- Email: [email protected]
Phenotypic variation in the mating preferences of female field crickets,Gryllus integer
Anim. Behav., 1995, 49, 1269–1281
Phenotypic variation in the mating preferences of female field crickets, Gryllus integer WILLIAM E. WAGNER, J*, ANNE-MARIE MURRAY & WILLIAM H. CADE Department of Biological Sciences, Brock University, St Catharines, Ontario L2S 3A1, Canada (Received 21 July 1993; initial acceptance 8 September 1993; final acceptance 24 May 1994; MS. number: 6764)
Abstract. Phenotypic variation in the mating preferences of female field crickets was examined. Males of this species produce a trilled calling song which varies in the number of pulses per trill, the inter-trill interval and the proportion of missing pulses within a trill. As a population, females preferred male calling songs with more pulses per trill and shorter inter-trill intervals in two-speaker choice tests, but did not discriminate between male song that varied in the proportion of missing pulses. Female preference functions were examined by sequentially presenting females with a series of songs that varied in only one parameter. As a population, the strength of the female preference for male calling song appeared to increase with the number of pulses per trill in the song. However, there was no significant variation in the strength of the preference for male calling song with either the inter-trill interval or proportion of missing pulses in the song. There was significant variation between individual females in their preference functions based on the number of pulses per trill and the inter-trill interval in male song, but not based on the proportion of missing pulses in male song. Females appeared to differ in how strongly they preferred more pulses per trill. In contrast, females appeared to differ not only in the strength of their preference based on inter-trill interval, but also in whether they preferred longer or shorter inter-trill intervals. The repeatability of preference functions within females was relatively high for number of pulses per trill (0·50) and inter-trill interval (0·59), but low for proportion of missing pulses ("0·02). Correlations between female preference functions were also examined. Females that strongly preferred more pulses per trill tended to strongly prefer shorter inter-trill intervals. In addition, females that strongly preferred shorter inter-trill intervals tended to prefer a higher proportion of missing pulses. These results suggest that selection can act on female preference functions in field crickets, and that direct selection on one preference function can result in indirect selection on other preference functions.
Female mate choice is known to have an important effect on the evolution of male morphology and behaviour (reviewed by Thornhill & Alcock 1983; Searcy & Andersson 1986; Kirkpatrick 1987). However, while numerous studies have shown that females select mates based on variation in male displays, resulting in sexual selection on these displays, little is known about the detailed structure of female preference functions (i.e. variation in the strength of a female preference with variation in a male display). The structure of female preference functions will have at least two important consequences for the nature *Present address: Nebraska Behavioral Biology Group, School of Biological Sciences, University of Nebraska, Lincoln, NE 68588, U.S.A. 0003–3472/95/051269+13 $08.00/0
?
of sexual selection on male morphology and behaviour. First, the shape of the average female preference function will determine whether female choice results in directional, stabilizing, or disruptive sexual selection on a male trait (e.g Gerhardt 1992); and second, the shape of the average female preference function will, in part, determine the strength of sexual selection by female choice; the steeper the preference function, the stronger should be sexual selection by female choice. The factors that can influence the evolution of female mating preferences have received considerable recent attention (reviewed in Bradbury & Andersson 1987; Kirkpatrick & Ryan 1991). For selection to influence the evolution of female preferences, there must be variation between 1995 The Association for the Study of Animal Behaviour
1269
1270
Animal Behaviour, 49, 5
females in their mating preferences. Few studies, however, have shown that mating preferences differ among females within a population (exceptions include Majerus et al. 1982; Sappington & Taylor 1990; Ryan et al. 1992). Even fewer studies have examined the repeatability of preferences within females (for discussions of repeatability see Lessells & Boag 1987; Boake 1989). Because the repeatability of a trait will set an upper limit on the degree to which variation in the trait derives from heritable differences among individuals (Falconer 1981), an examination of the repeatability of female mating preferences can provide information on the potential for genetic variation in mating preferences (see also Møller 1994). If the repeatability of preferences within females is low, the heritability of preferences must also be low (Falconer 1981). Between-female variation in mating preferences can also have important consequences for the nature of sexual selection on male traits. First, variation between females can reduce the strength of sexual selection; sexual selection will be stronger when females are unanimous in their preferences than when there is variation between females in their preferences. Second, variation in female mating preferences can result in frequencydependent sexual selection on male traits (e.g. O’Donald & Majerus 1984; Partridge & Hill 1984). If some females prefer one extreme of a male trait while other females prefer the opposite extreme, the selective advantage of the trait will depend on the proportion of males exhibiting each extreme relative to the proportion of females preferring each extreme. Selection on mating preferences, as on any trait, will be direct, indirect, or both (e.g. Lande & Arnold 1983; Endler 1986). Indirect selection can act on mating preferences, in part, through the correlation of one preference function with another preference function that is under direct selection. Patterns of phenotypic correlation between preference functions can thus have important consequences for selection on mating preferences. No study that we are aware of, however, has examined phenotypic correlations between female preferences. It is unknown, for example, whether females that exhibit strong preferences based on one male trait also exhibit strong preferences based on other traits, or if females that exhibit strong preferences based on one male trait show weak or no preferences based on other traits.
Inter-trill interval
100 ms
Pulse
Missing pulse
Trill
Figure 1. Sonagram of the calling song of male G. integer.
In this paper we examine phenotypic variation in female calling song preferences in a Texas population of the field cricket, G. integer. Male field crickets generally sing from the ground within or near burrows (Alexander 1961), and females select mates, in part, based on variation in male calling song (e.g. Popov & Shuvalov 1977; Hedrick 1986; Simmons 1988). In G. integer, males produce a trilled calling song, and there is variation between males in the number of pulses per trill, the proportion of missing pulses within a trill and the inter-trill interval (Fig. 1; Souroukis et al. 1992). Previous studies have examined the evolution of male calling and mating behaviour (e.g. Cade 1975, 1981; Cade & Wyatt 1984; Cade & Cade 1992) and female mating frequency (Sakaluk & Cade 1980; Solymar & Cade 1990a). Little is known about female mating preferences based on male calling song. Hedrick (1986) examined the population-level mating preferences of females from a California population of G. integer. However, it appears that the California and Texas field crickets are different species (Weissman et al. 1980) First, we examined the population-level mating preferences of females based on the number of pulses per trill, the inter-trill interval and the proportion of missing pulses in male calling song using a traditional two-stimulus choice design (e.g. Gerhardt 1974; Ryan & Wagner 1987; Wagner & Sullivan 1995). Second, we examined the shapes of the average preference functions of females based on the number of pulses per trill, the inter-trill interval and the proportion of missing pulses in male calling song. Third, we examined variation between and within females in their preference functions. And fourth, we examined phenotypic correlations between female preference functions.
Wagner et al.: Phenotypic variation in female preferences
1271
METHODS Rearing Methods Female mating preferences were tested from 18 January to 6 February 1993. The females tested were either first or second generation offspring of field-caught females from Austin, Texas. Fieldcaught females were transported to the laboratory and housed individually in plastic containers with moist vermiculite for oviposition, water (glass vials plugged with cotton), Purina cat chow, and cardboard egg cartons. Because field-caught females were isolated from males upon capture, all eggs were fertilized by males in the field. First generation offspring were derived from females collected in October 1992. Second generation offspring were derived from females collected in July 1992. For the latter, one female offspring from a field-caught female was mated to a randomly selected male from another family to produce second generation offspring. Offspring were reared in containers with their siblings. Crickets were maintained in an environmental chamber at 30)C, 40% relative humidity, and a 12:12 h light:dark cycle. Within 2 days of their final moult, females were isolated in individual containers. Because females are not receptive to mating until approximately 3–4 days after their final moult (Solymar & Cade 1990b), and because none of the isolated females deposited fertilized eggs in the cotton of the water vials, the only substrate available for oviposition, the test females were known to be virgins. Female preferences were tested within 7–14 days of their final moult. Females are receptive to mating at this time (personal observation). Testing Apparatus Female preferences were tested using a PC Kugel, a device modelled after the MacKugel of Doherty & Pires (1987). This device was similar to a track ball in design, consisting of a hollow plastic sphere (16·2 cm diameter, 34·4 g) supported on a column of air (Fig. 2). The sphere was held in position on the air column by two rollers taken from a computer ‘mouse’. The rollers were level with the horizontal axis of the sphere and perpendicular to each other. A cricket tethered on top of the sphere turned the sphere as it walked towards or away from broadcast song. The displacement of the rollers this movement caused was
Hollow plastic sphere
Sensors
Air column
Output to computer
Air line
Figure 2. Illustration of the PC Kugel used to test female mating preferences in G. integer. See text for details.
monitored by a personal computer. As with a track ball, the direction and velocity of movement of the sphere can be determined by the displacement of each roller. Because the sphere was in contact only with the two rollers, there was little frictional resistance to sphere rotation. In addition, because the sphere was lightweight, there was little inertial resistance to sphere rotation. Thus crickets easily turned the sphere as they walked. The PC Kugel was placed in a circular test chamber (diameter 101·9 cm, height 47·4 cm). The top, bottom and wall of the chamber were lined with foam to improve the localizability of broadcast song. Four speakers mounted on metal posts were placed equidistant around the test chamber. Each speaker was 35·5 cm from the position of the tethered cricket. A cricket was held in position on top of the sphere using a rotating tether supported by a rod overhanging the sphere. The tether consisted of a small piece of balsa wood (approximate dimensions: 4 mm length, 2 mm width, 2 mm height) pierced on the upper surface by a thin 5 cm length of wire. The lower surface of the wood was fixed to the cricket with a drop of hot bee’s wax applied to the dorsal surface of the cricket’s thorax. The free end of the wire rested in a tubular sleeve at the end of an adjustable metal arm extending over the sphere. This arrangement allowed the cricket to turn 360) on the sphere and allowed approximately 5 mm of vertical movement, but restricted the cricket from moving laterally. Consequently, the cricket turned the sphere as it attempted to move in response to broadcast song but remained in position on top of the sphere.
1272
Animal Behaviour, 49, 5
Song was broadcast to a tethered cricket from one of the four speakers. In response, the cricket either failed to respond (i.e. remained motionless) or turned the sphere as it attempted to move towards or away from the speaker from which the song was broadcast. Once per second the direction and velocity of movement by the sphere were determined using a custom-designed computer program. This produced a series of vectors that provided information on both the relative velocity of movement and the angle of movement relative to the speaker during each 1-s sampling period (see Doherty & Pires 1987). To calculate the absolute velocity of movement it would have been necessary to measure the relationship between velocity of sphere rotation and vector length. Such measurements were not made. However, the velocity of sphere rotation is directly proportional to vector length. Consequently, vector lengths provide reliable measures of the relative velocity of sphere rotation during movement by females. A female’s net movement towards or away from a speaker for the entire test period (i.e. the strength of a female’s preference for the broadcast song) was summarized by a ‘vector score’, calculated as Vector score=Ó[cos (vector angle)#vector length] where the angular direction of the active speaker is 0) (see also Doherty & Pires 1987). Movement towards the active speaker results in positive vector scores (cos (0))=1) while movement away from the active speaker results in negative vector scores (cos (180))= "1). In addition, greater movement towards or away from the active speaker results in greater positive or negative vector scores. Movement perpendicular to the active speaker results in vector scores of 0 (cos (90))=0, cos (270))=0). Using this method, the vector score of a female moving directly towards the active speaker will be higher than the vector score of a female moving at the same velocity at a 10) angle with respect to the active speaker. These vector scores are unitless because the absolute relationship between sphere rotation and vector length was not determined (see above). Following a series of tests, the cricket and tether were removed from the test chamber. A hot needle was then used to gently loosen the wax fixing the cricket to the tether. The cricket was not injured by this process and could thus be retested on subsequent days.
