- Email: [email protected]
Evaluating the design of mate-choice experiments: the effect of amplexus on mate choice by female barking treefrogs,Hyla gratiosa
Anim. Behav., 1996, 51, 881–890
Evaluating the design of mate-choice experiments: the effect of amplexus on mate choice by female barking treefrogs, Hyla gratiosa CHRISTOPHER G. MURPHY & H. CARL GERHARDT Division of Biological Sciences, University of Missouri, U.S.A. (Received 21 November 1994; initial acceptance 19 March 1995; final acceptance 11 August 1995; MS. number 7163)
Abstract. The proper design of mate-choice experiments is critical to detecting and characterizing mating preferences. Female anuran amphibians tested in mate-choice experiments are almost always collected in amplexus (i.e. after they have chosen a mate); it has been assumed that these females are as discriminating as they are before they enter amplexus. This assumption was tested by testing females of the barking treefrog, Hyla gratiosa, twice within the same night, once before they entered amplexus (‘before’ test) and once after (‘after’ test). Two such experiments were conducted in which females were given a choice between two calls that differed either in repetition rate or fundamental frequency. In both experiments, the proportion of females choosing each alternative did not differ between the before and after tests. Some females switched their choice between the before and after tests; in both experiments, equal numbers of females switched in one direction as in the other. The response of females in the after tests was not affected by either the time that they had spent in amplexus or the time of night when they were tested. Females that had not been tested earlier in the season took significantly less time to make their choice in the after test than in the before test, but females that had been previously tested did not. The results support the assumption that female anuran amphibians captured in amplexus are as discriminating as they are when first selecting mates in natural choruses. ?
The evolution of mate choice is one of the most exciting and controversial areas in evolutionary biology (Halliday 1983; Bradbury & Davies 1987; Pomiankowski 1988; Kirkpatrick & Ryan 1991; Andersson 1994). Two prerequisites for the study of the evolution of mate choice are demonstrating that females show choice and determining which male characters influence mate choice. Accomplishing these goals requires that mate-choice experiments be designed properly; improperly designed experiments could lead to incorrect conclusions about the existence of mate choice, the characteristics that influence mate choice or the degree of discrimination exhibited by females. Anuran amphibians have become model systems in which to examine mate choice (reviewed in Gerhardt 1988, 1994). Because females of Correspondence: C. G. Murphy, Department of Biology, James Madison University, Harrisonburg, VA 22807, U.S.A. (email: [email protected]). H. C. Gerhardt is at the Division of Biological Sciences, Tucker Hall, University of Missouri, Columbia, MO 65211, U.S.A. 0003–3472/96/040881+10 $18.00/0
?
1996 The Association for the Study of Animal Behaviour
many of these species base their choice of mates predominantly or exclusively on acoustic signals, cues in additional sensory modalities are unnecessary for eliciting natural mating behaviour from females. Techniques for recording, analysing and synthesizing such signals (e.g. Schwartz 1994) allow researchers to generate acoustic stimuli that are as attractive as natural calls (Gerhardt 1981) and to vary virtually any property of calls. Females behave toward speakers broadcasting advertisement calls as they would toward calling males: they make a series of movements toward a speaker, eventually contacting it. For most species, such behaviour is an unambiguous indication of a choice, because a female approaching a calling male in this manner in a natural chorus almost always mates with that male (unless she is intercepted by a satellite male). Many playback studies have demonstrated that female anuran amphibians do indeed choose mates and that numerous properties of calls influence mate choice in these species (reviewed in Gerhardt 1988, 1994). 1996 The Association for the Study of Animal Behaviour
881
882
Animal Behaviour, 51, 4
In almost all playback experiments conducted with anuran amphibians (cf. Martof 1961; Martof & Thompson 1964; Lopez & Narins 1991), researchers tested females that had been captured in amplexus and separated from their mates. There are two advantages in using females captured in amplexus for mate-choice experiments. First, such females are often much easier to find than females that have not yet chosen a mate. Second, females found in amplexus may be more likely to respond in playback experiments (93%: Gerhardt 1981; 88%: Arak 1988) than are females that have not yet chosen a mate (33–37%: Lopez & Narins 1991; cf. Martof 1961). Despite these advantages, there are reasons to expect that females tested after they have entered amplexus may be less discriminating than females that have not yet entered amplexus. In many playback experiments, females are tested several hours after they have entered amplexus; by this time, choruses of many species have ended for the night. Under natural conditions, a female losing her male at this time (e.g. due to predation of the male) will have few, widely separated males from which to choose. Females that choose to by-pass nearby males for more distant males will often have to traverse relatively large distances, thereby increasing their expenditure of energy and risk of predation. In addition, females may have a limited time in which to complete oviposition, and time spent discriminating between males may reduce the time available for oviposition. Finally, in some species (e.g. hylid treefrogs), females that have already entered amplexus will lay eggs even if they lose their mate and do not remate; in these species, ovulation probably begins at the time of amplexus or shortly thereafter, and once ovulation begins oviposition follows in a matter of hours. For these reasons, selection may favour females that, when separated from their mate, reduced their discrimination and mate quickly with nearby males. Thus, testing females captured in amplexus may underestimate the discrimination of females selecting mates under natural conditions. Although playback experiments with anuran amphibians have been conducted for over 30 years, no one has tested the assumption that females collected in amplexus behave the same as females that have not yet chosen a mate. In this study, we tested this assumption using a paired design in which we compared the choices made by individual female barking treefrogs, Hyla gratiosa,
before and after entering amplexus. To increase the generality of our results, we conducted two experiments. In one, we presented females with stimuli that differed in call repetition rate, a dynamic property that varies considerably during a bout of calling (Gerhardt 1991); in the other, we used stimuli that differed in frequency, a static property that varies little within a bout of calling (Gerhardt 1991).
METHODS Study Site and Species This study was conducted in the Apalachicola National Forest 4 km southwest of Tallahassee, Florida (30)22*N, 84)21*W). In this area, H. gratiosa breeds from March through August in semi-permanent ponds that lack fish. Shortly after sunset, males descend from the canopy of forests surrounding breeding ponds to begin calling in these ponds. They inflate themselves and float on the surface of the water, calling from the perimeter of ponds for 1–4 h. Females arrive at the pond shortly after males, approaching males and nudging them to initiate amplexus. Males do not defend oviposition sites, almost never behave as satellites, do not pursue females and do not attempt to dislodge alreadymated males. Pairs remain near the male’s call site for 1–2 h before disapppearing underwater to lay eggs. Male H. gratiosa produce advertisement calls consisting of a pulsatile onset followed by a harmonic series with two or three spectral peaks, the fundamental and one or two of the upper harmonics (Oldham & Gerhardt 1975; Gerhardt 1981). In the population monitored in this study, the average (&) fundamental frequency was 470 Hz&34 (range=410–550, N=30), the upper frequency peak was usually the fourth harmonic, the average call duration was 150&18.1 ms (range=112.8–118.8, N=30), and males produced an average of 54.9&5.1 calls per min (range= 42–66 calls/min, N=81). General Methods for Playback Experiments Each female was tested with the same playback experiment twice within the same night, once before and once after she entered amplexus. To
Murphy & Gerhardt: Effect of amplexus on mate choice obtain females before they entered amplexus, we used a drift fence to intercept females as they approached the pond (Murphy 1993). We transported the captured female to a playback arena located 0.45 km from the pond and tested her in a two-choice playback (‘before’ tests); we tested females as soon as possible after capture (about 15 min). Following the before test, we transported the female back to the pond and released her, allowing her to choose a mate. We collected the mated pair and transported it back to the playback arena, where we separated the female from her mate and retested her in the same playback used in the before test. Following this test (‘after’ test), we freeze-branded the female (Daugherty 1976) to prevent re-testing her in the same experiment (i.e. call-rate or fundamental frequency experiment). Freeze brands last for the duration of the breeding season (Murphy 1994). After the female was marked, we returned her and her mate to the pond so that the female could oviposit. Although no female was tested twice in the same before-and-after experiment, nine females were tested in the fundamental frequency experiment after having been tested an average (&) of 22.7&9.0 days (range=9–37 days) previously in other playback experiments (either the call-rate experiment or a set of playbacks for another study). We used two sites for playbacks. The first site was the shore of a dried-up pond; the substrate for this arena was sand, covered with short (7.5– 10 cm) vegetation. Eventually, the water level of the pond rose enough that other species of anurans (Gastrophryne carolinensis, H. femoralis) began calling there, producing background noise that made it difficult to adjust the sound pressure level (SPL) of the playback stimuli. Therefore, we began conducting playbacks at another site located on a nearby, unpaved road. This site also had a sandy substrate but lacked vegetation. We conducted all trials of the call-rate playback at the first site. We tested 9 of the 19 females in the fundamental frequency experiment at the first site and the remaining 10 at the second site. Within an experiment, each female was tested at the same site for both the before and after tests. At the beginning of a trial, we placed the female in an acoustically transparent, hardware–cloth cage 4.4 m from two speakers, which were separated by 3.5 m; the cage and the two speakers
883
formed a triangle. We used this arrangement, rather than releasing females equidistant between two speakers, because it more closely mimics the array of males that confront females when they choose mates in natural choruses (Murphy & Gerhardt, unpublished data). The distance between the two speakers (3.5 m) is very similar to the average inter-male distance in natural choruses (3.6 m; Murphy & Gerhardt, unpublished data). After beginning the playback, we released the female from the restraining cage by manually removing its lid. We observed the movements of females using a headlamp dimmed with a rheostat to an intensity at which the female was just visible. We scored a female as having chosen a speaker when she either passed within 10 cm of the speaker or circled it (in natural choruses, females sometimes make two or three close passes by a male before actually contacting him). We scored a ‘no response’ if the female did not leave the release cage after 3 min or if we lost sight of her when she hopped several metres outside of the triangle delineated by the release point and the speakers. We also recorded the time each female took to move to a speaker after the lid was removed from the restraining cage, as well as her path between the release point and the speaker. Stimuli used in playbacks were synthetic calls produced on an Amiga 500 computer (sampling rate of 20 kHz, 8-bits/sample) with custom software (J. Schwartz, unpublished data). All calls lasted 175 ms and had a pulsatile beginning followed by a harmonic series that consisted of the fundamental frequency plus the third and fourth harmonic, both of which were attenuated 6 dB relative to the fundamental; these calls are as attractive to females as are natural calls (Gerhardt 1981). Calls were recorded in stereo on to cassette tape with a Sony Pro Walkman recorder, played back from that tape recorder through a Sony XM-3021 stereo amplifier and broadcast from Realistic Minimus 7 speakers. This system was flat (&3 dB) over the frequency range encompassed by the experimental stimuli (0.4–2.0 kHz). The SPL (re 20 ìPa, ‘fast’ root-mean-square, Cweighting) of the two choices was equalized at 70 dB at the release point with a Realistic sound level meter; this intensity is similar to that measured at the same distance from calling males in natural choruses (70–72 db; Murphy, unpublished data).
