possibility that females assessed the average size of the chorus by averaging signal intensity over multiple nights, assessing chorus size on control nights during the usual chorusing period, and on experimental nights during the period after males were released inside the fence. However, the results do rule out the possibility that nightly variation in numbers of females is the result of nightly variation in signal output from choruses. Therefore, the observed correlations between the numbers of males and females in the chorus cannot be explained by a direct influence of chorus size on visitation by females.
It has been widely assumed that sounds emanating from choruses of anuran amphibians function in long-range communication and that the intensity of these sounds affects female visitation rates (Gerhardt and Klump, 1988; Henzi et al., 1995; Wells, 1977). The results of the exclusion experiment call into question this assumption for H. gratiosa, a surprising finding given that females of this species are likely to be able to detect chorus sounds from as far away as 320 m (Gerhardt and Klump, 1988). Nonetheless, it appears that the intense signals emanating from choruses of H. gratiosa do not, in fact, function in long-range communication and that the long distance over which choruses are audible is perhaps better viewed as an epiphenomenon of intense, short-range vocal competition for females.
Only one other study of an anuran amphibian has attempted to test experimentally the hypothesis that groups of males attract females. Schwartz (1994) gave female Hyla microcephala choices between groups of varying numbers of speakers (one versus two, one versus seven, and two versus six) broadcasting synthetic male calls. Females chose the larger group of speakers more often in two of the three tests (one versus seven, and two versus six), but speakers in the larger groups did not attract more females per speaker than did smaller groups. However, interpretation of these results is complicated by the fact that the distance between speaker groups was small relative to that between natural choruses; females may have interpreted the choice presented them as being among different males within a single chorus, rather than between different choruses (Schwartz, 1994). Additional experiments with other species of anuran amphibians are needed to determine whether the results of the present study represent the rule or the exception.
Experiments with birds and insects support the hypothesis that visitation to leks or choruses by females is influenced directly by the numbers of males present. Experiments with insects in which the numbers of males (Walker, 1983), number of loudspeakers broadcasting male song (Cade, 1981; Doolan, 1981; Shelly and Greenfield, 1991), or extent of resources on which leks form (Jones and Quinnell, 2002) were manipulated, have demonstrated that larger groups attract more females than smaller groups, although only two of these studies (Cade, 1981; Doolan, 1981) found that the number of females attracted per male was greater for groups than for solitary males. Experiments with birds in which either the number of males (Lank and Smith, 1992) or loudspeakers broadcasting male vocalizations (Hovi et al., 1997) demonstrated that larger groups attract more females than smaller groups, both absolutely and on a per-male basis.
For H. gratiosa, positive correlations between the numbers of males and females attending the chorus seem best explained, not by greater attraction of females by larger choruses, but by similar responses of both sexes to the same environmental variables. A multiple regression model incorporating 13 environmental variables explained substantial variance in the nightly numbers of males and females, and comparisons of the models for the two sexes revealed that males and females respond in a similar manner to the same variables. It is not possible to determine, without additional experimentation, whether both sexes respond to the same variables because each sex benefits from attending the chorus under particular conditions or because one or both sexes benefit from using environmental variables to predict the abundance of members of the opposite sex.
Three other studies of anuran amphibians have conducted multiple regression analyses to compare the responses of both sexes to environmental variables. Sinsch (1988) found that the numbers of males and females natterjack toads (Bufo calamita) were related to the same two environmental variables. Ritke et al. (1992) determined that the only significant predictor of the numbers of male and female gray treefrogs (Hyla chrysoscelis) was the number of individuals of the opposite sex. Henzi et al. (1995) found that the numbers of male and female painted reed frogs (Hyperolius marmoratus) were predicted by different regression models.