Two-speaker Choice Tests To determine whether there are populationlevel mating preferences based on the number of pulses per trill, the inter-trill interval, or the proportion of missing pulses in male calling song, female preferences were tested using three pairs of stimuli. Stimuli were constructed by digitizing a natural pulse and constructing pulse trains with a pulse rate of 70 pulses/s. First, preferences based on the number of pulses per trill in male song were tested; females were presented with a choice between male song of 15 and 70 pulses per trill (inter-trill interval=270 ms, proportion of missing pulses=0·094 for each stimulus). Second, preferences based on the inter-trill interval of male song were tested; females were presented with a choice between male song with inter-trill intervals of 150 and 450 ms (pulses per trill=32, proportion of missing pulses=0·094 for each stimulus). And third, preferences based on the proportion of missing pulses in male song were tested; females were presented with a choice between male song with 0 and 15% missing pulses (pulses per trill=32, inter-trill interval=270 ms for each stimulus). The number of pulses per trill, inter-trill interval, and proportion of missing pulses in the test stimuli lie within the natural range of variation in male song (Souroukis et al. 1992). Each female was tested with all three stimulus pairs. The two stimuli of a stimulus pair were broadcast antiphonally from opposite sides of the test chamber. Each stimulus was 84·6 dB SPL (re 2#10 "5 N/m2) in amplitude at the position of the cricket. The first stimulus was broadcast from a speaker for 30 s, followed by a 2-s no-stimulus period. The second stimulus was then broadcast from the opposite speaker in the same manner. Broadcasts were continued for seven repetitions of each stimulus, yielding 14 vector scores. The vector scores were averaged separately for each stimulus for each female, giving an average vector score for each stimulus for each female. Each female was tested only once with a given pair of stimuli. Between females the stimuli were switched between the speakers to control for any effect of side biases on the results. To determine whether females preferred one stimulus over its alternative, pairs of average vector scores were compared with a two-tailed paired t-test. There are a number of reasons to believe that calling song preferences in G. integer are mating
Wagner et al.: Phenotypic variation in female preferences preferences. First, phonotaxis and receptivity to mating arise at the same time ontogenetically in this species (Solymar & Cade 1990b). Second, females deprived of males become more phonotactic, while females housed with males become less phonotactic (Cade 1979). And third, mating usually follows phonotactically mediated movement towards a male (personal observation). Thus there is an association between movement towards a calling male and mating. Preference Functions To quantify female preference functions based on the number of pulses per trill, the inter-trill interval and the proportion of missing pulses in male calling song, we tested female responses to three sets of eight stimuli. Stimuli were constructed as described above. Within each stimulus set, stimuli were presented sequentially in a randomized order to females, and the speaker through which each stimulus was broadcast was randomly selected. Each stimulus was broadcast for 120 s, and stimulus presentations were separated by 30 s. Each stimulus was 84·6 dB SPL (re 2#10 "5 N/m2) in amplitude at the position of the cricket. The strength of a female’s preference for a given song type (i.e. stimulus) was estimated by the net vector score resulting from a female’s movement during the stimulus presentation. First, we examined preference functions based on the number of pulses per trill in male song. Females were presented with eight male songs that varied only in the number of pulses per trill: 10, 20, 30, 40, 50, 60, 70 and 80 pulses per trill (inter-trill interval=270 ms, proportion of missing pulses=0·094 for each stimulus). Second, we examined preference functions based on the intertrill interval in male song. Females were presented with eight male songs that varied only in inter-trill interval: 100, 150, 200, 250, 300, 350, 400 and 450 ms (pulses per trill=32, proportion of missing pulses=0·094 for each stimulus). And third, we examined preference functions based on the proportion of missing pulses in male song. Females were presented with eight male songs that varied only in the proportion of missing pulses: 0·00, 0·04, 0·08, 0·12, 0·16, 0·20, 0·24 and 0·28 (pulses per trill=32, inter-trill interval=270 ms for each stimulus). All values fall within the natural range of variation for G. integer calling song (Souroukis et al. 1992).
1273
Based on female responses to each stimulus within a stimulus set, we examined populationlevel preference functions based on the number of pulses per trill, the inter-trill interval and the proportion of missing pulses in male calling song using repeated-measures analysis of variance. Specifically, we tested the hypothesis that female responses (i.e. vector scores) varied between stimuli within each stimulus set. Twenty-five females were tested at least once with each of the three stimulus sets. Nineteen of these females were tested twice, with the two testing periods separated by a minimum of 3 days. For females tested on more than 1 day, female responses to the stimuli were averaged across days. We also examined the extent of variation within and between females in their preference functions. For the 19 females tested on 2 different days, we estimated the preference functions of each female for each male trait separately for each of the two tests by regressing vector score on stimulus value. We used the resulting univariate regression coefficients to describe the direction and strength of each preference function for each female: a positive preference function (i.e. regression slopes) indicates an increase in the strength of the preference with an increase in stimulus value, a negative preference function indicates a decrease in the strength of the preference with an increase in stimulus value, and a preference function of zero indicates no variation in the strength of the preference with stimulus value. Females that did not move in response to broadcast song were discarded. Consequently, preference functions of zero did not derive in whole or in part from unresponsive females. This method assumes that female preference functions can be adequately described by a linear function. Variation in univariate regression coefficients was then compared between females using analysis of variance, where groups represent females. In addition, based on the variance components from the analysis of variance, we estimated the repeatability of preference functions within females (Lessells & Boag 1987). Repeatability represents the intraclass correlation coefficient, and it provides a measure of the consistency of a trait within individuals. Finally, we calculated correlations between female preference functions using Pearson product–moment correlations. For females tested on more than 1 day, preference functions (i.e. regression coefficients) were averaged across days. All statistics were calculated
Animal Behaviour, 49, 5
1274
Table I. Population-level mating preferences of female G. integer in two-speaker choice tests (df=19 for all statistical comparisons)
Test
Stimuli
Mean preference score (&)
Number of pulses per trill Inter-trill interval (ms) Proportion of missing pulses
15 70 150 450 0·00 0·15
2491 4909 5573 2604 3581 4201
6000 (a) 5000 4000
Paired t-test
3000 2000
(1060) (893) (1230) (621) (1419) (913)
2·17*
1000
2·73*
0 –1000
0·43
–2000 10 20 30 40 50 60 70 80 Number of pulses per trill
Mean preference score measures the relative amount of movement by females towards each stimulus. *P<0·05.
4000 (b)
RESULTS Relative Mating Preferences Results of the two-stimulus choice tests showed that females moved more towards the male song with more pulses per trill and the male song with the shorter inter-trill interval but did not differ significantly in their movement towards song with few and many missing pulses (Table I). These results suggest that, as a population, females prefer male calling song with more pulses per trill and shorter inter-trill intervals. However, females do not appear to discriminate between male calling song based on the proportion of missing pulses.
Preference score
using the SYSTAT statistical package (Wilkinson 1989).
3000
2000
1000
0
5000
100 150 200 250 300 350 400 450 Inter-trill interval (ms)
(c)
4000
3000
Population-level Preference Functions Females sequentially presented eight randomly ordered songs that varied only in the number of pulses per trill showed significant variation in the amount of movement towards a song (i.e. vector score; repeated-measures ANOVA: F7,168 =4·69, P<0·001). The strength of the female preference appeared to increase with the number of pulses per trill in the song (Fig. 3a); that is, females moved more towards songs with many pulses per trill than songs with few pulses per trill. There was no significant variation in female movement towards eight randomly ordered songs
2000
1000 0 0·04 0·08 0·12 0·16 0·20 0·24 0·28 Proportion of missing pulses Figure 3. Average preference functions of female G. integer based on (a) the number of pulses per trill, (b) the inter-trill interval and (c) the proportion of missing pulses in male calling song (N=25 for each stimulus). Average (&) preference scores for each stimulus are shown.
Wagner et al.: Phenotypic variation in female preferences
1275
Table II. Variation in preference functions between and within female G. integer (N=19) Preference Number of pulses per trill Inter-trill interval Proportion of missing pulses
df
Source
Mean squares
F-ratio
n0
17 18 17 18 17 18
Between females Within females Between females Within females Between females Within females
1995·53 673·84 49·58 12·78 38 107·22 39 826·21
2·96*
2·0
0·50
3·88**
2·0
0·59
0·53
2·0
"0·02
Repeatability
Preference functions are univariate regression coefficients from the regression of vector score on stimulus value. *P<0·05; **P<0·01.
that varied only in inter-trill interval (repeatedmeasures ANOVA: F7,168 =1·01, P=0·424). The average population preference function was essentially flat over the range of inter-trill intervals tested (Fig. 3b). There was also no significant variation in female movement towards eight randomly ordered songs that varied only in the proportion of missing pulses (repeated-measures ANOVA: F7,168 =0·97, P=0·455). The average preference function was essentially flat over the range of missing pulses tested (Fig. 3c). Phenotypic Variation in Preference Functions The preference functions of 19 females were quantified on 2 days, separated by a minimum period of 3 days (X&=3·42&0·22, maximum=4). Multiple measures of female preference functions allowed evaluation of between- and within-female variation in preference functions. Preference functions were derived by regressing the amount of movement by a female towards a stimulus (i.e. vector score) on stimulus value. In this analysis, then, preference functions are estimated by univariate regression coefficients. First, we examined variation between females in their preference functions (i.e. regression coefficients). There was significant variation between females in their preference functions based on the number of pulses per trill and the inter-trill interval in male calling song (Table II). In contrast, there was no significant variation between females in their preference functions based on the proportion of missing pulses in male calling song (Table II). Female preference functions did not significantly differ between the two trial periods (pulses per trill: paired t=0·19, df=18, P=0·86; inter-trill
interval: paired t=1·00, df=18, P=0·34; proportion of missing pulses: paired t=0·27, df=18, P=0·79). Consequently, there was no directional change in female preference functions between trial periods. Second, we examined the frequency distributions of female preference functions (i.e. regression coefficients). Preference functions were averaged within females so that each female was represented only once in each frequency distribution. In addition, data from females tested only once with each stimulus (N=6) were included. Most female preference functions based on the number of pulses per trill in male calling song were positive (Fig. 4a; X&=29·9&7·7). Consequently, while there was consistent betweenfemale variation in preference functions based on the number of pulses per trill, females appeared to vary in how strongly they preferred more pulses per trill, not in whether they preferred more or fewer pulses per trill. In contrast, while there was a slight bias towards negative preference functions based on the inter-trill interval in male calling song (Fig. 4b; X&= "0·8&1·3), a substantial proportion of females exhibited positive preference functions. Thus there appeared to be between-female variation not only in the strength of their preferences based on inter-trill interval, but also in whether they preferred shorter or longer inter-trill intervals. Female preference functions based on the proportion of missing pulses in male calling song varied substantially in both magnitude and sign (Fig. 4c; X&=12·2&52·5), but because there was no significant variation between females in preference functions based on this trait, much of this variation can be accounted for by withinfemale variation in preference functions (see below).