Animal Behaviour, 51, 4
% choosing stimulus
884
100
P=0.004 N=15
P=0.002 N=13
P=0.034 N=43
80 60 40 20 0
45 versus 65 45 versus 55 55 versus 65 Stimulus (calls/min)
Figure 1. Percentage of females choosing low call-rate (.) and high call-rate (/) alternatives in two-choice playback experiments. Females used in these experiments were captured in amplexus. Significance values are two-tailed P-values from a binomial test of the null hypothesis of no preference (i.e. equal percentage of females choosing each alternative).
Call-rate Experiment In this experiment, we tested whether amplexus affected discrimination based on call repetition rate. To determine which call rates to use in the before-and-after tests, we conducted pilot experiments with females captured in amplexus. In the pilot experiments, we placed females midway between two speakers separated by 4 m and equalized the SPL of the two choices at 74 dB SPL at the release point. Females overwhelmingly preferred call rates either at the upper end of the natural distribution (65 calls/min) or at the population average (55 calls/min) over those near the lower end of the natural distribution (45 calls/min; Fig. 1); a lower percentage of females preferred the high call rate (65 calls/min) over the average call rate (55 calls/min; Fig. 1). We chose this latter stimulus set (high versus average call rates) to test the hypothesis that females are more discriminating before than after they enter amplexus, because it was the only set for which females captured before entering amplexus could be substantially more discriminating than they were in the pilot experiments. We recorded a synthetic call with a fundamental frequency of 400 Hz on to one channel of a cassette tape at 65 calls/min and on to the other channel at 55 calls/min. Because the two stimuli differed in repetition rate, the phase relationship of the calls from the two speakers changed during the playback.
We tested each female once before and once after she had entered amplexus. After placing the female in the restraining cage, we played back the stimuli for 30 s before releasing the female. To control for side biases, we switched the stimuli between speakers for each new female; for each female, however, we broadcast the same stimulus from the same speaker in both the before and after tests. Following the before test, we transported females back to the pond. We released 18 of the 20 females inside the fence and collected them and their mate 0.5–1.5 h later; we released the remaining two females inside the fence, observed them until they entered amplexus, and immediately collected them and their mates. We did not separate females from their mates until immediately before their after tests. We tested 20 females before and after amplexus; six additional females did not respond in the before test. We tested females in the before test within an average of (&) of 14.1&1.3 min of their capture in the fence (range=12–18 min) and in the after test within an average of 40.2& 12.7 min (range=16–66 min) of their capture following mating. Fundamental Frequency Experiment In this experiment, we tested whether amplexus affected discrimination based on fundamental frequency. Previous two-choice playback experiments conducted with females captured in amplexus showed that females generally prefer calls with intermediate fundamental frequencies over calls with either higher or lower fundamental frequencies (Gerhardt 1981; Murphy & Gerhardt, unpublished data); however, which frequency was preferred differed between females (Murphy & Gerhardt, unpublished data). Hence, we could not make a directional prediction about which fundamental frequency females would prefer in the before and after tests, as we did for the call-rate experiments. Instead, we presented females with two synthetic calls that differed in fundamental frequency (450 Hz versus 500 Hz); this difference represents 1.5 standard deviations of the distribution of fundamental frequencies in this population. We scored a female as ‘discriminating’ if she chose the same stimulus in each of four trials and as ‘indiscriminate’ if she did not; the probability of a female choosing the same stimulus in all four trials by chance alone (i.e. in the absence of
Murphy & Gerhardt: Effect of amplexus on mate choice preferences or carry-over effects) is 0.063. Our results are not affected if we use a less conservative measure of ‘discriminating’ (i.e. choice of the same stimulus in three or more of the four trials). We conducted four trials, rather than some greater number, to minimize the amount of time that elapsed between capture of the female and the beginning of her last trial. We chose alternatives of 450 Hz and 500 Hz because this stimulus set produced the greatest proportion of indiscriminate females in pilot playbacks with females from amplexus (indiscriminate: N=11; discriminating: N450 Hz =4, N500 Hz =2); a greater proportion of indiscriminate females in tests with females captured in amplexus should increase the likelihood of detecting greater discrimination by females tested before entering amplexus. We recorded each stimulus on separate channels of a cassette tape at a rate of 60 calls/min with a fixed timing relationship, such that the two stimuli always alternated with each other, with 0.5 s elapsing between the beginning of one stimulus and the beginning of the other. We tested each female in four trials before she had entered amplexus and in four trials after she had entered amplexus. To control for side biases, we switched stimuli between speakers after the first two trials in both the before and after tests of each female. Females were released from the restraining cage after each stimulus had been repeated five times (i.e. after 5 s). We reduced the restraint time in this experiment relative to that in the call-rate experiment to minimize the total amount of time between the capture of a female and the beginning of her last trial; by doing so, we hoped to minimize any decrease in a female’s discrimination that might be caused by a delay between her capture in the fence and testing. After the before test, we transported females back to the pond, where we observed all but one until they entered amplexus with a male. We immediately collected amplexed pairs and held them for testing. We captured the one female that we did not observe entering amplexus about 0.5 h after releasing her inside the fence. We did not separate females from their mates until immediately before their first trial in the after test. We tested 19 females before and after amplexus; four additional females did not respond in the before test. We tested females in their first trial of the before test within an average (&) of 15.1&3.5 min of their capture in the fence
885
(range=12–28 min) and in their last trial of the before test within an average of 25.2&4.0 min of their capture in the fence (range=21–36 min). We tested females in their first trial of the after test within an average of 79.2&31.4 min (range= 21–128 min) of their capture following mating and in the last trial of the after test within an average of 87.4&32.0 min (range=28–135 min) of their capture following mating. Response Times in Before and After Tests For the call-rate experiment, we used a paired test to compare the response time of each female in the before and after tests. We defined response time as the time a female took to move to a speaker after we had removed the lid from the restraining cage. For the fundamental frequency experiment, we used repeated-measured ANOVA (Tabachnick & Fidell 1989) to compare the response times of females in the four trials of the before test with their response times in the four trials of the after test. We entered as independent variables two within-subject factors, amplexus state (before or after) and trial (first, second, third, fourth), and one between-subject variable, whether the female had been tested previously (yes, no). Response times were log-transformed to meet the assumption of normality. In conducting a posteriori comparisons following the overall ANOVA, we controlled comparison-wide type I error rates with the sequential Bonferroni method (Rice 1989). Because there were no cases where a test was judged statistically significant (P<0.05) without the use of this technique but nonsignificant with it, however, we present only unadjusted P-values. Effect of Time in Amplexus and Time of Night on Responses in the After Test Our before-and-after experiments explicitly tested whether amplexus influences the discrimination of females. The amplexus state of a female may not be the sole determinant of her degree of discrimination, however: discrimination might decline as females spend more time in amplexus or as the time remaining for the chorus to continue decreases (i.e. as time of night increases). Because our experiments did not completely decouple amplexus state from these two variables, we conducted two analyses to determine whether time
Animal Behaviour, 51, 4
886
spent in amplexus and time of testing influenced discrimination. First, we used multiple logistic regression analysis (Hosmer & Lemeshow 1989) to determine whether the probability that females tested in the fundamental frequency experiment became less discriminating between the before and after tests was influenced by the time that females spent in amplexus prior to their testing in the after test and the time of night (hours after sunset) of their first trial in the after test. We defined a female as becoming less discriminating if she chose one stimulus four of four times in the before test but fewer times in the after test. Second, we examined the influence of time spent in amplexus and time of night on response times in the after experiments. For females tested in the fundamental frequency experiment, we conducted a multiple regression analysis in which we regressed average response time in the after trials on two independent variables, time since amplexus and time of the night that the after tests were begun. Because transformations failed to normalize the residuals, we conducted the analysis using the ranks of the data (Conover 1980). For females tested in the call-rate experiment, we examined the correlation between response time in the after test and time of testing. We could not examine the relationship between response time and time spent in amplexus for these females because we recorded this latter variable for only one of them. RESULTS Choices of Females in Before and After Tests For the call-rate experiment, the hypothesis that females are more discriminating before they enter amplexus, in conjunction with the results of our pilot experiment (Fig. 1), predicts that (1) a greater proportion of females should choose the fast call rate (65 calls/min) in the before test than in the after test, and (2) females that switch their choice between the before and after tests should be more likely to switch from choosing the fast call rate to choosing the slow call rate than vice versa. The logic behind this prediction is as follows. If females become less discriminating after entering amplexus, then a lower proportion of females will choose the high call rate in the after test than in the before test. For more females to choose the slow call in the after test than in the before test, more females must have switched their choice between the before and after tests from
Table I. Responses of females tested in the call-rate experiment Choice in ‘after’ test Choice in ‘before’ test Fast call rate* Slow call rate† Total
Fast call rate*
Slow call rate†
Total
9 4 13
5 2 7
14 6 20
Chi-squared goodness-of-fit test using results from before test as the expected values for the after test: ÷2 =0.238, df=1, P (one-tailed)=0.31. *Fast call rate=65 calls/min. †Slow call rate=55 calls/min.
