Can Frogs Have Sex Continuously for Up to 48 Hours
Abstract
Preference functions, which quantify preference strength relative to variation in male traits or signals, are central to understanding mechanisms and consequences of female choice. Female tree frogs (Hyla versicolor) choose mates on the basis of advertisement calls and prefer long calls to short calls. Here we show, in two experimental designs, that preference strength increased significantly as the difference in call duration was increased only if the absolute durations of alternative stimuli were below average. Hence preference strength was a non-linear function of duration, and females did not base preferences solely on the percentage difference in duration. In experiments simulating costly choice (unequal playback levels), non-linear effects were more pronounced than in the conventional design (equal playback levels). Repeated estimates of preference strength using the unequal-playback design revealed significant among-female variation. These patterns of preference suggest that selection by female choice for males producing calls of average duration over males producing very short calls is stronger than selection for males producing very long calls over males producing calls of average duration. Female preferences, especially in tests simulating a potentially costly choice, could reflect differences in the net benefits to females of mating with males producing calls of different duration.
A first step in evaluating theoretical models concerned with female mate choice is to quantify preferences based on differences in male courtship signals (Andersson, 1994). Preference functions, which show the relationship between preference strength and variation in male signals, defines the form of selection on male signals and can influence the co-evolution of preferences and signals (Pomiankowski, 1988). Despite their importance for sexual selection theory, robust estimates of preference functions are available for only a few species (guppies: Basolo, 1990, 1998; tree frogs: Gerhardt, 1991; Murphy and Gerhardt, 2000; crickets: Hedrick and Weber, 1998; Wagner et al., 1994; Shaw and Herlihy, 2000; Shaw, 2000; katydids: Ritchie, 1996).
Studies of preference functions are often limited in one or two ways. First, if too few signal variants are tested, the shape of the preference function can only be crudely estimated. More precise estimates are required, for example, to distinguish between stabilizing and weakly directional selection if the basic function is unimodal (review; Gerhardt, 1991), and directional preferences can be non-linear (e.g., Basolo, 1990). That is, although higher values of some trait or property are generally favored, some ranges of values may be much more strongly favored or disfavored than other ranges. Second, despite the key assumption of most theoretical models that heritable variation in preference exists (review; Andersson, 1994), most studies provide only population-level estimates of preference functions, that is, the proportions of females choosing among different signals (reviews by Gerhardt, 1991; Jennions and Petrie, 1997; Wagner, 1998). Although a few studies have demonstrated genetic variation in preference (Bakker and Pomiankowski, 1995), other studies have failed to demonstrate significant phenotypic, much less genotypic variation (Butlin, 1993; Gerhardt, 1991; Gerhardt et al., 1996; Kime et al., 1998). Whether these failures reflect a real lack of variability or the insensitivity of behavioral assays is an open question.
Our present study, which tested many signal variants spanning the range of variation in male calls, was designed to generate detailed preference functions based on call duration in the gray tree frog (Hyla versicolor). Call duration is a highly repeatable property in this species (Gerhardt et al., 1996; Runkle et al., 1994; Sullivan and Hinshaw, 1992). In one of our designs, repeated testing of individuals also allowed us to estimate individual consistency and among-female variation in preference strength. We focused on call duration for two reasons. First, previous experiments (Gerhardt et al., 1996) suggested that the preference function might be non-linear. Specifically, in tests that simulated costly choice, females preferred calls of average duration to very short calls more strongly than they preferred calls of very long duration to calls of average duration. Males might thus be more likely to increase the probability of attracting a female by lengthening their calls if they are producing short calls than if they are producing calls of at least average duration. Second, comparisons of larval fitness traits (growth rate, size at metamorphosis, larval period) showed significant differences between the half-sibling offspring of long and short callers (Welch et al., 1998), thus indicating that the consequences of choices based on call duration, which is a reliable indicator of energetic costs to the male (Wells and Taigen, 1986), can be a selective force on female preferences. More detailed information about the shape of the preference function, and the demonstration of among-female variation in preference strength set the stage for future research concerning the heritability of preferences and thus the potential for an evolutionary response to selection.
METHODS
Overview of experiments
We used two playback designs that take advantage of the fact that female gray tree frogs initiate matings with calling males and prefer long calls to short calls (Gerhardt, 1991; Gerhardt et al., 1996; Klump and Gerhardt, 1987; Sullivan and Hinshaw, 1992). In the first, conventional design (equal playback level) population-level estimates of preference strength were generated in terms of the proportions of females choosing the longer of two alternatives. In the second design (unequal playback level), the preference strength of individual females was estimated in terms of the sound pressure level (SPL) difference favoring a short-call alternative at which the preference for the long-call alternative was maintained. This experimental design, and others employed in the study by Gerhardt et al. (1996), simulates potentially costly choices because the distance from a sound source is negatively correlated with playback level (see Discussion). We used the same set of eight alternatives in both of these designs (see Table 1). In addition, for three of these alternatives and one additional alternative, repeated testing using the second design provided data about the consistency of individuals and among-female variation in preference strength.