Animal Behaviour, 49, 5
1276 10
Table III. Phenotypic correlations between female preference functions (i.e. regression coefficients from the regression of vector score on stimulus value) in G. integer (N=25)
(a)
8 6
Number of pulses per trill
4 2 0 –40 –20 0
20 40 60 80 100 120 140 160
Number of pulses per trill Inter-trill interval Proportion of missing pulses
Inter-trill interval
Proportion of missing pulses
— "0·482*
—
— "0·415* 0·056
10 Number of females
(b)
*P<0·05.
8 6 4 2 0 –20 –16 –12 –8 –4
0
4
8
12 16 20
10 (c) 8
based on both the number of pulses per trill (0·50) and the inter-trill interval (0·59) in male calling song were relatively high, while the repeatability of female preference functions based on the proportion of missing pulses in male calling song was quite low ("0·02; negative repeatabilities result from F-ratios<1: Lessells & Boag 1987). Preference functions based on the number of pulses per trill and the inter-trill interval in male calling song thus not only differ between females, but also appear to be relatively stable within females.
6
Phenotypic Correlations between Preference Functions
4 2 0
–800 –400 0 400 800 200 600 1000 –1000 –600 –200 Preference function
Figure 4. Frequency distributions of the preference functions of individual female G. integer based on (a) the number of pulses per trill, (b) the inter-trill interval and (c) the proportion of missing pulses in male calling song. Female preference functions are represented as the univariate regression coefficient from the regression of female movement towards a stimulus on stimulus value. The more positive the slope, the more strongly females prefer greater stimulus values. The more negative the slope, the more strongly females prefer lesser stimulus values. Slopes of zero indicate no discrimination between stimulus values.
The pattern of phenotypic correlation between female preference functions (i.e. regression coefficients) was examined. Female preference functions based on the number of pulses per trill and the inter-trill interval in male calling song were negatively correlated (Table III). Similarly, female preference functions based on the intertrill interval and the proportion of missing pulses in male calling song were negatively correlated (Table III). There was no significant correlation between female preference functions based on the number of pulses per trill and the proportion of missing pulses in male calling song (Table III). DISCUSSION
Third, we quantified the repeatability of preference fucntions within females (Table II). The repeatabilities of female preference functions
Relative Mating Preferences As a population, female G. integer preferred male calling song with more pulses per trill and
Wagner et al.: Phenotypic variation in female preferences shorter inter-trill intervals in two-speaker choice tests. These results suggest that female choice will favour males that produce calling song with more pulses per trill (i.e. longer trills) and shorter intertrill intervals (i.e. higher trill rates). Female choice has been shown to favour longer calling bouts, longer chirp and trill durations, and higher chirp rates in other species of cricket (Popov & Shuvalov 1977; Pollack & Hoy 1981; Hedrick 1986; Simmons 1988; Stout & McGhee 1988). In some animals females select mates based on song characters correlated with size or age (e.g. Thornhill & Alcock 1983; Ryan 1985; Zuk 1987; Simmons 1988; Ryan et al. 1990). However, neither the number of pulses per trill nor the inter-trill interval in male calling song is correlated with male size or age in G. integer (Souroukis et al. 1992). Female choice based on these song characters thus will not result in mating with larger or older males. Although missing pulses in a song may result from wearing of the stridulatory file with age in some crickets (e.g. Stiedl et al. 1991), the proportion of missing pulses in male calling song does not appear to vary with age (or size) in G. integer (Souroukis et al. 1992). Consequently, even if females preferred more missing pulses in male calling song, such preferences would not result in mating with older (or larger) males. In many animals, including many crickets, females prefer to mate with males that produce more energetic signals (Ryan & Keddy-Hector 1992). Because calling songs with more pulses per trill and shorter inter-trill intervals contain greater acoustic energy, female G. integer also appear to prefer some song characters correlated with greater energetic content. The reasons females prefer these song characters is currently unclear. Possibilities include biases in the sensory or information processing system of females that evolved in another context (e.g. Basolo 1990a; Ryan 1990; Kirkpatrick & Ryan 1991; Endler 1992), a direct benefit associated with mating with males producing these song types (e.g. Thornhill 1983; Price 1984; Ryan 1985), the greater localizability of these song types, and a heritable fitness benefit passed to offspring that is correlated with these song types (e.g. Hamilton & Zuk 1982; Pomiankowksi 1987; Grafen 1990). The absence of female choice based on the proportion of missing pulses in male calling song, however, suggests that the energetic content of
1277
male song is not, by itself, sufficient to explain the preferences of female G. integer; the lower the proportion of missing pulses in a song, the higher the energetic content of the song. For example, a song with 28% missing pulses has 28% less acoustic energy than a song with 0% missing pulses. Population-level Preference Functions A female preference function describes variation in the strength of a preference for a male trait with variation in the expression of the trait. The average preference function of female G. integer based on the number of pulses per trill in male calling song was significant and positive; the strength of a female’s preference generally increased with the number of pulses per trill in a song. This result agrees with that obtained in the two-speaker choice tests. It is unclear whether the average preference function based on the number of pulses per trill is linear or non-linear. Regardless of the shape of the preference function, however, female choice will favour males with more pulses per trill in their calling song. If female preference functions are linear, however, sexual selection by female choice may be relatively constant across the natural range of male phenotypes. In contrast, non-linear preference functions (e.g. Basolo 1990b) will result in variation in the strength of sexual selection over the range of male phenotypes; males with phenotypes in the range where female preference functions are steeper will be under stronger sexual selection than males with phenotypes in the range where female preference functions are shallower. There was no significant variation in the strength of female preferences with variation in the inter-trill interval in male calling song, although there was a weak, non-significant negative relationship. This result contrasts markedly with the preference for shorter inter-trill intervals shown by females in the two-speaker choice tests. In part, this difference may reflect the difference between experimental designs that test mate choice and those that test mating preferences. Mate choice can be defined as differential mating by females as a result of the possession of mating preferences (Heisler et al. 1987). When two traits are presented in isolation to females, females might show a 10% higher response to one of the traits. This does not, however, necessarily suggest
1278
Animal Behaviour, 49, 5
that when the traits are simultaneously presented to females that they will show a 10% greater response to the same trait. Mate choice is generally an all-or-none response, with a female mating with one of a number of available males. Thus, while the strength of a female’s inherent responsiveness to two traits may differ by 10%, when presented with a choice they may only orient to the more preferred trait, or spend the vast majority of their time orienting to that trait. Choice tests, then, might amplify female responses to different stimuli, and we might expect preferences to be more distinct in simultaneous choice tests than in tests where stimuli are presented individually. Amplification of a weak preference can potentially account for the disparity between the preference for shorter trill intervals measured in the two-speaker choice tests and the results of the preference function measurements. There was no significant variation in the strength of female preferences with variation in the proportion of missing pulses in male calling song. This result agrees with those obtained in the two-speaker choice tests. Consequently, sexual selection by female choice does not appear to act on the proportion of missing pulses in male calling song. Phenotypic Variation in Female Mating Preferences There was significant variation between females in their preference functions based on both the number of pulses per trill and the inter-trill interval in male calling song. There was, however, no significant variation between females in their preference functions based on the proportion of missing pulses in male calling song. Only a handful of studies have shown phenotypic variation in female mating preferences within a population. In Colias eurytheme butterflies, females of different colour morphs show different pheromone preferences (Sappington & Taylor 1990). In cricket frogs, Acris crepitans, a female’s preference for male call dominant frequency is partially size dependent; larger females prefer lower frequencies than do smaller females (Ryan et al. 1992). In green tree frogs, Hyla cinerea, females differ in the degree to which they discriminate between conspecific and hybrid song (Gerhardt 1974, 1992), and in barn swallows, Hirundo rustica, females consistently differ in the tail length of their mates (Møller
1994). In contrast, Ritchie (1992) failed to detect differences among female bushcrickets, Ephippiger ephippiger, in their mating preferences. In part, demonstrations of phenotypic variation in mating preferences have been hampered by the use of two-stimulus, dichotomous choice designs in which female responses are scored as either yes or no for each stimulus within a stimulus pair. When females differ in the strength of their preferences, but not in direction of their preferences, such choice designs will often be too insensitive to detect variation among females. More appropriate experimental designs may reveal that phenotypic variation in female preferences is widespread. The effect of individual variation in female preference functions on the nature of sexual selection on male traits will depend, in part, on the frequency distribution of female preference functions. If most females have preference functions of similar sign (i.e. if most females prefer one extreme of the male trait distribution), sexual selection on male traits will be directional. However, if some females prefer one extreme of the male trait distribution while other females prefer the opposite extreme, sexual selection on male traits may be frequency dependent (e.g. O’Donald & Majerus 1984). Assuming a constant distribution of female preferences, the fitness consequences of a male trait variant will then depend on the frequency of males possessing that variant relative to the frequency of females preferring that variant. Because female G. integer appeared to differ primarily in the strength of their preferences for more pulses per trill in male calling song, sexual selection by female choice should result in directional sexual selection on the number of pulses per trill. In contrast, because females appeared to differ in whether they preferred longer or shorter inter-trill intervals in male calling song, sexual selection by female choice may result in frequency-dependent sexual selection on inter-trill interval. The repeatability of female preference functions based on the number of pulses per trill and the inter-trill interval in male calling song was relatively high (0·50 and 0·59, respectively). There is thus not only significant variation between female G. integer in their preference functions, but also a relatively high consistency within females in their preference functions. Møller (1994) also found high repeatabilities for female tail length preferences in barn swallows. In contrast, Boake
Wagner et al.: Phenotypic variation in female preferences (1989) estimated a zero repeatability for the pheromone preferences of female flour beetles, Tribolium castaneum. There were significant correlations between the preference functions of females (i.e. between regressions slopes). First, preference functions based on the number of pulses per trill and the inter-trill interval in male calling song were negatively correlated. Consequently, it appears that females that were relatively nondiscriminatory based on the number of pulses per trill tended to prefer longer inter-trill intervals, whereas females that strongly preferred more pulses per trill tended to strongly prefer shorter inter-trill intervals. Second, preference functions based on the inter-trill interval and proportion of missing pulses in male calling song were negatively correlated. Consequently, it appears that females that strongly preferred shorter inter-trill intervals tended to prefer a high proportion of missing pulses, whereas females that strongly preferred longer inter-trill intervals tended to prefer a smaller proportion of missing pulses. It is unknown whether male traits vary in their importance to females. It is possible that some traits are ignored by females during mate choice under natural conditions, even though they discriminate between variants of the trait in controlled choice tests (e.g. Wagner & Sullivan 1995). However, female choice on one male trait clearly has the potential to result in correlated sexual selection on other traits. For example, sexual selection by female choice on the number of pulses per trill in male calling song may result in correlated selection on the inter-trill interval in male calling song. The Evolution of Cricket Mating Preferences? Phenotypic variation in female mating preferences is a necessary condition for selection on mating preferences. Because so few studies have documented such variation, experimental and correlational studies of the factors influencing the evolution of female preferences have been nearly impossible. Instead, most studies on the evolution of female preferences have relied either on comparative data or on demonstrating the adaptive significance of population-level mating preferences (e.g. Thornhill & Alcock 1983; Basolo 1990a; Houde & Endler 1990; Kirkpatrick & Ryan 1991). The existence of variation between
1279
females in their mating preferences, however, allows quantitative evaluation of the fitness consequences of different female preferences. Because there are consistent differences between female G. integer in some of their preference functions, selection can potentially act on female preferences in this species. In addition, phenotypic correlations between female preference functions can result in indirect selection on preference functions correlated with the preference function under direct selection. For example, direct selection for stronger preferences for more pulses per trill in male calling song will result in indirect selection for stronger preferences for shorter inter-trill intervals. Thus, G. integer provide an excellent model for examining selection on mating preferences. For selection to result in the evolution of female mating preferences, there must be genetic variation in female preference functions. Only a few studies have examined genetic variation in female preferences (e.g. Majerus et al. 1982; Kearns et al. 1992; Charalambous et al. 1994). It is currently impossible to determine whether phenotypic variation in the preference functions of female G. integer derives, in part, from genetic variation. However, because the repeatability of preference functions within females is high, there is the potential for genetic variation in female preference functions, and thus for selection to result in evolutionary changes in female preferences. The existence of genetic variation in female preference functions in G. integer remains to be examined.