the fast call to the slow call than vice versa. Neither prediction was supported. The proportion of females choosing the fast call in the before test (70%) was only slightly greater than that in the after test (65%), and this difference was not statistically significant (Table I). The number of females that switched from choosing the fast call rate to choosing the slow call rate (N=5) was not significantly greater than the number that switched their choices in the opposite direction (N=4, Conover 1980, McNemar test, T2 =5, N=9, P (one-tailed)=0.37). For the fundamental frequency experiment, the hypothesis that females are more discriminating before they enter amplexus predicts that (1) the proportion of discriminating females (i.e. females choosing one alternative in all four trials) should be greater in the before test than in the after test, and (2) females that switch their discrimination status between the before and after tests should be more likely to switch from discriminating to indiscriminate than vice versa. Neither prediction was supported. Although the proportion of discriminating females in the before test (31.6%) was slightly greater than that in the after test (26.3%), this difference was not statistically significant (Table II). In the before test, four of the six discriminating females chose the 450-Hz call, and two chose the 500-Hz call. In the after test, one of the five discriminating females chose the 450-Hz call, and the other four chose the 500-Hz call. The number of females that switched from being discriminating to being indiscriminate (N=5) was not significantly greater than the number that switched in the opposite direction (N=4, McNemar test, T2 =5, N=9, P (one-tailed)=0.37).
Murphy & Gerhardt: Effect of amplexus on mate choice
887
Table II. Responses of females tested in the fundamental-frequency experiment Response in ‘after’ test Response in ‘before’ test* Discriminating Indiscriminate Total
Discriminating
Indiscriminate
Total
1 4 5
5 9 14
6 13 19
*‘Discriminating’ females chose the same stimuli in all four trials; ‘indiscriminate’ females chose each stimulus at least once during the four trails (see text). Chi-squared goodness-of-fit test using results from before test as the expected values for the after test: ÷2 =0.244, df=1, P (one-tailed)=0.31. Table III. Power of the chi-squared goodness-of-fit test to detect a difference between the observed proportion of females choosing an alternative in the ‘after’ test and several hypothetical proportions for the ‘before’ test
Experiment
Proportion observed in ‘after’ test
Call rate† (N=20)
0.65 (13:7)
Fundamental frequency‡ (N=19)
0.26 (5:14)
Hypothetical proportion in ‘before’ test 0.80 0.85 0.90 0.47 0.53 0.58 0.63
(16:4) (17:3) (18:2) (9:10) (10:9) (11:8) (12:7)
Power* 0.51 0.83 0.99 0.58 0.76 0.89 0.97
*á (one-tailed)=0.05; calculated using Milligan’s (1979) approximation. †Proportions are proportions of females choosing high call-rate; numbers in parentheses represent ratio of females choosing high call-rate to females choosing low call-rate. ‡Proportions are proportions of ‘discriminating’ females; numbers in parentheses represent ratio of ‘discriminating’ females to ‘indiscriminate’ females. See text for definitions of discriminating and indiscriminate.
We conducted power analyses (Cohen 1988) to determine whether our sample sizes provided the chi-squared goodness-of-fit test with sufficient power to reject the null hypothesis that the proportion of females choosing one alternative was greater in the before test than in the after test. Based on Cohen’s (1988) recommendation that a power of 0.80 is desirable, the sample size of our call-rate experiment (N=20) was sufficient to detect a difference between the observed proportion (65%) of females choosing the high callrate in the after test and proportions of females choosing the high call-rate in the before test of 85% or greater (Table III). The sample size of our fundamental frequency experiment (N=19) was sufficient to detect a difference between the observed proportion of discriminating females in the after test (26%) and proportions of discrimi-
nating females in the before test of 58% or greater (Table III). Response Times in Before and After Tests The hypothesis that females tested before entering amplexus are more discriminating than females tested after entering amplexus predicts that females should make a choice more quickly in the after tests than in the before tests. Failure to support this prediction would not falsify the hypothesis, but support of this prediction would be consistent with the hypothesis. Females tested in the call-rate experiment took significantly longer to make a choice in the before test (median=2.5 min; interquartile range=1.2–3.1 min) than in the after test (median=0.8 min; inter-quartile range=0.6– 1.4 min; Wilcoxon matched-pairs signed-ranked
Animal Behaviour, 51, 4
888
2.0 Average response time (min)
6
Number of females
5 4 3 2 1 0
** 1.5
1.0
*
0.5
0.0
Before
After Amplexus state
–1 0 1 2 3 4 5 6 Difference in response time between before and after tests (min)
Figure 2. Difference between the before and after tests in the time required by individual females to make a choice in the call-rate experiment. Positive values (.) indicate females that took more time to respond in the before test than in the after test; negative values (/) indicate females that took less time to respond in the before test than in the after test. Arrow indicates median.
test, z= "3.55, P=0.0004; Fig. 2). Whether females tested in the fundamental frequency experiment took longer to make a choice in the before test than in the after test depended on whether we had tested them previously that season (either in the call-rate experiment or in other playback experiments). A repeated-measures ANOVA revealed that the effect of previous testing approached significance (F1,17 =4.12, P=0.058; N=19), the effect of amplexus state (before or after) was highly significant (F1,17 =20.67, P<0.001), and the interaction between previous treatment and amplexus state was significant (F1,17 =4.90, P=0.041). Because the response times of individual females did not differ between trials within before or after tests (F3,51 =2.18, P=0.10), and because there were no interactions between trial and amplexus status or previous treatment (P§0.36), we explored further the effect of previous testing and amplexus state on response time by averaging each female’s four trials within a before or after test and comparing these averages. Females that had not been tested earlier in the season took significantly longer to make their choice in the before test than in the after tests (paired t-test of log-transformed averages, t=4.70, P=0.001, N=10; Fig. 3), but females that we
Figure 3. Median (&one quartile) response times of females tested in the before and after tests of the fundamental frequency playback, shown separately for females that had (,) and had not been (-) tested in a playback experiment earlier in the season (see text). Response times for each female are the average of the four trials of the particular test (before or after). **P=0.001, paired t-test on log-transformed data, *P=0.0165, two-sample t-test on log-transformed data.
had tested earlier in the season took the same amount of time to respond in the before and after tests (paired t-test of log-transformed averages, t=1.47, P=0.18, N=9; Fig. 3). Previously tested females took significantly longer than previously untested females to make a choice in the before test (two-sample t-test of log-transformed averages, t=2.66, P=0.0165, N=19) but not in the after test (two-sample t-test of log-transformed averages, t=0.04, P=0.97, N=19; Fig. 3). The reduction in response times between the before and after tests did not occur because females took more direct paths to their chosen speaker in the after test than in the before test. To test this hypothesis, we qualitatively compared the paths taken by each female in her before and after test in the call-rate experiment. We categorized each pair of paths as being more direct in the before test, equally direct in both tests, or more direct in the after test. We examined paths from the call-rate experiment rather than from the fundamental frequency experiment, because the former included more females that had not been previously tested (N=20) than did the latter (N=10) and because comparisons were easier to make between the single before and after tests of the call-rate experiment than between the four before trials and four after trials of the fundamental frequency
Murphy & Gerhardt: Effect of amplexus on mate choice experiment. The paths of six females (30%) were more direct in the after test than in the before test, and the paths of 14 (70%) were either equally direct in both tests (N=8) or more direct in the before test than in the after test (N=6). Thus, the reduction in response time between the before and after tests does not appear to be the result of females taking more direct paths to the speakers in the after tests.