Table 1
Acoustic stimuli used in choice-tests of females of Hyla versicolor
| Alternatives | Percentage difference | Call period(s) | Call duty cycles |
|---|---|---|---|
| The stimuli of Series 1 and 2 were used in both the equal and unequal playback designs. The stimulus listed under "Additional test" was tested in the unequal playback design. | |||
| aEstimates from call data collected during 1987 (n = 168 males) (Gerhardt et al., 1996). | |||
| bTests in which females were tested twice in the unequal-playback design. | |||
| Series 1: short-call (≈ mean = 17.8 pulses/calla) | |||
| 18 vs. 20 pulses/call | 11.1% | 4.50 | 0.194 vs. 0.217b |
| 18 vs. 22 pulses/call | 22.2% | 4.50 | 0.194 vs. 0.239 |
| 18 vs. 24 pulses/call | 33.3% | 4.50 | 0.194 vs. 0.261 |
| 18 vs. 27 pulses/call | 50% | 4.50 | 0.194 vs. 0.294 |
| Series 2: short-call (≈ 1.80 SD below population meana) | |||
| 10 vs. 11 pulses/call | 10% | 2.38 | 0.200 vs. 0.221 |
| 10 vs. 12 pulses/call | 20% | 4.50 | 0.111 vs. 0.120b |
| 10 vs. 13 pulses/call | 30% | 2.38 | 0.200 vs. 0.263 |
| 10 vs. 15 pulses/call | 50% | 2.38 | 0.200 vs. 0.315b |
| Additional test | |||
| 8 vs. 12 pulses/call | 50% | 4.50 | 0.083 vs. 0.120b |
| Alternatives | Percentage difference | Call period(s) | Call duty cycles |
|---|---|---|---|
| The stimuli of Series 1 and 2 were used in both the equal and unequal playback designs. The stimulus listed under "Additional test" was tested in the unequal playback design. | |||
| aEstimates from call data collected during 1987 (n = 168 males) (Gerhardt et al., 1996). | |||
| bTests in which females were tested twice in the unequal-playback design. | |||
| Series 1: short-call (≈ mean = 17.8 pulses/calla) | |||
| 18 vs. 20 pulses/call | 11.1% | 4.50 | 0.194 vs. 0.217b |
| 18 vs. 22 pulses/call | 22.2% | 4.50 | 0.194 vs. 0.239 |
| 18 vs. 24 pulses/call | 33.3% | 4.50 | 0.194 vs. 0.261 |
| 18 vs. 27 pulses/call | 50% | 4.50 | 0.194 vs. 0.294 |
| Series 2: short-call (≈ 1.80 SD below population meana) | |||
| 10 vs. 11 pulses/call | 10% | 2.38 | 0.200 vs. 0.221 |
| 10 vs. 12 pulses/call | 20% | 4.50 | 0.111 vs. 0.120b |
| 10 vs. 13 pulses/call | 30% | 2.38 | 0.200 vs. 0.263 |
| 10 vs. 15 pulses/call | 50% | 2.38 | 0.200 vs. 0.315b |
| Additional test | |||
| 8 vs. 12 pulses/call | 50% | 4.50 | 0.083 vs. 0.120b |
Table 1
Acoustic stimuli used in choice-tests of females of Hyla versicolor
| Alternatives | Percentage difference | Call period(s) | Call duty cycles |
|---|---|---|---|
| The stimuli of Series 1 and 2 were used in both the equal and unequal playback designs. The stimulus listed under "Additional test" was tested in the unequal playback design. | |||
| aEstimates from call data collected during 1987 (n = 168 males) (Gerhardt et al., 1996). | |||
| bTests in which females were tested twice in the unequal-playback design. | |||
| Series 1: short-call (≈ mean = 17.8 pulses/calla) | |||
| 18 vs. 20 pulses/call | 11.1% | 4.50 | 0.194 vs. 0.217b |
| 18 vs. 22 pulses/call | 22.2% | 4.50 | 0.194 vs. 0.239 |
| 18 vs. 24 pulses/call | 33.3% | 4.50 | 0.194 vs. 0.261 |
| 18 vs. 27 pulses/call | 50% | 4.50 | 0.194 vs. 0.294 |
| Series 2: short-call (≈ 1.80 SD below population meana) | |||
| 10 vs. 11 pulses/call | 10% | 2.38 | 0.200 vs. 0.221 |
| 10 vs. 12 pulses/call | 20% | 4.50 | 0.111 vs. 0.120b |
| 10 vs. 13 pulses/call | 30% | 2.38 | 0.200 vs. 0.263 |
| 10 vs. 15 pulses/call | 50% | 2.38 | 0.200 vs. 0.315b |
| Additional test | |||
| 8 vs. 12 pulses/call | 50% | 4.50 | 0.083 vs. 0.120b |
| Alternatives | Percentage difference | Call period(s) | Call duty cycles |
|---|---|---|---|
| The stimuli of Series 1 and 2 were used in both the equal and unequal playback designs. The stimulus listed under "Additional test" was tested in the unequal playback design. | |||
| aEstimates from call data collected during 1987 (n = 168 males) (Gerhardt et al., 1996). | |||
| bTests in which females were tested twice in the unequal-playback design. | |||
| Series 1: short-call (≈ mean = 17.8 pulses/calla) | |||
| 18 vs. 20 pulses/call | 11.1% | 4.50 | 0.194 vs. 0.217b |
| 18 vs. 22 pulses/call | 22.2% | 4.50 | 0.194 vs. 0.239 |
| 18 vs. 24 pulses/call | 33.3% | 4.50 | 0.194 vs. 0.261 |
| 18 vs. 27 pulses/call | 50% | 4.50 | 0.194 vs. 0.294 |
| Series 2: short-call (≈ 1.80 SD below population meana) | |||
| 10 vs. 11 pulses/call | 10% | 2.38 | 0.200 vs. 0.221 |
| 10 vs. 12 pulses/call | 20% | 4.50 | 0.111 vs. 0.120b |
| 10 vs. 13 pulses/call | 30% | 2.38 | 0.200 vs. 0.263 |
| 10 vs. 15 pulses/call | 50% | 2.38 | 0.200 vs. 0.315b |
| Additional test | |||
| 8 vs. 12 pulses/call | 50% | 4.50 | 0.083 vs. 0.120b |
General methods and test procedures
Females of H. versicolor of the northwestern mitochondrial DNA (cytochrome b) lineage of Ptacek et al. (1994) were collected in small ponds in Boone County (Baskett Wildlife and Three Creeks Conservation areas), Missouri during the 1994-1999 breeding seasons. Most females were released at the same ponds after testing in behavioral experiments, and all treatment of these animals conforms to a protocol approved by the Animal Care and Use Committee of the University of Missouri, Columbia.