ACKNOWLEDGMENTS We thank A. L. Basolo, J. A. Endler, M. J. Ryan and two anonymous referees for their criticism of the manuscript. We also thank R. Warner and his laboratory group for their feedback on the results. The Department of Technical Services at Brock University, especially Tom MacDonald, invested considerable time and effort in the development of the PC Kugel; without their perseverance, the research would have been impossible. We also thank J. A. Doherty for providing details on the construction of his MacKugel. Financial support was provided by a grant from the Natural Sciences and Engineering Research Council of Canada (grant no. A6174 to W.H.C.) and by NATO and NSF postdoctoral fellowships (to W.E.W.).
Animal Behaviour, 49, 5
1280 REFERENCES
Alexander, R. D. 1961. Aggressiveness, territoriality, and sexual behavior in field crickets (Orthoptera: Gryllidae). Behaviour, 17, 130–223. Basolo, A. L. 1990a. Female preference predates the evolution of the sword in swordtail fish. Science, 250, 808–810. Basolo, A. L. 1990b. Female preference for male sword length in the green swordtail, Xiphophorus helleri (Pisces: Poeciliidae). Anim. Behav., 40, 332–338. Boake, C. R. 1989. Repeatability: its role in evolutionary studies of mating behavior. Evol. Ecol., 3, 173–182. Bradbury, J. W. & Andersson, M. A. (Eds) 1987. Sexual Selection: Testing the Alternatives. Chichester: John Wiley. Cade, W. H. 1975. Acoustically orienting parasitoids: fly phonotaxis to cricket song. Science, 190, 1312–1313. Cade, W. H. 1979. Age of first mating in field crickets, Gryllus integer (Orthoptera: Gryllidae). Fla Entomol., 73, 193–195. Cade, W. H. 1981. Alternative male strategies: genetic differences in crickets. Science, 212, 563–564. Cade, W. H. & Cade, E. S. 1992. Male mating success, calling and searching behaviour at high and low densities in the field cricket, Gryllus integer. Anim. Behav., 43, 49–56. Cade, W. H. & Wyatt, D. R. 1984. Factors affecting calling behaviour in field crickets, Teleogryllus and Gryllus (age, weight, density, and parasites). Behaviour, 88, 61–75. Charalambous, M., Butlin, R. K. & Hewitt, G. M. 1994. Genetic variation in male song and female song preference in the grasshopper Chorthippus brunneus (Orthoptera: Acrididae). Anim. Behav., 47, 399–411. Doherty, J. A. & Pires, A. 1987. A new microcomputerbased method for measuring walking phonotaxis in field crickets (Gryllidae). J. exp. Biol., 130, 425–432. Endler, J. A. 1986. Natural Selection in the Wild. Princeton, New Jersey: Princeton University Press. Endler, J. A. 1992. Signals, signal conditions and the direction of evolution. Am. Nat. Suppl., 139, 125–153. Falconer, D. S. 1981. Introduction to Quantitative Genetics. New York: Longman. Gerhardt, H. C. 1974. Vocalisations of some hybrid treefrogs: acoustic and behavioural analyses. Behaviour, 49, 130–151. Gerhardt, H. C. 1992. Female mate choice in treefrogs: static and dynamic acoustic criteria. Anim. Behav., 42, 615–635. Grafen, A. 1990. Sexual selection unhandicapped by the Fisher process. J. theor. Biol., 144, 473–516. Hamilton, W. D. & Zuk, M. 1982. Heritable true fitness and bright birds: a role for parasites? Science, 218, 384–387. Hedrick, A. V. 1986. Female preference for male calling bout duration in a field cricket. Behav. Ecol. Sociobiol., 19, 73–77. Heisler, L., Andersson, M. B., Arnold, S. J., Boake, C. R., Borgia, G., Hausfater, G., Kirkpatrick, M., Lande, R., Maynard Smith, J., O’Donald, P., Thornhill, A. R. & Weissing, F. J. 1987. The evolution
of mating preferences and sexually selected traits: group report. In: Sexual Selection: Testing the Alternatives (Ed. by J. W. Bradbury & M. B. Andersson), pp. 96–118. Chichester: John Wiley. Houde, A. E. & Endler, J. A. 1990. Correlated evolution of female mating preferences and male color patterns in the guppy Poecilia reticulata. Science, 248, 1405–1408. Kearns, P. W. E., Tomlinson, I. P. M., Veltman, C. J. & O’Donald, P. 1992. Non-random mating in Adalia bipunctata (the two-spot ladybird). II. Further tests for female mating preference. Heredity, 68, 385–389. Kirkpatrick, M. 1987. Sexual selection by female choice in polygynous animals. A. Rev. Ecol. Syst., 68, 385–389. Kirkpatrick, M. 1987. Sexual selection by female choice in polygynous animals. A. Rev. Ecol. Syst., 18, 43–70. Kirkpatrick, M. & Ryan M. J. 1991. The evolution of mating preferences and the paradox of the lek. Nature, Lond., 350, 33–38. Lande, R. & Arnold, S. J. 1983. The measurement of selection on correlated characters. Evolution, 37, 1210–1226. Lessells, C. M. & Boag, P. T. 1987. Unrepeatable repeatabilities: a common mistake. Auk, 104, 116–121. Majerus, M. E. N., O’Donald, P. & Weir, J. 1982. Female mating preference is genetic. Nature, Lond., 300, 521–523. Møller, A. P. 1994. Repeatability of female choice in a monogamous swallow. Anim. Behav., 47, 643–648. O’Donald, P. & Majerus, M. E. N. 1984. Polymorphism of melanic ladybirds maintained by frequencydependent sexual selection. Biol. J. Linn. Soc., 23, 101–111. Partridge, L. & Hill, W. G. 1984. Mechanisms for frequency-dependent mating success. Biol. J. Linn. Soc., 23, 113–132. Pollack, G. S. & Hoy, R. 1981. Phonotaxis to individual rhythmic components of a complex cricket calling song. J. comp. Physiol., 144, 367–373. Pomiankowski, A. 1987. The ‘handicap principle’ does work sometimes. Proc. R. Soc. Lond. B, 127, 123–145. Popov, A. V. & Shuvalov, V. F. 1977. Phonotactic behavior of crickets. J. comp. Physiol., 119, 111–126. Price, T. D. 1984. Sexual selection on body size, territory and plumage variables in a population of Darwin’s finches. Evolution, 38, 327–341. Ritchie, M. G. 1992. Variation in male song and female preference within a population of Ephippiger ephippiger (Orthoptera: Tettigoniidae). Anim. Behav., 43, 845–855. Ryan, M. J. 1985. The Túngara Frog. Chicago: University of Chicago Press. Ryan, M. J. 1990. Sexual selection, sensory systems, and sensory exploitation. Oxf. Surv. evol. Biol., 7, 157–195. Ryan, M. J., Hews, D. K. & Wagner, W. E., Jr. 1990. Sexual selection on alleles that determine body size in the swordtail Xiphophorus nigrensis. Behav. Ecol. Sociobiol., 26, 231–237. Ryan, M. J. & Keddy-Hector, A. C. 1992. Directional patterns of female mate choice and the role of sensory biases. Am. Nat. Suppl., 139, 4–35.
Wagner et al.: Phenotypic variation in female preferences Ryan, M. J., Perrill, S. A. & Wilczynski, W. 1992. Auditory tuning and call frequency predict population-based mating preferences in the cricket frog, Acris crepitans. Am. Nat., 139, 1370–1383. Ryan, M. J. & Wagner, W. E., Jr. 1987. Mating asymmetries between species: female swordtails prefer to mate with heterospecifics. Science, 236, 595–597. Sakaluk, S. K. & Cade, W. H. 1980. Female mating frequency and progeny production in singly and doubly mated house and field crickets. Can. J. Zool., 58, 404–411. Sappington, T. W. & Taylor, O. R. 1990. Disruptive sexual selection in Colias eurytheme butterflies. Proc. natn. Acad. Sci. U.S.A., 87, 6132–6135. Searcy, W. A. & Andersson, M. 1986. Sexual selection and the evolution of song. A. Rev. Ecol. Syst., 17, 507–534. Simmons, L. W. 1988. The calling song of the field cricket, Gryllus bimaculatus (De Geer): constraints on transmission and its role in intermale competition and female choice. Anim. Behav., 36, 380–394. Solymar, B. & Cade, W. H. 1990a. Heritable variation for female mating frequency in field crickets, Gryllus integer. Behav. Ecol. Sociobiol., 26, 73–76. Solymar, B. & Cade, W. H. 1990b. Age of first mating in field crickets, Gryllus integer (Orthoptera: Gryllidae). Fla Entomol., 73, 193–195. Souroukis, K., Cade, W. H. & Rowell, G. 1992. Factors that possibly influence variation in the calling song of field crickets: temperature, time, and male size, age, and wing morphology. Can. J. Zool., 70, 950–955.
1281
Stiedl, O., Bickmeyer, U. & Kalmring, K. 1991. Tooth impact rate alteration in the song of males of Ephippiger ephippiger Fiebig (Orthoptera, Tettigoniidae) and its consequences for phonotaxic behaviour of females. Bioacoustics, 3, 1–16. Stout, J. F. & McGhee, R. 1988. Attractiveness of the male Acheta domestica calling song to females. II. The relative importance of syllable period, intensity and chirp rate. J. comp. Physiol., 164, 277–287. Sullivan, B. K. & Hinshaw, S. H. 1992. Female choice and selection on male calling behaviour in the grey treefrog, Hyla versicolor. Anim. Behav., 44, 733–744. Thornhill, R. 1983. Cryptic female choice and its implications in the scorpionfly Harpobittacus nigriceps. Am. Nat., 122, 765–788. Thornhill, R. & Alcock, J. 1983. The Evolution of Insect Mating Systems. Cambridge, Massachusetts: Harvard University Press. Wagner, W. E., Jr & Sullivan, B. K. 1995. Sexual selection in the Gulf Coast toad, Bufo valliceps: female choice based on variable characters. Anim. Behav., 49, 305–319. Weissman, D. B., Rentz, D. C. F., Alexander, R. D. & Loher, W. 1980. Field crickets (Gryllus and Acheta) of California and Baja California, Mexico (Orthoptera: Gryllidae: Gryllinae). Trans. Am. entomol. Soc., 106, 327–356. Wilkinson, L. 1989. SYSTAT: The System for Statistics. Evanston, Illinois: SYSTAT. Zuk, M. 1987. Variability in attractiveness of male field crickets (Orthoptera: Gryllidae) to females. Anim. Behav., 35, 1240–1248.