889
(Spearman rank correlations, N=18, time spent in amplexus: rS = "0.10, P=0.67; time of testing: rS =0.20, P=0.20). There was no correlation between time of testing and response time in the after tests for the call-rate experiment (Spearman rank correlation, rS =0.31, P=0.17, N=20).
DISCUSSION Effect of Time in Amplexus and Time of Night on Responses in the After Test If the time that females spent in amplexus before testing or the time of night when they were tested influenced their discrimination, then these two variables should be positively related to the probability that females will become less discriminating between the before and after tests. Five of the females tested in the fundamental frequency experiment became less discriminating between the before and after tests (i.e. chose one stimulus four of four times in the before test but fewer times in the after test), but 13 females either did not change their degree of discrimination or became more discriminating (one additional female became less discriminating, but we did not record the time this female spent in amplexus). Neither time spent in amplexus nor time of testing was related to the probability that a female would become less discriminating between the before and after test; this result held for both the multiple logistic regression model (change in scaled deviance=3.68, ÷2 test, df=3, P>0.50) and univariate logistic regression models (time in amplexus: change in scaled deviance=0.79, ÷2 test, df=1, P>0.50; time of testing: change in scaled deviance=0.17, ÷2 test, df=1, P>0.75). The hypothesis that discrimination by females is influenced by the time that they spend in amplexus before testing and the time of night when they are tested also predicts that these two variables should be negatively related to the response times of females in the after tests. Failure to support this prediction would not falsify the hypothesis, but support of this prediction would be consistent with the hypothesis. In the fundamental frequency experiment, there was no relationship between average response times in the after tests and either time spent in amplexus or time of testing in either the multiple regression analysis (overall model of ranked data: r2 =0.09, P=0.50, N=18) or the univariate correlations
The results of our experiments support the hypothesis that females of H. gratiosa captured in amplexus and tested in two-choice playback experiments are as discriminating as females that have not yet entered amplexus. Our sample sizes (19–20) provided adequate power to detect a difference between the before and after tests of at least 0.20 in the proportion of females choosing the high call-rate in the call-rate experiment and a difference of at least 0.32 in the proportion of discriminating females in the fundamental frequency experiment. Thus, we conclude that there are no large differences in the discrimination of females tested before and after they enter amplexus, although we cannot rule out the possibility that smaller differences exist. Although we transported females to the playback arena and tested them as quickly as possible, we cannot rule out the possibility that the short delay between the capture and testing of females in the before test reduced the discrimination of these females enough to offset any greater discrimination that they might have shown because they had not yet entered amplexus. In actuality, this concern has more to do with the effect of handling and transport on discrimination than it has to do with the effect of amplexus on discrimination. Thus, given that females are to be transported to a playback arena away from the breeding pond where they are captured, our results suggest that whether females are collected before or after they enter amplexus will have little effect on the preferences revealed by playback experiments. The only difference that we observed in the behaviour of females between the before and after tests was that females that had not been tested earlier in the season made their choices more quickly in the after test than in the before test (Figs 2, 3). Because previously tested females did not show this reduction, and because they spent less time making a choice in the before test than
Animal Behaviour, 51, 4
890
did previously untested females, we conclude that previously untested females did not reduce their response times between the before and after tests because they had entered amplexus. Rather, we hypothesize that these females reduced their response times in the after test because they had gained some sort of experience during the before test. As a result of being tested earlier in the night, these females may have become habituated to handling and transport or may have become familiar with something about the playback arena or the alternative stimuli. This hypothesis would also explain why females responded equally quickly in the before and after tests when tested a second time during the season: these females may have gained experience from having been tested earlier in the season. Additional experiments are required to test this hypothesis. Our study provides support for the long-held assumption that the preferences of females captured in amplexus accurately reflect the preferences of females that have not yet mated; this assumption is implicit in virtually all mate-choice experiments with anuran amphibians. Because we conducted before and after tests with two different call properties, one static and one dynamic, our results appear to be general with respect to these different classes of call properties. Similar tests with other species are required, however, before the practice of testing females captured in amplexus can be considered to be fully validated. ACKNOWLEDGMENTS We are grateful to the United States Forest Service for permission to conduct this study, Cris Copley, Scott Hamilton, Dave Johnson and Amy Winkeler for excellent and dedicated assistance in the field, Josh Schwartz for synthesizing the stimuli used in playbacks, and Rick Howard, Josh Schwartz and Kent Wells for providing helpful comments on the manuscript. This research was supported by a National Science Foundation grant (IBN-9122219) to H.C.G. REFERENCES Andersson, M. 1994. Sexual Selection. Princeton, New Jersey: Princeton University Press. Arak, A. 1988. Female mate choice in the natterjack toad: active choice or passive attraction? Behav. Ecol. Sociobiol., 22, 317–327.
Bradbury, J. W. & Davies, N. B. 1987. Relative roles of intra- and intersexual selection. In: Sexual Selection: Testing the Alternatives (Ed. by J. W. Bradbury & M. B. Andersson), pp. 143–163. New York: John Wiley. Cohen, J. 1988. Statistical Power Analysis for the Behavioral Sciences. 2nd edn. Hillsdale, New Jersey: Lawrence Erlbaum Associates. Conover, W. J. 1980. Practical Nonparametric Statistics. 2nd edn. New York: John Wiley. Daugherty, C. H. 1976. Freeze-branding as a technique for marking anurans. Copeia, 1976, 836–838. Gerhardt, H. C. 1981. Mating call recognition in the barking treefrog (Hyla gratiosa): responses to synthetic calls and comparisons with the green treefrog (Hyla cinerea). J. comp. Physiol. A, 144, 17–25. Gerhardt, H. C. 1988. Acoustic properties used in call recognition by frogs and toads. In: The Evolution of the Amphibian Auditory System (Ed. by B. Fritzsch, M. Ryan, W. Wilczynski, T. Hetherington & W. Walkowiak), pp. 455–483. New York: John Wiley. Gerhardt, H. C. 1991. Female mate choice in treefrogs: static and dynamic acoustic criteria. Anim. Behav., 42, 615–635. Gerhardt, H. C. 1994. Evolution of vocalization in frogs and toads. A. Rev. Ecol. Syst., 25, 293–324. Halliday, T. R. 1983. The study of mate choice. In: Mate Choice (Ed. by P. Bateson), pp. 3–32. Cambridge: Cambridge University Press. Hosmer, D. W. Jr & Lemeshow, S. 1989. Applied Logistic Regression. New York: John Wiley. Kirkpatrick, M. & Ryan, M. J. 1991. The evolution of mating preferences and the paradox of the lek. Nature, Lond., 350, 33–38. Lopez, P. T. & Narins, P. M. 1991. Mate choice in the neotropical frog, Eleutherodactylus coqui. Anim. Behav., 41, 757–772. Milligan, G. W. 1979. A computer program for calculating power of the chi-squared test. Educational and Psychological Measurement, 39, 681–684. Martof, B. S. 1961. Vocalization as an isolating mechanism in frogs. Am. Midl. Nat., 65, 118–126. Martof, B. S. & Thompson, E. F., Jr 1964. A behavioral analysis of the mating call of the chorus frog, Pseudacris triseriata. Am. Midl. Nat., 71, 198–209. Murphy, C. G. 1993. A modified drift fence for capturing treefrogs. Herpetol. Rev., 24, 143–145. Murphy, C. G. 1994. Chorus tenure of male barking treefrogs, Hyla gratiosa. Anim. Behav., 48, 763–777. Oldham, R. S. & Gerhardt, H. C. 1975. Behavioral isolation of the treefrogs Hyla cinerea and Hyla gratiosa. Copeia, 1975, 223–231. Pomiankowski, A. 1988. The evolution of female mate preferences for male genetic quality. Oxford Surveys in Evolutionary Biology, 5, 136–184. Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution, 43, 223–225. Schwartz, J. J. 1994. Male advertisement and female choice in frogs: new findings and recent approaches to the study of communication in a dynamic acoustic environment. Am. Zool., 34, 616–624. Tabachnick, B. G. & Fidell, L. S. 1989. Using Multivariate Statistics. New York: Harper Collins.