Preference strength was determined by testing females in a temperature-regulated (20°C ± 1.5°C), semi-anechoic chamber. We placed each female in an acoustically transparent, hardware-cloth cage located midway between two speakers. Details about the playback system and chamber acoustics are provided in Gerhardt (1994) and Gerhardt et al. (1996). For every test and every trial (see below), we remotely removed the top of the cage, allowing the female to move in any direction; the female was released after playing back the alternative stimuli three times each. Our criteria for a response were phonotactic orientation (head scanning, zigzag hopping, or crawling correlated with the playback of a signal; Rheinlaender et al., 1979) and movement to within 10 cm or 50 cm (see below) of one of the speakers. Even though control tests indicate that no side biases exist in the chamber (see below), stimuli were switched periodically between speakers between tests of different animals or tests of different pairs of alternatives.
Acoustic stimuli
Details about the synthesis of acoustic signals are provided in Gerhardt et al. (1996). We varied the duration of synthetic calls by altering the number of pulses/call, as do male tree frogs. The smallest biologically meaningful step-size thus equals one pulse duration and period. Because pulse rate (20 pulses/s = pulse period of 50 ms) and pulse duration (25 ms) were the same in all stimuli, call duration in ms can thus be easily computed by multiplying the number of pulses per call by 50 and subtracting 25 ms. The pulse rate was chosen to match the mean value in the study population at 20°C, and the spectral properties of all synthetic calls had population-typical values (carrier frequencies of 2.2 kHz [0 dB] and 1.1 kHz [-6 dB]).
In both experimental designs (next section), we pitted a short call against an alternative stimulus that had exactly or approximately 10, 20, 30, or 50% more pulses per call (see Table 1). In one series of tests, the short-call alternative had 18 pulses/call (duration = 875 ms), which was close to the grand mean (17.8 pulses/call) in the calls of 168 males from the same population recorded in the 1987 season (Gerhardt et al., 1996). In the second series of tests, the short alternative had 10 pulses/call (duration = 475 ms), which is about 1.80 standard deviations below the mean (SD = 4.3 pulses/call). In one additional test, used only in the design using unequal playback levels, the short-call alternative had 8 pulses/call, which is about 2.28 standard deviations below the mean, and the long-call alternative had 50% more pulses (12 pulses/call). The timing relationship between the two alternatives in all tests was fixed so that there were equal periods of silence between the end of one alternative and the beginning of the next alternative.
The call period of the short-call alternatives in every test except one in the two test series was chosen so that the call duty cycle (ratio of call duration to call period) was close to the grand mean (0.190, range of means, 0.178-0.210) of the calls of males recorded over four seasons (Gerhardt et al., 1996). Call duty cycle increased as the call duration increased, but all values fell within the range of natural variation. In one test within the second series and in the one additional test (see Table 1), the call duty cycle of the shorter call was well below the grand mean, but still within the range of natural variation (0.080-0.340). Even though we have no reason to suspect that the difference in duty cycle would affect preference strength, the data were not used in any overall statistical comparisons with data from the other tests in which the call duty cycle of the short-call alternatives was about 0.20. These data nevertheless provided additional estimates of preference strength and repeatability for duration differences of 20% and 50%.
Experimental designs
Population-level estimates of preference strength: equal playback level
Females were offered choices of two alternatives whose playback level was equalized at 85 dB SPL (re 20 ÎŒPa, "fast" root-mean-square) at the female release point. When a female moved to within 10 cm of a speaker, we tallied a response (one response per female for any given two-stimulus choice); the minimum number of females tested was 16. Preference strength is estimated as the proportion of females choosing the longer alternative. If a female was tested in more than one two-stimulus test, a time-out period of at least 2 min elapsed between such tests. During this time, the female was removed from the chamber and the SPLs of the new signals were equalized before she was replaced at the release point. The time-out period was usually much longer (30-45 min) because we often tested a series of females in one test before switching to a new test. Females not being tested were always maintained in an incubator outside of the test chamber while another female was being tested and during adjustments of playback levels.