Phenotypic variation in the mating preferences of female field crickets, Gryllus integer WILLIAM E. WAGNER, J*, ANNE-MARIE MURRAY & WILLIAM H. CADE Department of Biological Sciences, Brock University, St Catharines, Ontario L2S 3A1, Canada (Received 21 July 1993; initial acceptance 8 September 1993; final acceptance 24 May 1994; MS. number: 6764)
Abstract. Phenotypic variation in the mating preferences of female field crickets was examined. Males of this species produce a trilled calling song which varies in the number of pulses per trill, the inter-trill interval and the proportion of missing pulses within a trill. As a population, females preferred male calling songs with more pulses per trill and shorter inter-trill intervals in two-speaker choice tests, but did not discriminate between male song that varied in the proportion of missing pulses. Female preference functions were examined by sequentially presenting females with a series of songs that varied in only one parameter. As a population, the strength of the female preference for male calling song appeared to increase with the number of pulses per trill in the song. However, there was no significant variation in the strength of the preference for male calling song with either the inter-trill interval or proportion of missing pulses in the song. There was significant variation between individual females in their preference functions based on the number of pulses per trill and the inter-trill interval in male song, but not based on the proportion of missing pulses in male song. Females appeared to differ in how strongly they preferred more pulses per trill. In contrast, females appeared to differ not only in the strength of their preference based on inter-trill interval, but also in whether they preferred longer or shorter inter-trill intervals. The repeatability of preference functions within females was relatively high for number of pulses per trill (0·50) and inter-trill interval (0·59), but low for proportion of missing pulses ("0·02). Correlations between female preference functions were also examined. Females that strongly preferred more pulses per trill tended to strongly prefer shorter inter-trill intervals. In addition, females that strongly preferred shorter inter-trill intervals tended to prefer a higher proportion of missing pulses. These results suggest that selection can act on female preference functions in field crickets, and that direct selection on one preference function can result in indirect selection on other preference functions.
Female mate choice is known to have an important effect on the evolution of male morphology and behaviour (reviewed by Thornhill & Alcock 1983; Searcy & Andersson 1986; Kirkpatrick 1987). However, while numerous studies have shown that females select mates based on variation in male displays, resulting in sexual selection on these displays, little is known about the detailed structure of female preference functions (i.e. variation in the strength of a female preference with variation in a male display). The structure of female preference functions will have at least two important consequences for the nature *Present address: Nebraska Behavioral Biology Group, School of Biological Sciences, University of Nebraska, Lincoln, NE 68588, U.S.A. 0003–3472/95/051269+13 $08.00/0
?
of sexual selection on male morphology and behaviour. First, the shape of the average female preference function will determine whether female choice results in directional, stabilizing, or disruptive sexual selection on a male trait (e.g Gerhardt 1992); and second, the shape of the average female preference function will, in part, determine the strength of sexual selection by female choice; the steeper the preference function, the stronger should be sexual selection by female choice. The factors that can influence the evolution of female mating preferences have received considerable recent attention (reviewed in Bradbury & Andersson 1987; Kirkpatrick & Ryan 1991). For selection to influence the evolution of female preferences, there must be variation between 1995 The Association for the Study of Animal Behaviour
1269
1270
Animal Behaviour, 49, 5
females in their mating preferences. Few studies, however, have shown that mating preferences differ among females within a population (exceptions include Majerus et al. 1982; Sappington & Taylor 1990; Ryan et al. 1992). Even fewer studies have examined the repeatability of preferences within females (for discussions of repeatability see Lessells & Boag 1987; Boake 1989). Because the repeatability of a trait will set an upper limit on the degree to which variation in the trait derives from heritable differences among individuals (Falconer 1981), an examination of the repeatability of female mating preferences can provide information on the potential for genetic variation in mating preferences (see also Møller 1994). If the repeatability of preferences within females is low, the heritability of preferences must also be low (Falconer 1981). Between-female variation in mating preferences can also have important consequences for the nature of sexual selection on male traits. First, variation between females can reduce the strength of sexual selection; sexual selection will be stronger when females are unanimous in their preferences than when there is variation between females in their preferences. Second, variation in female mating preferences can result in frequencydependent sexual selection on male traits (e.g. O’Donald & Majerus 1984; Partridge & Hill 1984). If some females prefer one extreme of a male trait while other females prefer the opposite extreme, the selective advantage of the trait will depend on the proportion of males exhibiting each extreme relative to the proportion of females preferring each extreme. Selection on mating preferences, as on any trait, will be direct, indirect, or both (e.g. Lande & Arnold 1983; Endler 1986). Indirect selection can act on mating preferences, in part, through the correlation of one preference function with another preference function that is under direct selection. Patterns of phenotypic correlation between preference functions can thus have important consequences for selection on mating preferences. No study that we are aware of, however, has examined phenotypic correlations between female preferences. It is unknown, for example, whether females that exhibit strong preferences based on one male trait also exhibit strong preferences based on other traits, or if females that exhibit strong preferences based on one male trait show weak or no preferences based on other traits.
Inter-trill interval
100 ms
Pulse
Missing pulse
Trill
Figure 1. Sonagram of the calling song of male G. integer.
In this paper we examine phenotypic variation in female calling song preferences in a Texas population of the field cricket, G. integer. Male field crickets generally sing from the ground within or near burrows (Alexander 1961), and females select mates, in part, based on variation in male calling song (e.g. Popov & Shuvalov 1977; Hedrick 1986; Simmons 1988). In G. integer, males produce a trilled calling song, and there is variation between males in the number of pulses per trill, the proportion of missing pulses within a trill and the inter-trill interval (Fig. 1; Souroukis et al. 1992). Previous studies have examined the evolution of male calling and mating behaviour (e.g. Cade 1975, 1981; Cade & Wyatt 1984; Cade & Cade 1992) and female mating frequency (Sakaluk & Cade 1980; Solymar & Cade 1990a). Little is known about female mating preferences based on male calling song. Hedrick (1986) examined the population-level mating preferences of females from a California population of G. integer. However, it appears that the California and Texas field crickets are different species (Weissman et al. 1980) First, we examined the population-level mating preferences of females based on the number of pulses per trill, the inter-trill interval and the proportion of missing pulses in male calling song using a traditional two-stimulus choice design (e.g. Gerhardt 1974; Ryan & Wagner 1987; Wagner & Sullivan 1995). Second, we examined the shapes of the average preference functions of females based on the number of pulses per trill, the inter-trill interval and the proportion of missing pulses in male calling song. Third, we examined variation between and within females in their preference functions. And fourth, we examined phenotypic correlations between female preference functions.
Wagner et al.: Phenotypic variation in female preferences
1271
METHODS Rearing Methods Female mating preferences were tested from 18 January to 6 February 1993. The females tested were either first or second generation offspring of field-caught females from Austin, Texas. Fieldcaught females were transported to the laboratory and housed individually in plastic containers with moist vermiculite for oviposition, water (glass vials plugged with cotton), Purina cat chow, and cardboard egg cartons. Because field-caught females were isolated from males upon capture, all eggs were fertilized by males in the field. First generation offspring were derived from females collected in October 1992. Second generation offspring were derived from females collected in July 1992. For the latter, one female offspring from a field-caught female was mated to a randomly selected male from another family to produce second generation offspring. Offspring were reared in containers with their siblings. Crickets were maintained in an environmental chamber at 30)C, 40% relative humidity, and a 12:12 h light:dark cycle. Within 2 days of their final moult, females were isolated in individual containers. Because females are not receptive to mating until approximately 3–4 days after their final moult (Solymar & Cade 1990b), and because none of the isolated females deposited fertilized eggs in the cotton of the water vials, the only substrate available for oviposition, the test females were known to be virgins. Female preferences were tested within 7–14 days of their final moult. Females are receptive to mating at this time (personal observation). Testing Apparatus Female preferences were tested using a PC Kugel, a device modelled after the MacKugel of Doherty & Pires (1987). This device was similar to a track ball in design, consisting of a hollow plastic sphere (16·2 cm diameter, 34·4 g) supported on a column of air (Fig. 2). The sphere was held in position on the air column by two rollers taken from a computer ‘mouse’. The rollers were level with the horizontal axis of the sphere and perpendicular to each other. A cricket tethered on top of the sphere turned the sphere as it walked towards or away from broadcast song. The displacement of the rollers this movement caused was
Hollow plastic sphere
Sensors
Air column
Output to computer
Air line
Figure 2. Illustration of the PC Kugel used to test female mating preferences in G. integer. See text for details.
monitored by a personal computer. As with a track ball, the direction and velocity of movement of the sphere can be determined by the displacement of each roller. Because the sphere was in contact only with the two rollers, there was little frictional resistance to sphere rotation. In addition, because the sphere was lightweight, there was little inertial resistance to sphere rotation. Thus crickets easily turned the sphere as they walked. The PC Kugel was placed in a circular test chamber (diameter 101·9 cm, height 47·4 cm). The top, bottom and wall of the chamber were lined with foam to improve the localizability of broadcast song. Four speakers mounted on metal posts were placed equidistant around the test chamber. Each speaker was 35·5 cm from the position of the tethered cricket. A cricket was held in position on top of the sphere using a rotating tether supported by a rod overhanging the sphere. The tether consisted of a small piece of balsa wood (approximate dimensions: 4 mm length, 2 mm width, 2 mm height) pierced on the upper surface by a thin 5 cm length of wire. The lower surface of the wood was fixed to the cricket with a drop of hot bee’s wax applied to the dorsal surface of the cricket’s thorax. The free end of the wire rested in a tubular sleeve at the end of an adjustable metal arm extending over the sphere. This arrangement allowed the cricket to turn 360) on the sphere and allowed approximately 5 mm of vertical movement, but restricted the cricket from moving laterally. Consequently, the cricket turned the sphere as it attempted to move in response to broadcast song but remained in position on top of the sphere.
1272
Animal Behaviour, 49, 5
Song was broadcast to a tethered cricket from one of the four speakers. In response, the cricket either failed to respond (i.e. remained motionless) or turned the sphere as it attempted to move towards or away from the speaker from which the song was broadcast. Once per second the direction and velocity of movement by the sphere were determined using a custom-designed computer program. This produced a series of vectors that provided information on both the relative velocity of movement and the angle of movement relative to the speaker during each 1-s sampling period (see Doherty & Pires 1987). To calculate the absolute velocity of movement it would have been necessary to measure the relationship between velocity of sphere rotation and vector length. Such measurements were not made. However, the velocity of sphere rotation is directly proportional to vector length. Consequently, vector lengths provide reliable measures of the relative velocity of sphere rotation during movement by females. A female’s net movement towards or away from a speaker for the entire test period (i.e. the strength of a female’s preference for the broadcast song) was summarized by a ‘vector score’, calculated as Vector score=Ó[cos (vector angle)#vector length] where the angular direction of the active speaker is 0) (see also Doherty & Pires 1987). Movement towards the active speaker results in positive vector scores (cos (0))=1) while movement away from the active speaker results in negative vector scores (cos (180))= "1). In addition, greater movement towards or away from the active speaker results in greater positive or negative vector scores. Movement perpendicular to the active speaker results in vector scores of 0 (cos (90))=0, cos (270))=0). Using this method, the vector score of a female moving directly towards the active speaker will be higher than the vector score of a female moving at the same velocity at a 10) angle with respect to the active speaker. These vector scores are unitless because the absolute relationship between sphere rotation and vector length was not determined (see above). Following a series of tests, the cricket and tether were removed from the test chamber. A hot needle was then used to gently loosen the wax fixing the cricket to the tether. The cricket was not injured by this process and could thus be retested on subsequent days.