Evaluating the design of mate-choice experiments: the effect of amplexus on mate choice by female barking treefrogs, Hyla gratiosa CHRISTOPHER G. MURPHY & H. CARL GERHARDT Division of Biological Sciences, University of Missouri, U.S.A. (Received 21 November 1994; initial acceptance 19 March 1995; final acceptance 11 August 1995; MS. number 7163)
Abstract. The proper design of mate-choice experiments is critical to detecting and characterizing mating preferences. Female anuran amphibians tested in mate-choice experiments are almost always collected in amplexus (i.e. after they have chosen a mate); it has been assumed that these females are as discriminating as they are before they enter amplexus. This assumption was tested by testing females of the barking treefrog, Hyla gratiosa, twice within the same night, once before they entered amplexus (‘before’ test) and once after (‘after’ test). Two such experiments were conducted in which females were given a choice between two calls that differed either in repetition rate or fundamental frequency. In both experiments, the proportion of females choosing each alternative did not differ between the before and after tests. Some females switched their choice between the before and after tests; in both experiments, equal numbers of females switched in one direction as in the other. The response of females in the after tests was not affected by either the time that they had spent in amplexus or the time of night when they were tested. Females that had not been tested earlier in the season took significantly less time to make their choice in the after test than in the before test, but females that had been previously tested did not. The results support the assumption that female anuran amphibians captured in amplexus are as discriminating as they are when first selecting mates in natural choruses. ?
The evolution of mate choice is one of the most exciting and controversial areas in evolutionary biology (Halliday 1983; Bradbury & Davies 1987; Pomiankowski 1988; Kirkpatrick & Ryan 1991; Andersson 1994). Two prerequisites for the study of the evolution of mate choice are demonstrating that females show choice and determining which male characters influence mate choice. Accomplishing these goals requires that mate-choice experiments be designed properly; improperly designed experiments could lead to incorrect conclusions about the existence of mate choice, the characteristics that influence mate choice or the degree of discrimination exhibited by females. Anuran amphibians have become model systems in which to examine mate choice (reviewed in Gerhardt 1988, 1994). Because females of Correspondence: C. G. Murphy, Department of Biology, James Madison University, Harrisonburg, VA 22807, U.S.A. (email: [email protected]). H. C. Gerhardt is at the Division of Biological Sciences, Tucker Hall, University of Missouri, Columbia, MO 65211, U.S.A. 0003–3472/96/040881+10 $18.00/0
?
1996 The Association for the Study of Animal Behaviour
many of these species base their choice of mates predominantly or exclusively on acoustic signals, cues in additional sensory modalities are unnecessary for eliciting natural mating behaviour from females. Techniques for recording, analysing and synthesizing such signals (e.g. Schwartz 1994) allow researchers to generate acoustic stimuli that are as attractive as natural calls (Gerhardt 1981) and to vary virtually any property of calls. Females behave toward speakers broadcasting advertisement calls as they would toward calling males: they make a series of movements toward a speaker, eventually contacting it. For most species, such behaviour is an unambiguous indication of a choice, because a female approaching a calling male in this manner in a natural chorus almost always mates with that male (unless she is intercepted by a satellite male). Many playback studies have demonstrated that female anuran amphibians do indeed choose mates and that numerous properties of calls influence mate choice in these species (reviewed in Gerhardt 1988, 1994). 1996 The Association for the Study of Animal Behaviour
881
882
Animal Behaviour, 51, 4
In almost all playback experiments conducted with anuran amphibians (cf. Martof 1961; Martof & Thompson 1964; Lopez & Narins 1991), researchers tested females that had been captured in amplexus and separated from their mates. There are two advantages in using females captured in amplexus for mate-choice experiments. First, such females are often much easier to find than females that have not yet chosen a mate. Second, females found in amplexus may be more likely to respond in playback experiments (93%: Gerhardt 1981; 88%: Arak 1988) than are females that have not yet chosen a mate (33–37%: Lopez & Narins 1991; cf. Martof 1961). Despite these advantages, there are reasons to expect that females tested after they have entered amplexus may be less discriminating than females that have not yet entered amplexus. In many playback experiments, females are tested several hours after they have entered amplexus; by this time, choruses of many species have ended for the night. Under natural conditions, a female losing her male at this time (e.g. due to predation of the male) will have few, widely separated males from which to choose. Females that choose to by-pass nearby males for more distant males will often have to traverse relatively large distances, thereby increasing their expenditure of energy and risk of predation. In addition, females may have a limited time in which to complete oviposition, and time spent discriminating between males may reduce the time available for oviposition. Finally, in some species (e.g. hylid treefrogs), females that have already entered amplexus will lay eggs even if they lose their mate and do not remate; in these species, ovulation probably begins at the time of amplexus or shortly thereafter, and once ovulation begins oviposition follows in a matter of hours. For these reasons, selection may favour females that, when separated from their mate, reduced their discrimination and mate quickly with nearby males. Thus, testing females captured in amplexus may underestimate the discrimination of females selecting mates under natural conditions. Although playback experiments with anuran amphibians have been conducted for over 30 years, no one has tested the assumption that females collected in amplexus behave the same as females that have not yet chosen a mate. In this study, we tested this assumption using a paired design in which we compared the choices made by individual female barking treefrogs, Hyla gratiosa,
before and after entering amplexus. To increase the generality of our results, we conducted two experiments. In one, we presented females with stimuli that differed in call repetition rate, a dynamic property that varies considerably during a bout of calling (Gerhardt 1991); in the other, we used stimuli that differed in frequency, a static property that varies little within a bout of calling (Gerhardt 1991).
METHODS Study Site and Species This study was conducted in the Apalachicola National Forest 4 km southwest of Tallahassee, Florida (30)22*N, 84)21*W). In this area, H. gratiosa breeds from March through August in semi-permanent ponds that lack fish. Shortly after sunset, males descend from the canopy of forests surrounding breeding ponds to begin calling in these ponds. They inflate themselves and float on the surface of the water, calling from the perimeter of ponds for 1–4 h. Females arrive at the pond shortly after males, approaching males and nudging them to initiate amplexus. Males do not defend oviposition sites, almost never behave as satellites, do not pursue females and do not attempt to dislodge alreadymated males. Pairs remain near the male’s call site for 1–2 h before disapppearing underwater to lay eggs. Male H. gratiosa produce advertisement calls consisting of a pulsatile onset followed by a harmonic series with two or three spectral peaks, the fundamental and one or two of the upper harmonics (Oldham & Gerhardt 1975; Gerhardt 1981). In the population monitored in this study, the average (&) fundamental frequency was 470 Hz&34 (range=410–550, N=30), the upper frequency peak was usually the fourth harmonic, the average call duration was 150&18.1 ms (range=112.8–118.8, N=30), and males produced an average of 54.9&5.1 calls per min (range= 42–66 calls/min, N=81). General Methods for Playback Experiments Each female was tested with the same playback experiment twice within the same night, once before and once after she entered amplexus. To
Murphy & Gerhardt: Effect of amplexus on mate choice obtain females before they entered amplexus, we used a drift fence to intercept females as they approached the pond (Murphy 1993). We transported the captured female to a playback arena located 0.45 km from the pond and tested her in a two-choice playback (‘before’ tests); we tested females as soon as possible after capture (about 15 min). Following the before test, we transported the female back to the pond and released her, allowing her to choose a mate. We collected the mated pair and transported it back to the playback arena, where we separated the female from her mate and retested her in the same playback used in the before test. Following this test (‘after’ test), we freeze-branded the female (Daugherty 1976) to prevent re-testing her in the same experiment (i.e. call-rate or fundamental frequency experiment). Freeze brands last for the duration of the breeding season (Murphy 1994). After the female was marked, we returned her and her mate to the pond so that the female could oviposit. Although no female was tested twice in the same before-and-after experiment, nine females were tested in the fundamental frequency experiment after having been tested an average (&) of 22.7&9.0 days (range=9–37 days) previously in other playback experiments (either the call-rate experiment or a set of playbacks for another study). We used two sites for playbacks. The first site was the shore of a dried-up pond; the substrate for this arena was sand, covered with short (7.5– 10 cm) vegetation. Eventually, the water level of the pond rose enough that other species of anurans (Gastrophryne carolinensis, H. femoralis) began calling there, producing background noise that made it difficult to adjust the sound pressure level (SPL) of the playback stimuli. Therefore, we began conducting playbacks at another site located on a nearby, unpaved road. This site also had a sandy substrate but lacked vegetation. We conducted all trials of the call-rate playback at the first site. We tested 9 of the 19 females in the fundamental frequency experiment at the first site and the remaining 10 at the second site. Within an experiment, each female was tested at the same site for both the before and after tests. At the beginning of a trial, we placed the female in an acoustically transparent, hardware–cloth cage 4.4 m from two speakers, which were separated by 3.5 m; the cage and the two speakers
883
formed a triangle. We used this arrangement, rather than releasing females equidistant between two speakers, because it more closely mimics the array of males that confront females when they choose mates in natural choruses (Murphy & Gerhardt, unpublished data). The distance between the two speakers (3.5 m) is very similar to the average inter-male distance in natural choruses (3.6 m; Murphy & Gerhardt, unpublished data). After beginning the playback, we released the female from the restraining cage by manually removing its lid. We observed the movements of females using a headlamp dimmed with a rheostat to an intensity at which the female was just visible. We scored a female as having chosen a speaker when she either passed within 10 cm of the speaker or circled it (in natural choruses, females sometimes make two or three close passes by a male before actually contacting him). We scored a ‘no response’ if the female did not leave the release cage after 3 min or if we lost sight of her when she hopped several metres outside of the triangle delineated by the release point and the speakers. We also recorded the time each female took to move to a speaker after the lid was removed from the restraining cage, as well as her path between the release point and the speaker. Stimuli used in playbacks were synthetic calls produced on an Amiga 500 computer (sampling rate of 20 kHz, 8-bits/sample) with custom software (J. Schwartz, unpublished data). All calls lasted 175 ms and had a pulsatile beginning followed by a harmonic series that consisted of the fundamental frequency plus the third and fourth harmonic, both of which were attenuated 6 dB relative to the fundamental; these calls are as attractive to females as are natural calls (Gerhardt 1981). Calls were recorded in stereo on to cassette tape with a Sony Pro Walkman recorder, played back from that tape recorder through a Sony XM-3021 stereo amplifier and broadcast from Realistic Minimus 7 speakers. This system was flat (&3 dB) over the frequency range encompassed by the experimental stimuli (0.4–2.0 kHz). The SPL (re 20 ìPa, ‘fast’ root-mean-square, Cweighting) of the two choices was equalized at 70 dB at the release point with a Realistic sound level meter; this intensity is similar to that measured at the same distance from calling males in natural choruses (70–72 db; Murphy, unpublished data).