Estimates of preference strength in decibels (dB): unequal playback levels
We estimated preference strength by determining how much the SPL (in dB) of a long alternative could be reduced relative to the short alternative (always adjusted to 85 dB SPL) without abolishing the preference for the longer call. For each test condition, each female was tested a maximum of four times, and we concluded that she preferred the long call if she moved to within 50 cm of its source in at least three trials. After each trial, the female was replaced in the central release cage, the investigator left the chamber, alternatives were played back three times each, and the cage top was remotely removed as in the first trial. This procedure insured a time-out period of 1 min or more, depending on how many females were being tested on a given day. At the end of a series of trials at one SPL-difference, the female was removed from the chamber, and a new choice condition (adjustment of the SPL of the long call) was set before the next series of trials. If a female met our criterion for a preference, then the SPL of the longer call was decreased in 3-dB steps until she did not meet our criterion. Preference strength was then defined as the largest difference in SPL favoring the short alternative at which a female did meet our criterion. If a female did not prefer the longer call when its SPL was 3 dB less than that of the short call or when the alternatives had the same SPL, her preference strength was arbitrarily assigned a score of 0 dB.
Because females have a limited period of responsiveness (they stop responding when they oviposit), the relaxed response criterion (approach to within 50 cm rather than 10 cm) for each trial increased our chances of obtaining one or more complete estimates of preference strength for each animal. We justify this criterion in two ways. First, because we were estimating preference strength on the basis of a predetermined difference in SPL, females experienced such a difference only at or near the release point. Second, based on extensive experience with this species, we estimate that fewer than 1% of the females that move to within 50 cm of one speaker reverse their course and move to the other speaker. For example, in tests not reported in this paper, in only four of 1386 responses (362 females tested in the 1997-1998 seasons) did a female reverse course after moving to within 50 cm of one speaker and then move to within 50 cm the other speaker.
For tests of signals that differed by 10%, the SPL of both calls was 85 dB SPL (0 dB difference) in the initial set of trials. For most tests of signals whose duration differed by a greater percentage, the SPL of the longer alternative was initially adjusted to be 3 dB lower (82 dB SPL) because females nearly always chose the longer call when SPL was equalized. This procedure reduced the number of trials per female needed to obtain an estimate of preference strength, and again made it more likely that we would obtain one or more complete estimates before the end of the female's period of phonotactic responsiveness. Most females were used in different tests (different stimulus pairs; see Results). We varied the order of such tests, but obtained one estimate of preference strength for each test before changing to the next test.
Control experiment (unequal playback-level design): variability in female preference for differences in SPL
Our measure of preference strength in the unequal playback-level design is potentially confounded by variation among females in preferences for call intensity. We therefore conducted a control experiment in which females had a choice between two 18-pulse calls whose playback level differed by 3 dB (85 dB versus 82 dB SPL). If substantial numbers of females were to choose the less intense call, then we would be concerned that variation in female preferences based on intensity alone could confound our estimates of preference strength based on differences in call duration. Because all 12 females chose the more intense of two signals in three of a maximum of four trials (p <.001, two-tailed binomial test), we conclude there is little or no variation among females in preferences based on a small difference in SPL alone.
Control experiments (unequal playback-level design): carry over effects in repeated trials
Researchers routinely test individual frogs in several two-choice preference tests with an arbitrary time-out between tests (Gerhardt, 1991), and studies of other species of tree frogs did not detect carry over effects (Gerhardt, 1981; Murphy and Gerhardt, 1996). Because estimating preference strength in the unequal playback-level design required multiple trials with the same alternatives at the same playback levels, we conducted two control experiments to evaluate the possibility that a female gray tree frog's acoustic experience or choice in her initial trial might bias her to move toward the same speaker in a subsequent trial or to avoid it. In the first control experiment, conducted in 1996, a female first heard seven repetitions each of a short (12 pulses/call) and a much longer call (24 pulses/call) while she was still confined in the release cage. We then gave the female a choice between two calls of 12 pulses/call. Fourteen of 26 females (53.8%) responded to the speaker that had previously played back the long call; the others responded to the speaker that had previously played back the short call. This experiment demonstrates that a female's choice was not biased either towards or away from a speaker that previously played back a more attractive (long) call.
In the second control experiment, conducted in 1999, the female was allowed to respond to an attractive long call (27 pulses/call), and, when the female moved to within 50 cm of the speaker, we replaced her in the hardware cloth cage at the release point. After leaving the chamber, we then gave the female a choice of two calls of 18 pulses/call, which were played back three times each before the female was remotely released. Nine of 20 females (45.0%) responded to the same speaker to which they had initially moved in response to the call of 27 pulses/call; the others responded to the other speaker. Furthermore, 10 of these 20 females chose the stimulus played back from the south side of the chamber, and the others chose the stimulus played back from the opposite (north) side, thus indicating that side biases in the playback system or testing chamber were unlikely to have been a factor in any tests.
Estimates of the consistency and repeatability of preference strength
For four tests (see Table 1), we obtained two estimates of preference strength per female using the unequal playback-level design. The time-out period between the two testing periods of the same animal with the same two alternative stimuli was at least 30 min. These data were used for two purposes. First, they provide information about the consistency of preference strength and the reliability of our methodology. If, for example, estimates of preference strength were influenced by repeated testing, handling, the variable times between repeated testing, fatigue, or changes in motivation as oviposition became imminent, then we might expect to see many large differences in repeated estimates for the same animals, a skew in the direction of change in estimates, or both. Second, because a reasonable spread of preference strength was observed in three of the tests, we computed repeatability, which quantifies between-female variability.