Two-speaker Choice Tests To determine whether there are populationlevel mating preferences based on the number of pulses per trill, the inter-trill interval, or the proportion of missing pulses in male calling song, female preferences were tested using three pairs of stimuli. Stimuli were constructed by digitizing a natural pulse and constructing pulse trains with a pulse rate of 70 pulses/s. First, preferences based on the number of pulses per trill in male song were tested; females were presented with a choice between male song of 15 and 70 pulses per trill (inter-trill interval=270 ms, proportion of missing pulses=0·094 for each stimulus). Second, preferences based on the inter-trill interval of male song were tested; females were presented with a choice between male song with inter-trill intervals of 150 and 450 ms (pulses per trill=32, proportion of missing pulses=0·094 for each stimulus). And third, preferences based on the proportion of missing pulses in male song were tested; females were presented with a choice between male song with 0 and 15% missing pulses (pulses per trill=32, inter-trill interval=270 ms for each stimulus). The number of pulses per trill, inter-trill interval, and proportion of missing pulses in the test stimuli lie within the natural range of variation in male song (Souroukis et al. 1992). Each female was tested with all three stimulus pairs. The two stimuli of a stimulus pair were broadcast antiphonally from opposite sides of the test chamber. Each stimulus was 84·6 dB SPL (re 2#10 "5 N/m2) in amplitude at the position of the cricket. The first stimulus was broadcast from a speaker for 30 s, followed by a 2-s no-stimulus period. The second stimulus was then broadcast from the opposite speaker in the same manner. Broadcasts were continued for seven repetitions of each stimulus, yielding 14 vector scores. The vector scores were averaged separately for each stimulus for each female, giving an average vector score for each stimulus for each female. Each female was tested only once with a given pair of stimuli. Between females the stimuli were switched between the speakers to control for any effect of side biases on the results. To determine whether females preferred one stimulus over its alternative, pairs of average vector scores were compared with a two-tailed paired t-test. There are a number of reasons to believe that calling song preferences in G. integer are mating
Wagner et al.: Phenotypic variation in female preferences preferences. First, phonotaxis and receptivity to mating arise at the same time ontogenetically in this species (Solymar & Cade 1990b). Second, females deprived of males become more phonotactic, while females housed with males become less phonotactic (Cade 1979). And third, mating usually follows phonotactically mediated movement towards a male (personal observation). Thus there is an association between movement towards a calling male and mating. Preference Functions To quantify female preference functions based on the number of pulses per trill, the inter-trill interval and the proportion of missing pulses in male calling song, we tested female responses to three sets of eight stimuli. Stimuli were constructed as described above. Within each stimulus set, stimuli were presented sequentially in a randomized order to females, and the speaker through which each stimulus was broadcast was randomly selected. Each stimulus was broadcast for 120 s, and stimulus presentations were separated by 30 s. Each stimulus was 84·6 dB SPL (re 2#10 "5 N/m2) in amplitude at the position of the cricket. The strength of a female’s preference for a given song type (i.e. stimulus) was estimated by the net vector score resulting from a female’s movement during the stimulus presentation. First, we examined preference functions based on the number of pulses per trill in male song. Females were presented with eight male songs that varied only in the number of pulses per trill: 10, 20, 30, 40, 50, 60, 70 and 80 pulses per trill (inter-trill interval=270 ms, proportion of missing pulses=0·094 for each stimulus). Second, we examined preference functions based on the intertrill interval in male song. Females were presented with eight male songs that varied only in inter-trill interval: 100, 150, 200, 250, 300, 350, 400 and 450 ms (pulses per trill=32, proportion of missing pulses=0·094 for each stimulus). And third, we examined preference functions based on the proportion of missing pulses in male song. Females were presented with eight male songs that varied only in the proportion of missing pulses: 0·00, 0·04, 0·08, 0·12, 0·16, 0·20, 0·24 and 0·28 (pulses per trill=32, inter-trill interval=270 ms for each stimulus). All values fall within the natural range of variation for G. integer calling song (Souroukis et al. 1992).
1273
Based on female responses to each stimulus within a stimulus set, we examined populationlevel preference functions based on the number of pulses per trill, the inter-trill interval and the proportion of missing pulses in male calling song using repeated-measures analysis of variance. Specifically, we tested the hypothesis that female responses (i.e. vector scores) varied between stimuli within each stimulus set. Twenty-five females were tested at least once with each of the three stimulus sets. Nineteen of these females were tested twice, with the two testing periods separated by a minimum of 3 days. For females tested on more than 1 day, female responses to the stimuli were averaged across days. We also examined the extent of variation within and between females in their preference functions. For the 19 females tested on 2 different days, we estimated the preference functions of each female for each male trait separately for each of the two tests by regressing vector score on stimulus value. We used the resulting univariate regression coefficients to describe the direction and strength of each preference function for each female: a positive preference function (i.e. regression slopes) indicates an increase in the strength of the preference with an increase in stimulus value, a negative preference function indicates a decrease in the strength of the preference with an increase in stimulus value, and a preference function of zero indicates no variation in the strength of the preference with stimulus value. Females that did not move in response to broadcast song were discarded. Consequently, preference functions of zero did not derive in whole or in part from unresponsive females. This method assumes that female preference functions can be adequately described by a linear function. Variation in univariate regression coefficients was then compared between females using analysis of variance, where groups represent females. In addition, based on the variance components from the analysis of variance, we estimated the repeatability of preference functions within females (Lessells & Boag 1987). Repeatability represents the intraclass correlation coefficient, and it provides a measure of the consistency of a trait within individuals. Finally, we calculated correlations between female preference functions using Pearson product–moment correlations. For females tested on more than 1 day, preference functions (i.e. regression coefficients) were averaged across days. All statistics were calculated
Animal Behaviour, 49, 5
1274
Table I. Population-level mating preferences of female G. integer in two-speaker choice tests (df=19 for all statistical comparisons)
Test
Stimuli
Mean preference score (&)
Number of pulses per trill Inter-trill interval (ms) Proportion of missing pulses
15 70 150 450 0·00 0·15
2491 4909 5573 2604 3581 4201
6000 (a) 5000 4000
Paired t-test
3000 2000
(1060) (893) (1230) (621) (1419) (913)
2·17*
1000
2·73*
0 –1000
0·43
–2000 10 20 30 40 50 60 70 80 Number of pulses per trill
Mean preference score measures the relative amount of movement by females towards each stimulus. *P<0·05.
4000 (b)
RESULTS Relative Mating Preferences Results of the two-stimulus choice tests showed that females moved more towards the male song with more pulses per trill and the male song with the shorter inter-trill interval but did not differ significantly in their movement towards song with few and many missing pulses (Table I). These results suggest that, as a population, females prefer male calling song with more pulses per trill and shorter inter-trill intervals. However, females do not appear to discriminate between male calling song based on the proportion of missing pulses.
Preference score
using the SYSTAT statistical package (Wilkinson 1989).
3000
2000
1000
0
5000
100 150 200 250 300 350 400 450 Inter-trill interval (ms)
(c)
4000
3000
Population-level Preference Functions Females sequentially presented eight randomly ordered songs that varied only in the number of pulses per trill showed significant variation in the amount of movement towards a song (i.e. vector score; repeated-measures ANOVA: F7,168 =4·69, P<0·001). The strength of the female preference appeared to increase with the number of pulses per trill in the song (Fig. 3a); that is, females moved more towards songs with many pulses per trill than songs with few pulses per trill. There was no significant variation in female movement towards eight randomly ordered songs
2000
1000 0 0·04 0·08 0·12 0·16 0·20 0·24 0·28 Proportion of missing pulses Figure 3. Average preference functions of female G. integer based on (a) the number of pulses per trill, (b) the inter-trill interval and (c) the proportion of missing pulses in male calling song (N=25 for each stimulus). Average (&) preference scores for each stimulus are shown.
Wagner et al.: Phenotypic variation in female preferences
1275
Table II. Variation in preference functions between and within female G. integer (N=19) Preference Number of pulses per trill Inter-trill interval Proportion of missing pulses
df
Source
Mean squares
F-ratio
n0
17 18 17 18 17 18
Between females Within females Between females Within females Between females Within females
1995·53 673·84 49·58 12·78 38 107·22 39 826·21
2·96*
2·0
0·50
3·88**
2·0
0·59
0·53
2·0
"0·02
Repeatability
Preference functions are univariate regression coefficients from the regression of vector score on stimulus value. *P<0·05; **P<0·01.
that varied only in inter-trill interval (repeatedmeasures ANOVA: F7,168 =1·01, P=0·424). The average population preference function was essentially flat over the range of inter-trill intervals tested (Fig. 3b). There was also no significant variation in female movement towards eight randomly ordered songs that varied only in the proportion of missing pulses (repeated-measures ANOVA: F7,168 =0·97, P=0·455). The average preference function was essentially flat over the range of missing pulses tested (Fig. 3c). Phenotypic Variation in Preference Functions The preference functions of 19 females were quantified on 2 days, separated by a minimum period of 3 days (X&=3·42&0·22, maximum=4). Multiple measures of female preference functions allowed evaluation of between- and within-female variation in preference functions. Preference functions were derived by regressing the amount of movement by a female towards a stimulus (i.e. vector score) on stimulus value. In this analysis, then, preference functions are estimated by univariate regression coefficients. First, we examined variation between females in their preference functions (i.e. regression coefficients). There was significant variation between females in their preference functions based on the number of pulses per trill and the inter-trill interval in male calling song (Table II). In contrast, there was no significant variation between females in their preference functions based on the proportion of missing pulses in male calling song (Table II). Female preference functions did not significantly differ between the two trial periods (pulses per trill: paired t=0·19, df=18, P=0·86; inter-trill
interval: paired t=1·00, df=18, P=0·34; proportion of missing pulses: paired t=0·27, df=18, P=0·79). Consequently, there was no directional change in female preference functions between trial periods. Second, we examined the frequency distributions of female preference functions (i.e. regression coefficients). Preference functions were averaged within females so that each female was represented only once in each frequency distribution. In addition, data from females tested only once with each stimulus (N=6) were included. Most female preference functions based on the number of pulses per trill in male calling song were positive (Fig. 4a; X&=29·9&7·7). Consequently, while there was consistent betweenfemale variation in preference functions based on the number of pulses per trill, females appeared to vary in how strongly they preferred more pulses per trill, not in whether they preferred more or fewer pulses per trill. In contrast, while there was a slight bias towards negative preference functions based on the inter-trill interval in male calling song (Fig. 4b; X&= "0·8&1·3), a substantial proportion of females exhibited positive preference functions. Thus there appeared to be between-female variation not only in the strength of their preferences based on inter-trill interval, but also in whether they preferred shorter or longer inter-trill intervals. Female preference functions based on the proportion of missing pulses in male calling song varied substantially in both magnitude and sign (Fig. 4c; X&=12·2&52·5), but because there was no significant variation between females in preference functions based on this trait, much of this variation can be accounted for by withinfemale variation in preference functions (see below).