One likely explanation for why Sinsch (1988) and the present study found that the two sexes responded to the same environmental variables, whereas Ritke et al. (1992) and Henzi et al. (1995) did not, is that these latter two studies included in their regression models the numbers of individuals of the opposite sex, whereas the former two did not. Ritke et al. (1992) included the numbers of the opposite sex in both male and female models. Henzi et al. (1995) included male number in the female model, assuming that females base their decision to visit the chorus in part on the number of calling males (detected via signal intensity), but did not include female number in the male model because choruses of H. marmoratus form before females arrive.
Inclusion of the number of individuals of the opposite sex in both male and female models for H. gratiosa reduced both the number of statistically significant environmental variables for both sexes (females: three in 1987, zero in 1988, and one in 1990; males: four in 1987, three in 1988, six [no change] in 1990; cf. Table 3) and, more importantly, the number of significant environmental variables common to both sexes (one in 1987, zero in 1988, and one in 1990; cf. Table 3). Including the number of individuals of the opposite sex in only the female model also reduced the number of significant environmental variables common to both sexes (two in 1987, 0 in 1988, and one in 1990). Thus, including the numbers of the opposite sex into regression models would lead us to conclude that male and female H. gratiosa do not respond to the same environmental variables.
When investigating the effect of environmental variables on attendance by males and females, inclusion of the number of individuals of the opposite sex in regression models will be warranted only if individuals are directly influenced by the numbers of individuals of the opposite sex. Because male H. chrysoscelis arrive at the chorus before females (Ritke et al., 1990), it seems unlikely that males of this species can assess directly the numbers of females. Hence, inclusion of the number of females in the male model by Ritke et al. (1992) seems unwarranted. The results of the exclusion experiment in the present study suggest that we should be cautious in assuming that females are influenced directly by male numbers, even when the numbers of males is positively correlated with absolute and per-male number of females. Until the assumption of direct influence is verified for H. chrysoscelis and H. marmoratus, it may be premature to conclude that males and females of these two species respond either only to the numbers of individuals of the opposite sex (H. chrysoscelis; Ritke et al., 1992) or to different environmental variables (H. marmoratus; Henzi et al., 1995).
Evolution of chorusing in H. gratiosa
The results of the exclusion experiment bear directly on two hypotheses that might explain why male H. gratiosa call in choruses. The stimulus-pooling, or passive-attraction, hypothesis (Bradbury, 1981; Höglund and Alatalo, 1995) states that females do not compare choruses but instead visit the first chorus that they detect when they become receptive. Under this hypothesis, males display in groups because groups produce more continuous or more intense signals, and hence contact more females, than do the signals of solitary males. Because excluding males from the chorus did not reduce the number of visiting females, signal output from the chorus cannot influence female visitation, and it seems unlikely that the stimulus pooling hypothesis can explain the evolution or maintenance of chorusing in this species. These results add experimental support to the theoretical arguments against stimulus pooling as an explanation for group display (Bradbury, 1981).
The female preference hypothesis (for review, see Höglund and Alatalo, 1995) states that females prefer to mate with lekking or chorusing males over solitary males because groups offer greater choice of mates or greater protection from predation than do solitary males. This hypothesis predicts that females should prefer larger choruses over smaller choruses. To exert such a preference, females must sample more than one chorus, and signal output from choruses could allow females to assess chorus size without incurring the costs of visiting a chorus. However, female H. gratiosa visited the manipulated chorus even on exclusion nights when no males called and despite an active chorus only 100 m from the manipulated chorus, indicating that females do not exert choice among choruses on the night they mate. This result, in conjunction with the failure of the exclusion of males to reduce female visitation on the night after manipulation, and the lack of any long-term decline in female visitation owing to the manipulation, makes it unlikely that female H. gratiosa exert choice among choruses based on chorus size. Thus, it seems unlikely that chorusing evolved or is maintained in H. gratiosa by female preferences for larger choruses.