Statistical procedures
For parametric estimates and statistical tests (Statview or Statistica), values of preference strength in decibels were first converted to ranks (0 dB = 0, 3 dB = 1, 6 dB = 2, etc.) because decibels represent a nearly linear perceptual scale (Yost and Nielsen, 1985). Conover et al. (1981) provide the statistical justification for this procedure. Seaman et al. (1994) show that the 2×2 ANOVA we used is not affected adversely by interaction as are other designs using ranked data. Even though many females responded in more than one two-stimulus test, all statistical comparisons of the results of different two-stimulus tests used strictly independent data. That is, each female contributed only one response (equal playback-level design) or one estimate of preference strength (unequal play-back-level design) to the data set being analyzed.
RESULTS
We obtained at least one estimate of preference strength in the unequal playback-level design from 153 females during the 1996-1999 breeding seasons; 69 of these provided two or more estimates. Many of the same females, as well as 74 additional animals tested in 1994-1995, responded in at least one test of alternatives in the equal playback-level design, which provided population-level estimates of preference strength based on the proportions of females (one response per female per test) choosing the long-call alternative.
Estimates of the duration preference function: proportions of females choosing longer calls (equal playback-level design)
In Figure 1, we show the proportions (and lower 95% confidence limits) of females that chose the longer of two alternatives presented at the same playback level. In each test the majority of females chose the longer alternative: the results were all statistically significant (p-values <.05, one-tailed binomial test, with the a priori expectation that females would prefer the longer alternative). More importantly, however, preference strength (here, the probability that a female would choose the longer alternative) increased as the difference in duration was increased when the short alternative was much lower than average (10 pulses/call) but not when the short alternative was about average in duration (18 pulses/call). Although the lack of independent data (some females were tested in more than one of these tests) precludes a formal analysis of the overall differences in proportions, we did have independent samples of females for some important comparisons. When the short-call alternative had 10 pulses/call, the proportion of females (92%) preferring the longer alternative when the difference in duration was 50% was significantly higher than the proportion (67%) choosing the longer alternative when the difference in duration was 10% (G = 3.90, df = 1, p =.048). In the comparable contrast, using the short-call alternative of 18 pulses/call, the proportions of females choosing the longer alternatives of 11.1% and 50% were nearly the same (78% versus 77%, respectively). In tests of a 50% difference in duration, a higher proportion of females chose the longer call when the short-call alternative had 10 pulses/call than when the short-call alternative had 18 pulses/call (92% versus 77%, respectively), although this difference is not statistically significant (G = 1.93, df = 1, p =.164).
Figure 1
Preference functions expressed in terms of the proportions of females choosing (one response per female) the longer of two alternatives. The error bars show the lower 95% confidence limits, and the samples sizes (numbers of females) are given on each bar. Many of the females were tested in more than one of these tests, but statistical comparisons (shown above the histogram) were restricted to completely independent samples, that is, no female was tested in both of the tests being compared.
Estimates of the duration preference function: preference strength in decibels (unequal playback-level design)
In Figure 2, we summarize estimates of preference strength (in decibels) as a function of differences in absolute and relative call duration. As in the preference functions based on proportions, the function is steeper for tests of short-duration calls than that for tests of long-duration calls. Again, lack of independent data prevents a formal statistical analysis of the entire data set. Independent estimates (minimum of 25 frogs each) were, however, available for each of the four tests of duration differences of 10% (or 11.1%) and 50%. We subjected these data to a 2×2 (duration of short call × % difference in duration), between-subjects ANOVA. This analysis detected significant differences in preference strength for the main effects (F = 9.54, df = 1, 124, p =.002, values the same for both effects) and for the interaction between the main effects (F = 5.46, df = 1, 124, p =.021). Subsequent contrast analysis showed, in parallel to the results based on proportions of choices, that the preference was stronger by about 3 dB for the 50%-longer call than for the 10%-longer call when using the short-call alternative of 10 pulses/call (F = 13.46, df = 1, 124, p =.0036), but no such difference was found when using the short-call alternative of 18 pulses/call (F = 0.31, df = 1, 124, p =.578). Moreover, the trend that was non-significant in the equal playback-level design was highly significant: preference strength for 50%-longer calls was significantly higher when the short-call alternative had 10 pulses/call (5 dB) than when the short-call alternative had 18 pulses/call (2 dB) (F= 13.46, df = 1, 124, p =.0003). No such difference occurred when the alternatives differed by about 10% (F = 0.31, df = 1, 124, p =.578).
Figure 2
Preference functions expressed in terms of the mean preference strength (SPL differing favoring the short-call at which females continued to choose the long call) as a function of the percentage difference in the number of pulses/call (≅ duration). Triangles: short-call alternative had 18 pulses/call; squares: short-call alternative had 10 pulses/call; circles: test of 8 versus 12 pulses/call. Numbers of females are indicated next to each point. Error bars are ± two SE. Decibel values were first converted to a linear measures (ÎŒPa) from which means and standard errors were computed before reconversion to decibels, hence the asymmetry of the error bars.