Animal Behaviour, 49, 5
1276 10
Table III. Phenotypic correlations between female preference functions (i.e. regression coefficients from the regression of vector score on stimulus value) in G. integer (N=25)
(a)
8 6
Number of pulses per trill
4 2 0 –40 –20 0
20 40 60 80 100 120 140 160
Number of pulses per trill Inter-trill interval Proportion of missing pulses
Inter-trill interval
Proportion of missing pulses
— "0·482*
—
— "0·415* 0·056
10 Number of females
(b)
*P<0·05.
8 6 4 2 0 –20 –16 –12 –8 –4
0
4
8
12 16 20
10 (c) 8
based on both the number of pulses per trill (0·50) and the inter-trill interval (0·59) in male calling song were relatively high, while the repeatability of female preference functions based on the proportion of missing pulses in male calling song was quite low ("0·02; negative repeatabilities result from F-ratios<1: Lessells & Boag 1987). Preference functions based on the number of pulses per trill and the inter-trill interval in male calling song thus not only differ between females, but also appear to be relatively stable within females.
6
Phenotypic Correlations between Preference Functions
4 2 0
–800 –400 0 400 800 200 600 1000 –1000 –600 –200 Preference function
Figure 4. Frequency distributions of the preference functions of individual female G. integer based on (a) the number of pulses per trill, (b) the inter-trill interval and (c) the proportion of missing pulses in male calling song. Female preference functions are represented as the univariate regression coefficient from the regression of female movement towards a stimulus on stimulus value. The more positive the slope, the more strongly females prefer greater stimulus values. The more negative the slope, the more strongly females prefer lesser stimulus values. Slopes of zero indicate no discrimination between stimulus values.
The pattern of phenotypic correlation between female preference functions (i.e. regression coefficients) was examined. Female preference functions based on the number of pulses per trill and the inter-trill interval in male calling song were negatively correlated (Table III). Similarly, female preference functions based on the intertrill interval and the proportion of missing pulses in male calling song were negatively correlated (Table III). There was no significant correlation between female preference functions based on the number of pulses per trill and the proportion of missing pulses in male calling song (Table III). DISCUSSION
Third, we quantified the repeatability of preference fucntions within females (Table II). The repeatabilities of female preference functions
Relative Mating Preferences As a population, female G. integer preferred male calling song with more pulses per trill and
Wagner et al.: Phenotypic variation in female preferences shorter inter-trill intervals in two-speaker choice tests. These results suggest that female choice will favour males that produce calling song with more pulses per trill (i.e. longer trills) and shorter intertrill intervals (i.e. higher trill rates). Female choice has been shown to favour longer calling bouts, longer chirp and trill durations, and higher chirp rates in other species of cricket (Popov & Shuvalov 1977; Pollack & Hoy 1981; Hedrick 1986; Simmons 1988; Stout & McGhee 1988). In some animals females select mates based on song characters correlated with size or age (e.g. Thornhill & Alcock 1983; Ryan 1985; Zuk 1987; Simmons 1988; Ryan et al. 1990). However, neither the number of pulses per trill nor the inter-trill interval in male calling song is correlated with male size or age in G. integer (Souroukis et al. 1992). Female choice based on these song characters thus will not result in mating with larger or older males. Although missing pulses in a song may result from wearing of the stridulatory file with age in some crickets (e.g. Stiedl et al. 1991), the proportion of missing pulses in male calling song does not appear to vary with age (or size) in G. integer (Souroukis et al. 1992). Consequently, even if females preferred more missing pulses in male calling song, such preferences would not result in mating with older (or larger) males. In many animals, including many crickets, females prefer to mate with males that produce more energetic signals (Ryan & Keddy-Hector 1992). Because calling songs with more pulses per trill and shorter inter-trill intervals contain greater acoustic energy, female G. integer also appear to prefer some song characters correlated with greater energetic content. The reasons females prefer these song characters is currently unclear. Possibilities include biases in the sensory or information processing system of females that evolved in another context (e.g. Basolo 1990a; Ryan 1990; Kirkpatrick & Ryan 1991; Endler 1992), a direct benefit associated with mating with males producing these song types (e.g. Thornhill 1983; Price 1984; Ryan 1985), the greater localizability of these song types, and a heritable fitness benefit passed to offspring that is correlated with these song types (e.g. Hamilton & Zuk 1982; Pomiankowksi 1987; Grafen 1990). The absence of female choice based on the proportion of missing pulses in male calling song, however, suggests that the energetic content of
1277
male song is not, by itself, sufficient to explain the preferences of female G. integer; the lower the proportion of missing pulses in a song, the higher the energetic content of the song. For example, a song with 28% missing pulses has 28% less acoustic energy than a song with 0% missing pulses. Population-level Preference Functions A female preference function describes variation in the strength of a preference for a male trait with variation in the expression of the trait. The average preference function of female G. integer based on the number of pulses per trill in male calling song was significant and positive; the strength of a female’s preference generally increased with the number of pulses per trill in a song. This result agrees with that obtained in the two-speaker choice tests. It is unclear whether the average preference function based on the number of pulses per trill is linear or non-linear. Regardless of the shape of the preference function, however, female choice will favour males with more pulses per trill in their calling song. If female preference functions are linear, however, sexual selection by female choice may be relatively constant across the natural range of male phenotypes. In contrast, non-linear preference functions (e.g. Basolo 1990b) will result in variation in the strength of sexual selection over the range of male phenotypes; males with phenotypes in the range where female preference functions are steeper will be under stronger sexual selection than males with phenotypes in the range where female preference functions are shallower. There was no significant variation in the strength of female preferences with variation in the inter-trill interval in male calling song, although there was a weak, non-significant negative relationship. This result contrasts markedly with the preference for shorter inter-trill intervals shown by females in the two-speaker choice tests. In part, this difference may reflect the difference between experimental designs that test mate choice and those that test mating preferences. Mate choice can be defined as differential mating by females as a result of the possession of mating preferences (Heisler et al. 1987). When two traits are presented in isolation to females, females might show a 10% higher response to one of the traits. This does not, however, necessarily suggest
1278
Animal Behaviour, 49, 5
that when the traits are simultaneously presented to females that they will show a 10% greater response to the same trait. Mate choice is generally an all-or-none response, with a female mating with one of a number of available males. Thus, while the strength of a female’s inherent responsiveness to two traits may differ by 10%, when presented with a choice they may only orient to the more preferred trait, or spend the vast majority of their time orienting to that trait. Choice tests, then, might amplify female responses to different stimuli, and we might expect preferences to be more distinct in simultaneous choice tests than in tests where stimuli are presented individually. Amplification of a weak preference can potentially account for the disparity between the preference for shorter trill intervals measured in the two-speaker choice tests and the results of the preference function measurements. There was no significant variation in the strength of female preferences with variation in the proportion of missing pulses in male calling song. This result agrees with those obtained in the two-speaker choice tests. Consequently, sexual selection by female choice does not appear to act on the proportion of missing pulses in male calling song. Phenotypic Variation in Female Mating Preferences There was significant variation between females in their preference functions based on both the number of pulses per trill and the inter-trill interval in male calling song. There was, however, no significant variation between females in their preference functions based on the proportion of missing pulses in male calling song. Only a handful of studies have shown phenotypic variation in female mating preferences within a population. In Colias eurytheme butterflies, females of different colour morphs show different pheromone preferences (Sappington & Taylor 1990). In cricket frogs, Acris crepitans, a female’s preference for male call dominant frequency is partially size dependent; larger females prefer lower frequencies than do smaller females (Ryan et al. 1992). In green tree frogs, Hyla cinerea, females differ in the degree to which they discriminate between conspecific and hybrid song (Gerhardt 1974, 1992), and in barn swallows, Hirundo rustica, females consistently differ in the tail length of their mates (Møller
1994). In contrast, Ritchie (1992) failed to detect differences among female bushcrickets, Ephippiger ephippiger, in their mating preferences. In part, demonstrations of phenotypic variation in mating preferences have been hampered by the use of two-stimulus, dichotomous choice designs in which female responses are scored as either yes or no for each stimulus within a stimulus pair. When females differ in the strength of their preferences, but not in direction of their preferences, such choice designs will often be too insensitive to detect variation among females. More appropriate experimental designs may reveal that phenotypic variation in female preferences is widespread. The effect of individual variation in female preference functions on the nature of sexual selection on male traits will depend, in part, on the frequency distribution of female preference functions. If most females have preference functions of similar sign (i.e. if most females prefer one extreme of the male trait distribution), sexual selection on male traits will be directional. However, if some females prefer one extreme of the male trait distribution while other females prefer the opposite extreme, sexual selection on male traits may be frequency dependent (e.g. O’Donald & Majerus 1984). Assuming a constant distribution of female preferences, the fitness consequences of a male trait variant will then depend on the frequency of males possessing that variant relative to the frequency of females preferring that variant. Because female G. integer appeared to differ primarily in the strength of their preferences for more pulses per trill in male calling song, sexual selection by female choice should result in directional sexual selection on the number of pulses per trill. In contrast, because females appeared to differ in whether they preferred longer or shorter inter-trill intervals in male calling song, sexual selection by female choice may result in frequency-dependent sexual selection on inter-trill interval. The repeatability of female preference functions based on the number of pulses per trill and the inter-trill interval in male calling song was relatively high (0·50 and 0·59, respectively). There is thus not only significant variation between female G. integer in their preference functions, but also a relatively high consistency within females in their preference functions. Møller (1994) also found high repeatabilities for female tail length preferences in barn swallows. In contrast, Boake
Wagner et al.: Phenotypic variation in female preferences (1989) estimated a zero repeatability for the pheromone preferences of female flour beetles, Tribolium castaneum. There were significant correlations between the preference functions of females (i.e. between regressions slopes). First, preference functions based on the number of pulses per trill and the inter-trill interval in male calling song were negatively correlated. Consequently, it appears that females that were relatively nondiscriminatory based on the number of pulses per trill tended to prefer longer inter-trill intervals, whereas females that strongly preferred more pulses per trill tended to strongly prefer shorter inter-trill intervals. Second, preference functions based on the inter-trill interval and proportion of missing pulses in male calling song were negatively correlated. Consequently, it appears that females that strongly preferred shorter inter-trill intervals tended to prefer a high proportion of missing pulses, whereas females that strongly preferred longer inter-trill intervals tended to prefer a smaller proportion of missing pulses. It is unknown whether male traits vary in their importance to females. It is possible that some traits are ignored by females during mate choice under natural conditions, even though they discriminate between variants of the trait in controlled choice tests (e.g. Wagner & Sullivan 1995). However, female choice on one male trait clearly has the potential to result in correlated sexual selection on other traits. For example, sexual selection by female choice on the number of pulses per trill in male calling song may result in correlated selection on the inter-trill interval in male calling song. The Evolution of Cricket Mating Preferences? Phenotypic variation in female mating preferences is a necessary condition for selection on mating preferences. Because so few studies have documented such variation, experimental and correlational studies of the factors influencing the evolution of female preferences have been nearly impossible. Instead, most studies on the evolution of female preferences have relied either on comparative data or on demonstrating the adaptive significance of population-level mating preferences (e.g. Thornhill & Alcock 1983; Basolo 1990a; Houde & Endler 1990; Kirkpatrick & Ryan 1991). The existence of variation between
1279
females in their mating preferences, however, allows quantitative evaluation of the fitness consequences of different female preferences. Because there are consistent differences between female G. integer in some of their preference functions, selection can potentially act on female preferences in this species. In addition, phenotypic correlations between female preference functions can result in indirect selection on preference functions correlated with the preference function under direct selection. For example, direct selection for stronger preferences for more pulses per trill in male calling song will result in indirect selection for stronger preferences for shorter inter-trill intervals. Thus, G. integer provide an excellent model for examining selection on mating preferences. For selection to result in the evolution of female mating preferences, there must be genetic variation in female preference functions. Only a few studies have examined genetic variation in female preferences (e.g. Majerus et al. 1982; Kearns et al. 1992; Charalambous et al. 1994). It is currently impossible to determine whether phenotypic variation in the preference functions of female G. integer derives, in part, from genetic variation. However, because the repeatability of preference functions within females is high, there is the potential for genetic variation in female preference functions, and thus for selection to result in evolutionary changes in female preferences. The existence of genetic variation in female preference functions in G. integer remains to be examined.