Animal Behaviour, 51, 4
% choosing stimulus
884
100
P=0.004 N=15
P=0.002 N=13
P=0.034 N=43
80 60 40 20 0
45 versus 65 45 versus 55 55 versus 65 Stimulus (calls/min)
Figure 1. Percentage of females choosing low call-rate (.) and high call-rate (/) alternatives in two-choice playback experiments. Females used in these experiments were captured in amplexus. Significance values are two-tailed P-values from a binomial test of the null hypothesis of no preference (i.e. equal percentage of females choosing each alternative).
Call-rate Experiment In this experiment, we tested whether amplexus affected discrimination based on call repetition rate. To determine which call rates to use in the before-and-after tests, we conducted pilot experiments with females captured in amplexus. In the pilot experiments, we placed females midway between two speakers separated by 4 m and equalized the SPL of the two choices at 74 dB SPL at the release point. Females overwhelmingly preferred call rates either at the upper end of the natural distribution (65 calls/min) or at the population average (55 calls/min) over those near the lower end of the natural distribution (45 calls/min; Fig. 1); a lower percentage of females preferred the high call rate (65 calls/min) over the average call rate (55 calls/min; Fig. 1). We chose this latter stimulus set (high versus average call rates) to test the hypothesis that females are more discriminating before than after they enter amplexus, because it was the only set for which females captured before entering amplexus could be substantially more discriminating than they were in the pilot experiments. We recorded a synthetic call with a fundamental frequency of 400 Hz on to one channel of a cassette tape at 65 calls/min and on to the other channel at 55 calls/min. Because the two stimuli differed in repetition rate, the phase relationship of the calls from the two speakers changed during the playback.
We tested each female once before and once after she had entered amplexus. After placing the female in the restraining cage, we played back the stimuli for 30 s before releasing the female. To control for side biases, we switched the stimuli between speakers for each new female; for each female, however, we broadcast the same stimulus from the same speaker in both the before and after tests. Following the before test, we transported females back to the pond. We released 18 of the 20 females inside the fence and collected them and their mate 0.5–1.5 h later; we released the remaining two females inside the fence, observed them until they entered amplexus, and immediately collected them and their mates. We did not separate females from their mates until immediately before their after tests. We tested 20 females before and after amplexus; six additional females did not respond in the before test. We tested females in the before test within an average of (&) of 14.1&1.3 min of their capture in the fence (range=12–18 min) and in the after test within an average of 40.2& 12.7 min (range=16–66 min) of their capture following mating. Fundamental Frequency Experiment In this experiment, we tested whether amplexus affected discrimination based on fundamental frequency. Previous two-choice playback experiments conducted with females captured in amplexus showed that females generally prefer calls with intermediate fundamental frequencies over calls with either higher or lower fundamental frequencies (Gerhardt 1981; Murphy & Gerhardt, unpublished data); however, which frequency was preferred differed between females (Murphy & Gerhardt, unpublished data). Hence, we could not make a directional prediction about which fundamental frequency females would prefer in the before and after tests, as we did for the call-rate experiments. Instead, we presented females with two synthetic calls that differed in fundamental frequency (450 Hz versus 500 Hz); this difference represents 1.5 standard deviations of the distribution of fundamental frequencies in this population. We scored a female as ‘discriminating’ if she chose the same stimulus in each of four trials and as ‘indiscriminate’ if she did not; the probability of a female choosing the same stimulus in all four trials by chance alone (i.e. in the absence of
Murphy & Gerhardt: Effect of amplexus on mate choice preferences or carry-over effects) is 0.063. Our results are not affected if we use a less conservative measure of ‘discriminating’ (i.e. choice of the same stimulus in three or more of the four trials). We conducted four trials, rather than some greater number, to minimize the amount of time that elapsed between capture of the female and the beginning of her last trial. We chose alternatives of 450 Hz and 500 Hz because this stimulus set produced the greatest proportion of indiscriminate females in pilot playbacks with females from amplexus (indiscriminate: N=11; discriminating: N450 Hz =4, N500 Hz =2); a greater proportion of indiscriminate females in tests with females captured in amplexus should increase the likelihood of detecting greater discrimination by females tested before entering amplexus. We recorded each stimulus on separate channels of a cassette tape at a rate of 60 calls/min with a fixed timing relationship, such that the two stimuli always alternated with each other, with 0.5 s elapsing between the beginning of one stimulus and the beginning of the other. We tested each female in four trials before she had entered amplexus and in four trials after she had entered amplexus. To control for side biases, we switched stimuli between speakers after the first two trials in both the before and after tests of each female. Females were released from the restraining cage after each stimulus had been repeated five times (i.e. after 5 s). We reduced the restraint time in this experiment relative to that in the call-rate experiment to minimize the total amount of time between the capture of a female and the beginning of her last trial; by doing so, we hoped to minimize any decrease in a female’s discrimination that might be caused by a delay between her capture in the fence and testing. After the before test, we transported females back to the pond, where we observed all but one until they entered amplexus with a male. We immediately collected amplexed pairs and held them for testing. We captured the one female that we did not observe entering amplexus about 0.5 h after releasing her inside the fence. We did not separate females from their mates until immediately before their first trial in the after test. We tested 19 females before and after amplexus; four additional females did not respond in the before test. We tested females in their first trial of the before test within an average (&) of 15.1&3.5 min of their capture in the fence
885
(range=12–28 min) and in their last trial of the before test within an average of 25.2&4.0 min of their capture in the fence (range=21–36 min). We tested females in their first trial of the after test within an average of 79.2&31.4 min (range= 21–128 min) of their capture following mating and in the last trial of the after test within an average of 87.4&32.0 min (range=28–135 min) of their capture following mating. Response Times in Before and After Tests For the call-rate experiment, we used a paired test to compare the response time of each female in the before and after tests. We defined response time as the time a female took to move to a speaker after we had removed the lid from the restraining cage. For the fundamental frequency experiment, we used repeated-measured ANOVA (Tabachnick & Fidell 1989) to compare the response times of females in the four trials of the before test with their response times in the four trials of the after test. We entered as independent variables two within-subject factors, amplexus state (before or after) and trial (first, second, third, fourth), and one between-subject variable, whether the female had been tested previously (yes, no). Response times were log-transformed to meet the assumption of normality. In conducting a posteriori comparisons following the overall ANOVA, we controlled comparison-wide type I error rates with the sequential Bonferroni method (Rice 1989). Because there were no cases where a test was judged statistically significant (P<0.05) without the use of this technique but nonsignificant with it, however, we present only unadjusted P-values. Effect of Time in Amplexus and Time of Night on Responses in the After Test Our before-and-after experiments explicitly tested whether amplexus influences the discrimination of females. The amplexus state of a female may not be the sole determinant of her degree of discrimination, however: discrimination might decline as females spend more time in amplexus or as the time remaining for the chorus to continue decreases (i.e. as time of night increases). Because our experiments did not completely decouple amplexus state from these two variables, we conducted two analyses to determine whether time
Animal Behaviour, 51, 4
886
spent in amplexus and time of testing influenced discrimination. First, we used multiple logistic regression analysis (Hosmer & Lemeshow 1989) to determine whether the probability that females tested in the fundamental frequency experiment became less discriminating between the before and after tests was influenced by the time that females spent in amplexus prior to their testing in the after test and the time of night (hours after sunset) of their first trial in the after test. We defined a female as becoming less discriminating if she chose one stimulus four of four times in the before test but fewer times in the after test. Second, we examined the influence of time spent in amplexus and time of night on response times in the after experiments. For females tested in the fundamental frequency experiment, we conducted a multiple regression analysis in which we regressed average response time in the after trials on two independent variables, time since amplexus and time of the night that the after tests were begun. Because transformations failed to normalize the residuals, we conducted the analysis using the ranks of the data (Conover 1980). For females tested in the call-rate experiment, we examined the correlation between response time in the after test and time of testing. We could not examine the relationship between response time and time spent in amplexus for these females because we recorded this latter variable for only one of them. RESULTS Choices of Females in Before and After Tests For the call-rate experiment, the hypothesis that females are more discriminating before they enter amplexus, in conjunction with the results of our pilot experiment (Fig. 1), predicts that (1) a greater proportion of females should choose the fast call rate (65 calls/min) in the before test than in the after test, and (2) females that switch their choice between the before and after tests should be more likely to switch from choosing the fast call rate to choosing the slow call rate than vice versa. The logic behind this prediction is as follows. If females become less discriminating after entering amplexus, then a lower proportion of females will choose the high call rate in the after test than in the before test. For more females to choose the slow call in the after test than in the before test, more females must have switched their choice between the before and after tests from
Table I. Responses of females tested in the call-rate experiment Choice in ‘after’ test Choice in ‘before’ test Fast call rate* Slow call rate† Total
Fast call rate*
Slow call rate†
Total
9 4 13
5 2 7
14 6 20
Chi-squared goodness-of-fit test using results from before test as the expected values for the after test: ÷2 =0.238, df=1, P (one-tailed)=0.31. *Fast call rate=65 calls/min. †Slow call rate=55 calls/min.