The selective forces favoring the evolution or maintenance of chorusing by H. gratiosa remain unknown, but there are several possibilities. One likely hypothesis is that males gather in breeding ponds because of the high probability of encountering receptive females (i.e., the hotspot hypothesis; for review, see Höglund and Alatalo, 1995), either because females must visit the pond to oviposit or because females reject males that might call away from the pond, thereby eliminating the cost of carrying a male to the pond. Additionally, choruses may form because less attractive males choose to display near attractive males (i.e., the hotshot hypothesis; for review, see Höglund and Alatalo, 1995). Reduced risk of predation does not appear to be a factor favoring chorusing in this species, as the per-capita risk of predation for males increases with chorus size (Murphy 1992). Additional experiments are needed to test the hotspot and hotshot hypotheses for H. gratiosa.
|
ACKNOWLEDGEMENTS |
|---|
I thank the U.S. Forest Service for permission to conduct this study. Mike Blouin and Sharon Forster-Blouin helped in numerous ways, and Joe Travis and the Department of Biology at Florida State University provided logistical support. Rickie Domangue developed the program used to calculate power for Mann-Whitney tests. For field assistance, I thank Samantha Bates, Jim Bradley, Katy Duggan-Haas, Jessica Geyer, Veronique LaCapra, Erika Nowak, Tigerin Peare, Debbie Poelker, Shapell Randolph, Julie Spellerberg, Lei Ward, and Jolyne Woodmansee. Funding was provided by the Florida Nongame Wildlife Grant Program, Sigma Xi Grants-in-Aid of Research, the Theodore Roosevelt Memorial Fund of the American Museum of Natural History, and a Summer Faculty Assistance Grant from the College of Science and Mathematics, James Madison University. For helpful discussion and comments on previous versions of this paper, I thank Carl Gerhardt, Reid Harris, Steve Emlen, Paul Sherman, David Westneat, and two anonymous reviewers.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Alexander, RD, 1975. Natural selection and specialized chorusing behavior in acoustical insects. In: Insects, science and society (Pimentel D, ed). New York: Academic Press; 35–77.
Bradbury JW, 1981. The evolution of leks. In: Natural selection and social behavior (Alexander RD, Tinkle DW, eds). New York: Chiron; 138–169.
Bradbury JW, Vehrencamp SL, Gibson RM, 1989. Dispersion of displaying male sage grouse. I. Patterns of temporal variation. Behav Ecol Sociobiol 24:1-14.
Brooke PN, Alford RA, Schwartzkopf L, 2000. Environmental and social factors influence chorusing behavior in a tropical frog: examining various temporal and spatial scales. Behav Ecol Sociobiol 49:79-87.
Cade, WH, 1981. Field cricket spacing and the phonotaxis of crickets and parasitoid flies to clumped and isolated cricket songs. Z Tierpsychol 53:365-375.
Cohen J, 1988. Statistical power analysis for the behavioral sciences, 2nd ed. Hillsdale,. New Jersey: Lawrence Erlbaum Associates.
Doolan JM, 1981. Male spacing and the influence of female courtship behavior in the bladder cicada, Cystosoma sundersii. Westwood. Behav Ecol Sociobiol 9:269-276.[CrossRef]
Draper NR, Smith H, 1998. Applied regression analysis, 3rd ed. New York: Wiley.
Gelfand SA, 1997. Essentials of audiology. New York: Thieme.
Gerhardt HC, Klump GM, 1988. Phonotactic responses and selectivity of barking treefrogs (Hyla gratiosa) to chorus sounds. J Comp Physiol A 163:795-802.
Green AJ, 1990. Determinants of chorus participation and the effects of size, weight and competition on advertisement calling in the tungara frog, Physalaemus pustulosus (Leptodactylidae). Anim Behav 39:620-638.
Henzi SP, Dyson ML, Piper SE, Passamore NE, Bishop P, 1995. Chorus attendance by male and female painted reed frogs (Hyperolius marmoratus): environmental factors and selection pressures. Funct Ecol 9:485-491.
Höglund J, Alatalo RV, 1995. Leks. Princeton, New Jersey: Princeton University Press.
Hovi M, Alatalo RV, Halonen M, Lundberg A, 1997. Responses of male and female black grouse to male vocal display. Ethology 103:1032-1041.