The following qualitative results serve to emphasize the striking contrast between preference strength in tests using different short-call alternatives. When the short-call alternative had 10 pulses/call, nine of the 30 females (30%) chose a 50%-longer alternative (15 pulses/call) when the SPL of the longer call was 6 dB less than that of the shorter call; three females did so when the SPL of the longer call was 9 dB less. By contrast, when the short-call alternative had 18 pulses/call, only three of 31 females (9.6%) chose the 50%-longer alternative (27 pulses/call) when the SPL of the longer call was 6 dB less; none did so when the difference favoring the short-call alternative was 9 dB.
The mean preference strength at a duration difference of 50% for one additional test is also plotted in Figure 2. Here females were given a choice of calls with 8 versus 12 pulses/call. The mean preference strength was greater than 7 dB, and, remarkably, 14 of the 41 females (34%) chose the longer alternative even when its SPL was as much as 9 dB to 18 dB less than that of the shorter call.
Consistency and repeatability of estimates of preference strength
In Figure 3, we show the frequency distribution of the difference (in decibels) between two estimates of preference strength (n = 131 comparisons in four tests). About 50% of the second estimates were the same as the first estimate (0 dB difference) and only about 6% of the estimates differed by as much as 6 dB. There was also no significant skew in the direction of change (p >.1, two-tailed binomial).
Figure 3
Frequency distribution of changes in estimates of preference strength in female gray tree frogs. For most females, the second estimate was the same as the first estimate (0 dB difference). There was no significant tendency for second preference estimates to be higher than first estimates (minus values) than the reverse (positive values).
A formal analysis of repeatability was warranted for three tests because the initial range of preference estimates was reasonably large: low repeatability results not only from inconsistent measures or variability within individuals but also from a lack of variability between individuals (Falconer, 1989). Repeatability was not computed for the repeated estimates of preference strength in the test of 18 versus 20 pulses/s because only two of the 72 estimates were as great as 6 dB; all other estimates were either 0 or 3 dB.
The range of preference strength scores in the test of 8 versus 12 pulses/call was particularly large (0-18 dB), thus suggesting that considerable variation exists in preference strength. The two estimates of preference strength for each of 33 females tested with these alternatives are plotted in Figure 4. A formal analysis of repeatability (defined as the intra-class correlation coefficient = ri using ranked data, as above) yielded a value of 0.69 (F between-subject = 5.37, df = 32, p <.0001). We also found significant repeatability for two other tests for which we had sufficient independent data (20% difference, 10 versus 12 pulses/call; 50% difference, 10 versus 15 pulses/call) even though the range of preference-strength estimates was only 9 dB. The estimates were r = 0.50 (F between-subject = 2.97, df = 31, p =.0017) and r = 0.36 (F between-subject = 2.11, df = 32, p =.018), respectively. Significant repeatability of preference strength not only indicates consistency within females but also variation among females.
Figure 4
Plots of the first and second estimates of preference strength. The data are based on two estimates obtained from 33 females in the test of alternatives of 8 versus 12 pulses/call. Largest circles = 5 females; next-largest = 4 females; next largest = 3 females; next largest = 2 females; smallest = 1 female.
DISCUSSION
Shape of preference functions: non-linear effects
The most important result of our experiments is that both playback designs show a non-linear effect of increasing call duration. That is, when the short-call alternative was 10 pulses/call, preference strength increased as the difference in duration increased from 10 to 50%, but the same pattern did not occur when the short-call alternative was 18 pulses/call. Furthermore, preference strength for a test of a 50%-difference in duration was greater (significantly so in the unequal playback-level experiments) when the short-call alternative had 10 pulses/call than when the short-call alternative had 18 pulses/call. These results corroborate and extend the results of an earlier study (Gerhardt et al., 1996). In those experiments, females showed a much greater tendency to return to an interrupted source of long calls when the short-call alternative was well below the mean (12 pulses/call) than when the short alternative was close to the mean (18 pulses/call) (Gerhardt et al., 1996). Seventy-five percent of the females returned to the source of long calls in the first situation, and only about 9% and 25% did so in the second situation, when long-call alternatives had 24 and 30 pulses/call, respectively. Taken together, then, females more strongly prefer calls of average or above-average duration to very short alternatives than they favor very long alternatives to calls of about average duration. Further experiments are needed to determine the minimum duration at which increasing call duration results in an increase in preference strength.
Our results using pairs of calls with pulse numbers below the mean are similar to those of a study of preferences based on differences in sword length in the green swordtail fish (Basolo, 1990). Preference strength, which was estimated as the difference in time spent with each of two males differing in sword length, increased with the increasing percentage difference in sword length but appeared to asymptote as the difference in length became large. These results indicate that both in gray tree frogs and swordfish preference functions are not open-ended, but rather that increasing the value of a property well beyond the mean does not elicit concomitant increases in preference strength.
In terms of perceptual mechanisms, our results indicate that females do not simply evaluate the ratio of call duration in the two signals as predicted by Weber's law (Yost and Nielsen, 1985). In both of our designs, a 50% difference in duration led to different estimates of preference strength depending on the absolute durations of the pair of alternatives. Thus, female preferences based on duration in gray tree frogs depend on both the absolute and relative values of alternative stimuli.
Implications for sexual selection
Our results have significant implications for expected patterns of selection on male calls and, potentially, for selection on female preference functions via possible differences in the net benefits obtained by females from choosing males on the basis of call duration.