the fast call to the slow call than vice versa. Neither prediction was supported. The proportion of females choosing the fast call in the before test (70%) was only slightly greater than that in the after test (65%), and this difference was not statistically significant (Table I). The number of females that switched from choosing the fast call rate to choosing the slow call rate (N=5) was not significantly greater than the number that switched their choices in the opposite direction (N=4, Conover 1980, McNemar test, T2 =5, N=9, P (one-tailed)=0.37). For the fundamental frequency experiment, the hypothesis that females are more discriminating before they enter amplexus predicts that (1) the proportion of discriminating females (i.e. females choosing one alternative in all four trials) should be greater in the before test than in the after test, and (2) females that switch their discrimination status between the before and after tests should be more likely to switch from discriminating to indiscriminate than vice versa. Neither prediction was supported. Although the proportion of discriminating females in the before test (31.6%) was slightly greater than that in the after test (26.3%), this difference was not statistically significant (Table II). In the before test, four of the six discriminating females chose the 450-Hz call, and two chose the 500-Hz call. In the after test, one of the five discriminating females chose the 450-Hz call, and the other four chose the 500-Hz call. The number of females that switched from being discriminating to being indiscriminate (N=5) was not significantly greater than the number that switched in the opposite direction (N=4, McNemar test, T2 =5, N=9, P (one-tailed)=0.37).
Murphy & Gerhardt: Effect of amplexus on mate choice
887
Table II. Responses of females tested in the fundamental-frequency experiment Response in ‘after’ test Response in ‘before’ test* Discriminating Indiscriminate Total
Discriminating
Indiscriminate
Total
1 4 5
5 9 14
6 13 19
*‘Discriminating’ females chose the same stimuli in all four trials; ‘indiscriminate’ females chose each stimulus at least once during the four trails (see text). Chi-squared goodness-of-fit test using results from before test as the expected values for the after test: ÷2 =0.244, df=1, P (one-tailed)=0.31. Table III. Power of the chi-squared goodness-of-fit test to detect a difference between the observed proportion of females choosing an alternative in the ‘after’ test and several hypothetical proportions for the ‘before’ test
Experiment
Proportion observed in ‘after’ test
Call rate† (N=20)
0.65 (13:7)
Fundamental frequency‡ (N=19)
0.26 (5:14)
Hypothetical proportion in ‘before’ test 0.80 0.85 0.90 0.47 0.53 0.58 0.63
(16:4) (17:3) (18:2) (9:10) (10:9) (11:8) (12:7)
Power* 0.51 0.83 0.99 0.58 0.76 0.89 0.97
*á (one-tailed)=0.05; calculated using Milligan’s (1979) approximation. †Proportions are proportions of females choosing high call-rate; numbers in parentheses represent ratio of females choosing high call-rate to females choosing low call-rate. ‡Proportions are proportions of ‘discriminating’ females; numbers in parentheses represent ratio of ‘discriminating’ females to ‘indiscriminate’ females. See text for definitions of discriminating and indiscriminate.
We conducted power analyses (Cohen 1988) to determine whether our sample sizes provided the chi-squared goodness-of-fit test with sufficient power to reject the null hypothesis that the proportion of females choosing one alternative was greater in the before test than in the after test. Based on Cohen’s (1988) recommendation that a power of 0.80 is desirable, the sample size of our call-rate experiment (N=20) was sufficient to detect a difference between the observed proportion (65%) of females choosing the high callrate in the after test and proportions of females choosing the high call-rate in the before test of 85% or greater (Table III). The sample size of our fundamental frequency experiment (N=19) was sufficient to detect a difference between the observed proportion of discriminating females in the after test (26%) and proportions of discrimi-
nating females in the before test of 58% or greater (Table III). Response Times in Before and After Tests The hypothesis that females tested before entering amplexus are more discriminating than females tested after entering amplexus predicts that females should make a choice more quickly in the after tests than in the before tests. Failure to support this prediction would not falsify the hypothesis, but support of this prediction would be consistent with the hypothesis. Females tested in the call-rate experiment took significantly longer to make a choice in the before test (median=2.5 min; interquartile range=1.2–3.1 min) than in the after test (median=0.8 min; inter-quartile range=0.6– 1.4 min; Wilcoxon matched-pairs signed-ranked
Animal Behaviour, 51, 4
888
2.0 Average response time (min)
6
Number of females
5 4 3 2 1 0
** 1.5
1.0
*
0.5
0.0
Before
After Amplexus state
–1 0 1 2 3 4 5 6 Difference in response time between before and after tests (min)
Figure 2. Difference between the before and after tests in the time required by individual females to make a choice in the call-rate experiment. Positive values (.) indicate females that took more time to respond in the before test than in the after test; negative values (/) indicate females that took less time to respond in the before test than in the after test. Arrow indicates median.
test, z= "3.55, P=0.0004; Fig. 2). Whether females tested in the fundamental frequency experiment took longer to make a choice in the before test than in the after test depended on whether we had tested them previously that season (either in the call-rate experiment or in other playback experiments). A repeated-measures ANOVA revealed that the effect of previous testing approached significance (F1,17 =4.12, P=0.058; N=19), the effect of amplexus state (before or after) was highly significant (F1,17 =20.67, P<0.001), and the interaction between previous treatment and amplexus state was significant (F1,17 =4.90, P=0.041). Because the response times of individual females did not differ between trials within before or after tests (F3,51 =2.18, P=0.10), and because there were no interactions between trial and amplexus status or previous treatment (P§0.36), we explored further the effect of previous testing and amplexus state on response time by averaging each female’s four trials within a before or after test and comparing these averages. Females that had not been tested earlier in the season took significantly longer to make their choice in the before test than in the after tests (paired t-test of log-transformed averages, t=4.70, P=0.001, N=10; Fig. 3), but females that we
Figure 3. Median (&one quartile) response times of females tested in the before and after tests of the fundamental frequency playback, shown separately for females that had (,) and had not been (-) tested in a playback experiment earlier in the season (see text). Response times for each female are the average of the four trials of the particular test (before or after). **P=0.001, paired t-test on log-transformed data, *P=0.0165, two-sample t-test on log-transformed data.
had tested earlier in the season took the same amount of time to respond in the before and after tests (paired t-test of log-transformed averages, t=1.47, P=0.18, N=9; Fig. 3). Previously tested females took significantly longer than previously untested females to make a choice in the before test (two-sample t-test of log-transformed averages, t=2.66, P=0.0165, N=19) but not in the after test (two-sample t-test of log-transformed averages, t=0.04, P=0.97, N=19; Fig. 3). The reduction in response times between the before and after tests did not occur because females took more direct paths to their chosen speaker in the after test than in the before test. To test this hypothesis, we qualitatively compared the paths taken by each female in her before and after test in the call-rate experiment. We categorized each pair of paths as being more direct in the before test, equally direct in both tests, or more direct in the after test. We examined paths from the call-rate experiment rather than from the fundamental frequency experiment, because the former included more females that had not been previously tested (N=20) than did the latter (N=10) and because comparisons were easier to make between the single before and after tests of the call-rate experiment than between the four before trials and four after trials of the fundamental frequency
Murphy & Gerhardt: Effect of amplexus on mate choice experiment. The paths of six females (30%) were more direct in the after test than in the before test, and the paths of 14 (70%) were either equally direct in both tests (N=8) or more direct in the before test than in the after test (N=6). Thus, the reduction in response time between the before and after tests does not appear to be the result of females taking more direct paths to the speakers in the after tests.