Jones TM, Quinnell RJ, 2002. Testing predictions for the evolution of lekking in the sandfly Lutzomyia longipalps. Anim Behav 63:605-612.[CrossRef]
Lank DB, Smith CM, 1992. Females prefer larger leks: field experiments with ruffs (Philomachus pugnax). Behav Ecol Sociobiol 30:323-329.
Murphy CG, 1992. The mating system of the barking treefrog (Hyla gratiosa) (PhD dissertation). Ithaca, New York: Cornell University.
Murphy CG, 1993. A modified drift fence for capturing treefrogs. Herp Rev 24:143-145.
Murphy CG, 1994. Chorus tenure of male barking treefrogs (Hyla gratiosa). Anim Behav 48:763-777.[CrossRef]
Murphy CG, 1999. Nightly timing of chorusing by male barking treefrogs (Hyla gratiosa): the influence of female arrival and energy. Copeia 1999:333–347.
Murphy CG, Gerhardt HC, 1996. Evaluating experimental designs for determining mate choice: the effect of amplexus on mate choice by female barking treefrogs, Hyla gratiosa. Anim Behav 51:881-890.[CrossRef]
Murphy CG, Gerhardt HC, 2002. Mate sampling by female barking treefrogs (Hyla gratiosa). Behav Ecol 13:472-480.
Neter J, Wasserman W, Kutner MH, 1990. Applied linear statistical models. Homewood, IL: Irwin.
Phillips JB, 1990. Lek behavior in birds: do displaying males reduce nest predation? Anim Behav 39:555-565.
Ritke ME, Babb JG, Ritke MK, 1990. Life history of the gray treefrog (Hyla chrysoscelis) in western Tennessee. J Herpetol 24:135-141.
Ritke ME, Babb JG, Ritke MK, 1992. Temporal patterns of reproductive activity in the gray treefrog (Hyla chrysoscelis). J Herpetol 26:107-111.
Ryan MJ, Tuttle MD, Taft LK, 1981. The costs and benefits of frog chorusing behavior. Behav Ecol Sociobiol 8:273-278.[CrossRef]
Schwartz JJ, 1994. Male advertisement and female choice in frogs: recent findings and new approaches to the study of communication in a dynamic acoustic environment. Am Zool 34:616-624.
Shelly TE, 2001. Lek size and female visitation in two species of tephritid fruit flies. Anim Behav 62:33-40.[CrossRef]
Shelly TE, Greenfield MD, 1991. Dominions and desert clickers (Orthoptera: Acrididae): influences of resources and male signaling on female settlement patterns. Behav Ecol Sociobiol 28:133-140.[CrossRef]
Sinsch U, 1988. Temporal spacing of breeding activity in the natterjack toad, Bufo calamita. Oecologia 76:399-407.
Sokal RR, Rohlf FJ, 1981. Biometry, 2nd ed. New York: Freeman.
Sutherland WJ, 1996. From individual behaviour to population ecology. Oxford: Oxford University Press.
Tabachnick BG, Fidell LS, 1983. Using multivariate statistics. New York: Harper and Row.
Vehrencamp SL, Bradbury JW, Gibson RM, 1989. The energetic cost of display in male sage grouse. Anim Behav 38:885-896.[CrossRef]
Wagner WE, Sullivan BK, 1992. Chorus organization in the Gulf Coast toad (Bufo valliceps): male and female behavior and the opportunity for sexual selection. Copeia 1992:647–658.
Walker TJ, 1983. Mating modes and female choice in short-tailed crickets (Anurogyllus arboreus). In: Orthopteran mating systems: sexual competition in a diverse group of insects (Gwynne DT, Morris GK, eds). Boulder, Colorado: Westview; 240–267.
Wells KD, 1977. The social behavior of anuran amphibians. Anim Behav 25:666-693.[CrossRef]
Zar JH, 1996. Biostatistical analysis, 3rd ed. Upper Saddle River,. New Jersey: Prentice Hall.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||