ACKNOWLEDGMENTS We thank A. L. Basolo, J. A. Endler, M. J. Ryan and two anonymous referees for their criticism of the manuscript. We also thank R. Warner and his laboratory group for their feedback on the results. The Department of Technical Services at Brock University, especially Tom MacDonald, invested considerable time and effort in the development of the PC Kugel; without their perseverance, the research would have been impossible. We also thank J. A. Doherty for providing details on the construction of his MacKugel. Financial support was provided by a grant from the Natural Sciences and Engineering Research Council of Canada (grant no. A6174 to W.H.C.) and by NATO and NSF postdoctoral fellowships (to W.E.W.).
Animal Behaviour, 49, 5
1280 REFERENCES
Alexander, R. D. 1961. Aggressiveness, territoriality, and sexual behavior in field crickets (Orthoptera: Gryllidae). Behaviour, 17, 130–223. Basolo, A. L. 1990a. Female preference predates the evolution of the sword in swordtail fish. Science, 250, 808–810. Basolo, A. L. 1990b. Female preference for male sword length in the green swordtail, Xiphophorus helleri (Pisces: Poeciliidae). Anim. Behav., 40, 332–338. Boake, C. R. 1989. Repeatability: its role in evolutionary studies of mating behavior. Evol. Ecol., 3, 173–182. Bradbury, J. W. & Andersson, M. A. (Eds) 1987. Sexual Selection: Testing the Alternatives. Chichester: John Wiley. Cade, W. H. 1975. Acoustically orienting parasitoids: fly phonotaxis to cricket song. Science, 190, 1312–1313. Cade, W. H. 1979. Age of first mating in field crickets, Gryllus integer (Orthoptera: Gryllidae). Fla Entomol., 73, 193–195. Cade, W. H. 1981. Alternative male strategies: genetic differences in crickets. Science, 212, 563–564. Cade, W. H. & Cade, E. S. 1992. Male mating success, calling and searching behaviour at high and low densities in the field cricket, Gryllus integer. Anim. Behav., 43, 49–56. Cade, W. H. & Wyatt, D. R. 1984. Factors affecting calling behaviour in field crickets, Teleogryllus and Gryllus (age, weight, density, and parasites). Behaviour, 88, 61–75. Charalambous, M., Butlin, R. K. & Hewitt, G. M. 1994. Genetic variation in male song and female song preference in the grasshopper Chorthippus brunneus (Orthoptera: Acrididae). Anim. Behav., 47, 399–411. Doherty, J. A. & Pires, A. 1987. A new microcomputerbased method for measuring walking phonotaxis in field crickets (Gryllidae). J. exp. Biol., 130, 425–432. Endler, J. A. 1986. Natural Selection in the Wild. Princeton, New Jersey: Princeton University Press. Endler, J. A. 1992. Signals, signal conditions and the direction of evolution. Am. Nat. Suppl., 139, 125–153. Falconer, D. S. 1981. Introduction to Quantitative Genetics. New York: Longman. Gerhardt, H. C. 1974. Vocalisations of some hybrid treefrogs: acoustic and behavioural analyses. Behaviour, 49, 130–151. Gerhardt, H. C. 1992. Female mate choice in treefrogs: static and dynamic acoustic criteria. Anim. Behav., 42, 615–635. Grafen, A. 1990. Sexual selection unhandicapped by the Fisher process. J. theor. Biol., 144, 473–516. Hamilton, W. D. & Zuk, M. 1982. Heritable true fitness and bright birds: a role for parasites? Science, 218, 384–387. Hedrick, A. V. 1986. Female preference for male calling bout duration in a field cricket. Behav. Ecol. Sociobiol., 19, 73–77. Heisler, L., Andersson, M. B., Arnold, S. J., Boake, C. R., Borgia, G., Hausfater, G., Kirkpatrick, M., Lande, R., Maynard Smith, J., O’Donald, P., Thornhill, A. R. & Weissing, F. J. 1987. The evolution
of mating preferences and sexually selected traits: group report. In: Sexual Selection: Testing the Alternatives (Ed. by J. W. Bradbury & M. B. Andersson), pp. 96–118. Chichester: John Wiley. Houde, A. E. & Endler, J. A. 1990. Correlated evolution of female mating preferences and male color patterns in the guppy Poecilia reticulata. Science, 248, 1405–1408. Kearns, P. W. E., Tomlinson, I. P. M., Veltman, C. J. & O’Donald, P. 1992. Non-random mating in Adalia bipunctata (the two-spot ladybird). II. Further tests for female mating preference. Heredity, 68, 385–389. Kirkpatrick, M. 1987. Sexual selection by female choice in polygynous animals. A. Rev. Ecol. Syst., 68, 385–389. Kirkpatrick, M. 1987. Sexual selection by female choice in polygynous animals. A. Rev. Ecol. Syst., 18, 43–70. Kirkpatrick, M. & Ryan M. J. 1991. The evolution of mating preferences and the paradox of the lek. Nature, Lond., 350, 33–38. Lande, R. & Arnold, S. J. 1983. The measurement of selection on correlated characters. Evolution, 37, 1210–1226. Lessells, C. M. & Boag, P. T. 1987. Unrepeatable repeatabilities: a common mistake. Auk, 104, 116–121. Majerus, M. E. N., O’Donald, P. & Weir, J. 1982. Female mating preference is genetic. Nature, Lond., 300, 521–523. Møller, A. P. 1994. Repeatability of female choice in a monogamous swallow. Anim. Behav., 47, 643–648. O’Donald, P. & Majerus, M. E. N. 1984. Polymorphism of melanic ladybirds maintained by frequencydependent sexual selection. Biol. J. Linn. Soc., 23, 101–111. Partridge, L. & Hill, W. G. 1984. Mechanisms for frequency-dependent mating success. Biol. J. Linn. Soc., 23, 113–132. Pollack, G. S. & Hoy, R. 1981. Phonotaxis to individual rhythmic components of a complex cricket calling song. J. comp. Physiol., 144, 367–373. Pomiankowski, A. 1987. The ‘handicap principle’ does work sometimes. Proc. R. Soc. Lond. B, 127, 123–145. Popov, A. V. & Shuvalov, V. F. 1977. Phonotactic behavior of crickets. J. comp. Physiol., 119, 111–126. Price, T. D. 1984. Sexual selection on body size, territory and plumage variables in a population of Darwin’s finches. Evolution, 38, 327–341. Ritchie, M. G. 1992. Variation in male song and female preference within a population of Ephippiger ephippiger (Orthoptera: Tettigoniidae). Anim. Behav., 43, 845–855. Ryan, M. J. 1985. The Túngara Frog. Chicago: University of Chicago Press. Ryan, M. J. 1990. Sexual selection, sensory systems, and sensory exploitation. Oxf. Surv. evol. Biol., 7, 157–195. Ryan, M. J., Hews, D. K. & Wagner, W. E., Jr. 1990. Sexual selection on alleles that determine body size in the swordtail Xiphophorus nigrensis. Behav. Ecol. Sociobiol., 26, 231–237. Ryan, M. J. & Keddy-Hector, A. C. 1992. Directional patterns of female mate choice and the role of sensory biases. Am. Nat. Suppl., 139, 4–35.
Wagner et al.: Phenotypic variation in female preferences Ryan, M. J., Perrill, S. A. & Wilczynski, W. 1992. Auditory tuning and call frequency predict population-based mating preferences in the cricket frog, Acris crepitans. Am. Nat., 139, 1370–1383. Ryan, M. J. & Wagner, W. E., Jr. 1987. Mating asymmetries between species: female swordtails prefer to mate with heterospecifics. Science, 236, 595–597. Sakaluk, S. K. & Cade, W. H. 1980. Female mating frequency and progeny production in singly and doubly mated house and field crickets. Can. J. Zool., 58, 404–411. Sappington, T. W. & Taylor, O. R. 1990. Disruptive sexual selection in Colias eurytheme butterflies. Proc. natn. Acad. Sci. U.S.A., 87, 6132–6135. Searcy, W. A. & Andersson, M. 1986. Sexual selection and the evolution of song. A. Rev. Ecol. Syst., 17, 507–534. Simmons, L. W. 1988. The calling song of the field cricket, Gryllus bimaculatus (De Geer): constraints on transmission and its role in intermale competition and female choice. Anim. Behav., 36, 380–394. Solymar, B. & Cade, W. H. 1990a. Heritable variation for female mating frequency in field crickets, Gryllus integer. Behav. Ecol. Sociobiol., 26, 73–76. Solymar, B. & Cade, W. H. 1990b. Age of first mating in field crickets, Gryllus integer (Orthoptera: Gryllidae). Fla Entomol., 73, 193–195. Souroukis, K., Cade, W. H. & Rowell, G. 1992. Factors that possibly influence variation in the calling song of field crickets: temperature, time, and male size, age, and wing morphology. Can. J. Zool., 70, 950–955.
1281
Stiedl, O., Bickmeyer, U. & Kalmring, K. 1991. Tooth impact rate alteration in the song of males of Ephippiger ephippiger Fiebig (Orthoptera, Tettigoniidae) and its consequences for phonotaxic behaviour of females. Bioacoustics, 3, 1–16. Stout, J. F. & McGhee, R. 1988. Attractiveness of the male Acheta domestica calling song to females. II. The relative importance of syllable period, intensity and chirp rate. J. comp. Physiol., 164, 277–287. Sullivan, B. K. & Hinshaw, S. H. 1992. Female choice and selection on male calling behaviour in the grey treefrog, Hyla versicolor. Anim. Behav., 44, 733–744. Thornhill, R. 1983. Cryptic female choice and its implications in the scorpionfly Harpobittacus nigriceps. Am. Nat., 122, 765–788. Thornhill, R. & Alcock, J. 1983. The Evolution of Insect Mating Systems. Cambridge, Massachusetts: Harvard University Press. Wagner, W. E., Jr & Sullivan, B. K. 1995. Sexual selection in the Gulf Coast toad, Bufo valliceps: female choice based on variable characters. Anim. Behav., 49, 305–319. Weissman, D. B., Rentz, D. C. F., Alexander, R. D. & Loher, W. 1980. Field crickets (Gryllus and Acheta) of California and Baja California, Mexico (Orthoptera: Gryllidae: Gryllinae). Trans. Am. entomol. Soc., 106, 327–356. Wilkinson, L. 1989. SYSTAT: The System for Statistics. Evanston, Illinois: SYSTAT. Zuk, M. 1987. Variability in attractiveness of male field crickets (Orthoptera: Gryllidae) to females. Anim. Behav., 35, 1240–1248.