889
(Spearman rank correlations, N=18, time spent in amplexus: rS = "0.10, P=0.67; time of testing: rS =0.20, P=0.20). There was no correlation between time of testing and response time in the after tests for the call-rate experiment (Spearman rank correlation, rS =0.31, P=0.17, N=20).
DISCUSSION Effect of Time in Amplexus and Time of Night on Responses in the After Test If the time that females spent in amplexus before testing or the time of night when they were tested influenced their discrimination, then these two variables should be positively related to the probability that females will become less discriminating between the before and after tests. Five of the females tested in the fundamental frequency experiment became less discriminating between the before and after tests (i.e. chose one stimulus four of four times in the before test but fewer times in the after test), but 13 females either did not change their degree of discrimination or became more discriminating (one additional female became less discriminating, but we did not record the time this female spent in amplexus). Neither time spent in amplexus nor time of testing was related to the probability that a female would become less discriminating between the before and after test; this result held for both the multiple logistic regression model (change in scaled deviance=3.68, ÷2 test, df=3, P>0.50) and univariate logistic regression models (time in amplexus: change in scaled deviance=0.79, ÷2 test, df=1, P>0.50; time of testing: change in scaled deviance=0.17, ÷2 test, df=1, P>0.75). The hypothesis that discrimination by females is influenced by the time that they spend in amplexus before testing and the time of night when they are tested also predicts that these two variables should be negatively related to the response times of females in the after tests. Failure to support this prediction would not falsify the hypothesis, but support of this prediction would be consistent with the hypothesis. In the fundamental frequency experiment, there was no relationship between average response times in the after tests and either time spent in amplexus or time of testing in either the multiple regression analysis (overall model of ranked data: r2 =0.09, P=0.50, N=18) or the univariate correlations
The results of our experiments support the hypothesis that females of H. gratiosa captured in amplexus and tested in two-choice playback experiments are as discriminating as females that have not yet entered amplexus. Our sample sizes (19–20) provided adequate power to detect a difference between the before and after tests of at least 0.20 in the proportion of females choosing the high call-rate in the call-rate experiment and a difference of at least 0.32 in the proportion of discriminating females in the fundamental frequency experiment. Thus, we conclude that there are no large differences in the discrimination of females tested before and after they enter amplexus, although we cannot rule out the possibility that smaller differences exist. Although we transported females to the playback arena and tested them as quickly as possible, we cannot rule out the possibility that the short delay between the capture and testing of females in the before test reduced the discrimination of these females enough to offset any greater discrimination that they might have shown because they had not yet entered amplexus. In actuality, this concern has more to do with the effect of handling and transport on discrimination than it has to do with the effect of amplexus on discrimination. Thus, given that females are to be transported to a playback arena away from the breeding pond where they are captured, our results suggest that whether females are collected before or after they enter amplexus will have little effect on the preferences revealed by playback experiments. The only difference that we observed in the behaviour of females between the before and after tests was that females that had not been tested earlier in the season made their choices more quickly in the after test than in the before test (Figs 2, 3). Because previously tested females did not show this reduction, and because they spent less time making a choice in the before test than
Animal Behaviour, 51, 4
890
did previously untested females, we conclude that previously untested females did not reduce their response times between the before and after tests because they had entered amplexus. Rather, we hypothesize that these females reduced their response times in the after test because they had gained some sort of experience during the before test. As a result of being tested earlier in the night, these females may have become habituated to handling and transport or may have become familiar with something about the playback arena or the alternative stimuli. This hypothesis would also explain why females responded equally quickly in the before and after tests when tested a second time during the season: these females may have gained experience from having been tested earlier in the season. Additional experiments are required to test this hypothesis. Our study provides support for the long-held assumption that the preferences of females captured in amplexus accurately reflect the preferences of females that have not yet mated; this assumption is implicit in virtually all mate-choice experiments with anuran amphibians. Because we conducted before and after tests with two different call properties, one static and one dynamic, our results appear to be general with respect to these different classes of call properties. Similar tests with other species are required, however, before the practice of testing females captured in amplexus can be considered to be fully validated. ACKNOWLEDGMENTS We are grateful to the United States Forest Service for permission to conduct this study, Cris Copley, Scott Hamilton, Dave Johnson and Amy Winkeler for excellent and dedicated assistance in the field, Josh Schwartz for synthesizing the stimuli used in playbacks, and Rick Howard, Josh Schwartz and Kent Wells for providing helpful comments on the manuscript. This research was supported by a National Science Foundation grant (IBN-9122219) to H.C.G. REFERENCES Andersson, M. 1994. Sexual Selection. Princeton, New Jersey: Princeton University Press. Arak, A. 1988. Female mate choice in the natterjack toad: active choice or passive attraction? Behav. Ecol. Sociobiol., 22, 317–327.
Bradbury, J. W. & Davies, N. B. 1987. Relative roles of intra- and intersexual selection. In: Sexual Selection: Testing the Alternatives (Ed. by J. W. Bradbury & M. B. Andersson), pp. 143–163. New York: John Wiley. Cohen, J. 1988. Statistical Power Analysis for the Behavioral Sciences. 2nd edn. Hillsdale, New Jersey: Lawrence Erlbaum Associates. Conover, W. J. 1980. Practical Nonparametric Statistics. 2nd edn. New York: John Wiley. Daugherty, C. H. 1976. Freeze-branding as a technique for marking anurans. Copeia, 1976, 836–838. Gerhardt, H. C. 1981. Mating call recognition in the barking treefrog (Hyla gratiosa): responses to synthetic calls and comparisons with the green treefrog (Hyla cinerea). J. comp. Physiol. A, 144, 17–25. Gerhardt, H. C. 1988. Acoustic properties used in call recognition by frogs and toads. In: The Evolution of the Amphibian Auditory System (Ed. by B. Fritzsch, M. Ryan, W. Wilczynski, T. Hetherington & W. Walkowiak), pp. 455–483. New York: John Wiley. Gerhardt, H. C. 1991. Female mate choice in treefrogs: static and dynamic acoustic criteria. Anim. Behav., 42, 615–635. Gerhardt, H. C. 1994. Evolution of vocalization in frogs and toads. A. Rev. Ecol. Syst., 25, 293–324. Halliday, T. R. 1983. The study of mate choice. In: Mate Choice (Ed. by P. Bateson), pp. 3–32. Cambridge: Cambridge University Press. Hosmer, D. W. Jr & Lemeshow, S. 1989. Applied Logistic Regression. New York: John Wiley. Kirkpatrick, M. & Ryan, M. J. 1991. The evolution of mating preferences and the paradox of the lek. Nature, Lond., 350, 33–38. Lopez, P. T. & Narins, P. M. 1991. Mate choice in the neotropical frog, Eleutherodactylus coqui. Anim. Behav., 41, 757–772. Milligan, G. W. 1979. A computer program for calculating power of the chi-squared test. Educational and Psychological Measurement, 39, 681–684. Martof, B. S. 1961. Vocalization as an isolating mechanism in frogs. Am. Midl. Nat., 65, 118–126. Martof, B. S. & Thompson, E. F., Jr 1964. A behavioral analysis of the mating call of the chorus frog, Pseudacris triseriata. Am. Midl. Nat., 71, 198–209. Murphy, C. G. 1993. A modified drift fence for capturing treefrogs. Herpetol. Rev., 24, 143–145. Murphy, C. G. 1994. Chorus tenure of male barking treefrogs, Hyla gratiosa. Anim. Behav., 48, 763–777. Oldham, R. S. & Gerhardt, H. C. 1975. Behavioral isolation of the treefrogs Hyla cinerea and Hyla gratiosa. Copeia, 1975, 223–231. Pomiankowski, A. 1988. The evolution of female mate preferences for male genetic quality. Oxford Surveys in Evolutionary Biology, 5, 136–184. Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution, 43, 223–225. Schwartz, J. J. 1994. Male advertisement and female choice in frogs: new findings and recent approaches to the study of communication in a dynamic acoustic environment. Am. Zool., 34, 616–624. Tabachnick, B. G. & Fidell, L. S. 1989. Using Multivariate Statistics. New York: Harper Collins.