Male mating success
The pattern of preference we observed suggests that the upper bound on call duration is not only energetically constrained (Wells and Taigen, 1985), but also that increasing call duration beyond a certain point probably has diminishing effects on mating success. This non-linear effect was especially clear in the unequal playback-level design, which simulates costly choice. Because SPL is inversely related to distance, choosing a call of lower SPL over a call of higher SPL is equivalent to choosing a male that is further away. In our study population, moving extra distances possibly increases the risk of predation by sit-and-wait predators such as bullfrogs, green frogs, and snakes (Gerhardt et al., 1996). This pattern of preference effectively weakens selection on call duration and may contribute to the maintenance of its variation in natural populations.
Our results have a practical implication for field studies of male mating success based on variation in call duration. One such study of H. versicolor in Maine did not find a significant relationship between call duration and mating success (Sullivan and Hinshaw, 1992). However, the sample size was small, and hence the study lacked statistical power to detect small effects. Besides increasing sample size, future studies might be more likely to detect female choice on call duration if a special effort is made to compare the mating success of males producing very short calls with that of all other males.
Net benefits for females?
In general, we might expect to find non-linear female preference functions if females receive benefits as a result of some choices but not others, especially if choice is costly. For the gray tree frog, the shape of the non-linear preference functions suggest that the benefits females might receive from choosing males producing calls of at least average duration compensate for the cost of moving extra distances rather than mating with a nearby male producing very short calls. Given the choice of a male producing calls of average duration, however, the extra benefits of moving to a male producing very long calls might not be worthwhile. These ideas could be tested by assessing benefits in the form of increased fertilization success (Pfennig, 2000) or larval fitness (Welch et al., 1998).
Consistency and repeatability of preference strength
Estimates of preference strength were highly consistent within females, despite the multiple testing and handling between trials and tests at different SPL-differences (see Figure 3). Although the magnitude of the repeatability estimates of preference strength could be considered high to very high for behavioral traits (Arnold, 1986), the experimental design was unrealistic in terms of the time frame over which a female makes successive mate-choice decisions (Arnold, 1994). Trials that generated a second estimate of preference strength began as soon as 30 min after the end of the trials that generated the first estimate, whereas a female tree frog chooses one male to fertilize her entire clutch, a process that normally precludes a second mating on the same night. Some females return to mate again at intervals of a few weeks (Gerhardt et al., unpublished data), which would be the appropriate time interval for making successive estimates of preference strength.
Nevertheless, our repeatability estimates show that variation among females in preference strength can be demonstrated in the particular experimental design we used. We think it is reasonable to suggest that individual differences in preference strength might persist over time, and this hypothesis is, in principle, testable, even though extraordinary sampling efforts would be necessary. Other studies that have tested females multiple times in laboratory situations over relatively short time intervals have had mixed results (review; Jennions and Petrie, 1997). Despite their limitations, studies that reveal significant levels of repeatability encourage breeding or artificial selection experiments, which could directly estimate additive genetic variance in preference strength.
With respect to anuran studies, our present study shows that the usual method of testing preferences between signals of equal intensity in a binary fashion is less sensitive in detecting individual variation than procedures that might simulate costly choice. For example, in tests of the 12-pulse call over the 8-pulse call, preference strength was highly variable and repeatable. Yet females were nearly unanimous in choosing the longer of these two calls multiple times when these two alternatives were played at the same or nearly the same SPL. Thus, if preference strength were to be measured in terms of the proportions of repeated choices of the longer alternative in this particular test, repeatability would be close to zero. Wagner (1998) makes similar arguments in his appeal for more continuous measures of female preference, such as the speed and distance traveled toward a sound source, which were used to demonstrate repeatability of preference in field crickets (Wagner et al., 1994).
H. C. Walton is now at the Department of Evolution, Ecology and Organismal Biology, Ohio State University, Columbus, OH 43210.
We thank K. Blackwell, S. Bush, S. Conditt, J. Schul, V. Marshall, J. Schwartz, and A. Welch for collecting frogs, helping to test them, or both. C. Murphy, M. Ritchie, S. Bush, W. Wagner, Jr., M. Bee, M. Keller, and especially A. Welch provided helpful comments on the manuscript, help with the statistical analysis, or both. Funding was provided by National Science Foundation and National Institutes of Mental Health (Research Scientist Award) grants to H.C.G.
References
Andersson MB,
1994
.
Sexual selection
. Princeton, New Jersey: Princeton University Press.
Arnold SJ,
1986
. Laboratory and field approaches to the study of adaptation. In:
Predatory-prey relationships: perspectives and approaches from the study of lower vertebrates
(Feder ME, Lauder GV, eds). Chicago: University of Chicago Press;
157
-179.
Arnold SJ,
1994
. Multivariate inheritance and evolution: a review of concepts. In:
Quantitative genetic studies of behavioral evolution
(Boake CRB, ed.). Chicago: University of Chicago Press;
17
-48.
Bakker TCM, Pomiankowski A,
1995
. The genetic basis for female mate preferences.
J Evol Biol
8
:
129
-171.
Basolo AL,
1990
. Female preference for male sword length in the green swordtail, Xiphophorus helleri (Pisces: Poeciliidae).
Anim Behav
40
:
332
-338.
Basolo AL,
1998
. Evolutionary change in a receiver bias: a comparison of female preference functions.
Proc. R. Soc. Lond. B
265
:
1
-6.
Butlin RK,
1993
. The variability of mating signals and preferences in the Brown planthopper, Nilaparavata lugens (Homoptera: Delphacidae).
J Insect Behav
6
:
125
-140.
Conover WJ, Iman JL,
1981
. Rank transformation as a bridge between parametric and nonparametric statistics.
Am Statistician
35
:
124
-129.
Falconer DS,
1989
.
Introduction to quantitative genetics
, 3rd Ed. New York: John Wiley and Sons.
Gerhardt HC,
1981
. Mating call recognition in the green tree frog (Hyla cinerea) : importance of two frequency bands as a function of sound pressure level.
J Comp Physiol
144
:
9
-16.
Gerhardt HC,
1991
. Female mate choice in tree frogs: static and dynamic acoustic criteria.
Anim Behav
42
:
615
-635.
Gerhardt HC,
1994
. Reproductive character displacement of female mate choice in the gray tree frog Hyla chrysoscelis.
Anim Behav
47
:
959
-969.
Gerhardt HC, Doherty JA,
1988
. Acoustic communication in the gray tree frog, Hyla versicolor. evolutionary and neurobiological implications.
J Comp Physiol A
162
:
261
-278.
Gerhardt HC, Dyson ML, Tanner SD,
1996
. Dynamic acoustic properties of the advertisement calls of gray tree frogs: patterns of variability and female choice.
Behav Ecol
7
:
7
-18.
Hedrick AV, Weber T,
1998
. Variation in female responses to the fine structure of male song in the field cricket, Gryllus integer.
Behav Ecol
9
:
582
-591.
Jennions MD, Backwell PRY, Passmore NI,
1995
. Repeatability of mate choice: The effect of size in the African painted reed frog, Hyperolius marmoratus.
Anim Behav
49
:
181
-186.
Jennions MD, Petrie M,
1997
. Variation in mate choice and mating preferences: a review of causes and consequences.
Biol Rev
72
:
283
-327.
Kime NM, Rand AS, Kapfer M, Ryan MJ,
1998
. Consistency of female choice in the tĂșngara frog: a permissive preference for complex characters.
Anim Behav
55
:
641
-649.
Klump GM, Gerhardt HC,
1987
. Use of non-arbitrary acoustic criteria in mate choice by female gray tree frogs.
Nature
326
:
286
-288.
Murphy CG, Gerhardt HC,
1996
. Evaluating experimental designs for determining mate choice: the effect of amplexus on mate choice by barking tree frogs.
Anim. Behav
51
:
881
-890.
Murphy CG, Gerhardt HC,
2000
. Mating preference functions of individual female barking tree frogs, Hyla gratiosa for two properties of male advertisement calls.
Evolution
54
:
660
-669.
Pfennig, KS,
2000
. Female spadefoot toads compromise on mate quality to ensure conspecific matings.
Behav. Ecol
11
:
220
-227.
Pomiankowski A,
1988
. The evolution of female mating preferences for male genetic quality.
Oxford Surv Evol Biol
5
:
136
-184.
Ptacek MB, Gerhardt HC, Sage RD,
1994
. Speciation by polyploidy in tree frogs: multiple origins of the tetraploid, Hyla versicolor.
Evolution
48
:
898
-908.
Ritchie MG,
1996
. The shape of female mating preferences.
Proc Natl Acad Sci USA
93
:
14628
-14631.
Runkle LS, Wells, KD, Robb CC, Lance SL,
1994
. Individual, nightly and seasonal variation in calling behavior of the gray tree frog, Hyla versicolor. implications for energy expenditure.
Behav Ecol
5
:
318
-325.
Seaman JW Jr, Walls SG, Wise SE, Jaeger RG,
1994
. Caveat emptor: rank transform methods and interaction.
Trends Ecol Evol
9
:
261
-263.
Shaw KL,
2000
. Interspecific genetics of mate recognition: inheritance of female acoustic preference in Hawaiian crickets.
Evolution
54
:
1303
-1312.
Shaw KL, Herlihy D,
2000
. Acoustic preference functions and song variability in the Hawaiian cricket Laupala cerasina.
Proc R Soc Lond B
267
:
577
-584.
Sullivan BK, Hinshaw SH,
1992
. Variation in advertisement calls and male mating behaviour in the grey tree frog Hyla versicolor.
Anim Behav
44
:
733
-744.
Wagner WE Jr,
1998
. Measuring female mating preferences.
Anim Behav
55
:
1029
-1042.
Wagner WE Jr, Murray A-M, Cade WH,
1994
. Phenotypic variation in the mating preferences of female field crickets.
Anim Behav
49
:
27
-45.
Welch AM, Semlitsch RD, Gerhardt HC,
1998
. Call duration as a reliable indicator of genetic quality in the gray tree frog.
Science
280
:
1928
-1930.
Wells KD, Taigen TL,
1986
. The effect of social interactions on calling energetics in the gray tree frog (Hyla versicolor).
Behav Ecol Sociobiol
19
:
9
-18.
Yost WA, Nielsen DW,
1985
.
Fundamentals of hearing: an introduction
, 2nd Ed. New York: Holt, Rinehart and Winston.
International Society of Behavioral Ecology
International Society of Behavioral Ecology
Source: https://academic.oup.com/beheco/article/11/6/663/221632
0 Response to "Can Frogs Have Sex Continuously for Up to 48 Hours"
Enregistrer un commentaire