Inter-population variation in male mating behaviours in the sailfin mollie,Poecilia latipinna

Anim. Behav., 1996, 52, 59–71

Inter-population variation in male mating behaviours in the sailfin mollie, Poecilia latipinna MARGARET B. PTACEK & JOSEPH TRAVIS Department of Biological Science, Florida State University (Received 12 May 1994; initial acceptance 1 November 1994; final acceptance 20 October 1995; MS. number: 7003)

Abstract. Male sailfin mollies show size-dependent variation in sexual behaviour. The level of variation between six north Florida populations in rates of condition-dependent behaviours was estimated and whether behavioural variation is ordered with respect to male body size distributions was determined. In five of six populations, courtship display rates increased with male length, supporting previous evidence. Several results were not consistent with those reported elsewhere. Rates of gonopodial thrusting and gonoporal nibbling were not related to male body length. Courtship display rates adjusted for male body size were not ordered with respect to male body size distributions. High adjusted courtship rates were characteristic of populations where males were predominantly large and where they were predominantly small. No consistent pattern of variation existed between populations in the relationships of the three behaviour patterns to one another. Variation between populations in size-dependent behaviour is therefore far more extensive but less patterned than previously reported. These results imply that evolution can adjust the rates of the three behaviour patterns both independently of one another and of body size, and that the direction of behavioural evolution is not likely to be constrained by covariances within any single natural population. ?

studies of this type of intraspecific variation can be used as model systems for large-scale evolutionary processes. This type of intraspecific variation is often shown by the conditional expression of sexual behaviours, the statistical relationship between the rates at which the elements of a behavioural repertoire are expressed and the value of one or more phenotypic traits with which those behaviours are always associated (Ryan et al. 1992; Travis 1994b). Much is known about the existence of such trait associations, such as how territorial behaviour or rates of surreptitious mating attempts vary with male body size (Warner et al. 1975; Gross 1985; Ryan & Causey 1989). Little is known, however, about the extent to which such condition-dependent rates vary between conspecific populations, much less about their genetic bases or potential evolutionary significance (Travis 1994b). Patterns of size-dependent variation in male sexual behaviour in the sailfin mollie represent a potentially useful model system for examining these issues (reviewed by Travis 1994b). Within

Variation exists between conspecific populations in traits associated with fitness, such as trophic structures (James 1991), metabolic rates, acclimatory abilities, thermal and osmotic tolerances (Garland & Adolph 1991) and life-history parameters (Travis 1994a). Similarly wide variation is known for certain types of behaviour patterns, such as those involved with migration (Dingle 1994) and nest construction (Lynch 1994). Such extensive variation between populations raises three questions. (1) How much of this variation is under genetic control? (2) To the extent that there is genetic control, have the observed patterns of variation between populations been produced by natural selection? If so, then such patterns are strong evidence for the power of selection to match local phenotypes to local ecological conditions. (3) Do these patterns of intraspecific variation mirror interspecific variation? If so, then Correspondence: M. B. Ptacek, Department of Biological Science, Florida State University, Tallahassee, FL 32306-2043, U.S.A. (email: MPTACEK@GARNET. ACNS.FSU.EDU). 0003–3472/96/010059+13 $18.00/0

1996 The Association for the Study of Animal Behaviour

?

1996 The Association for the Study of Animal Behaviour

59

60

Animal Behaviour, 52, 1

any single population, larger males show higher rates of courtship displays and lower rates of gonopodial thrusting (forced insemination attempts). Male size, which is largely fixed at the completion of sexual maturation, varies widely within most populations, and the consequent variation in sexual behaviours can range over an order of magnitude between males at the extremes of the size distribution (e.g. 25 mm and 55 mm standard length). Between populations in the southeastern U.S.A., average male length can vary between 17 and 38 mm (Travis & Trexler 1987). Farr et al. (1986) found that males of the same body size taken from three populations in north Florida that differed in their size distributions showed a three-fold range in rates of courtship displays and gonopodial thrusts. This magnitude of variation could not be explained by a variety of hypotheses about social or ontogenetic influences on behaviour rates (Farr & Travis 1989; Travis & Woodward 1989; Sumner et al. 1994). Moreover, this variation was patterned: the rate of courtship displays, adjusted to a common body size, increased, and the adjusted rate of gonopodial thrusts decreased, with decreasing average body size in each population. Thus, males from populations where average body size is small behaved as though they were ‘larger’ than their absolute size compared to males from populations where average body size is large. At least two hypotheses can explain this pattern (Travis 1994b). First, direct selection operates on rates of behaviour independent of body size and will generate population differences in sizedependent behaviour patterns through genetic changes that influence only rates of behaviour. This hypothesis predicts a limited range of differences in behaviour patterns that may or may not show a pattern of directional association with body-size variation between populations. In the alternative hypothesis, direct selection operates on body size in different directions in different populations. Sexual selection could be optimizing (Verrell 1988), however; rates of sexual behaviours could be under selection independent of body-size values. Although genetic changes in behaviour patterns might occur through pleiotropic influences with body size, genes affecting behaviour patterns independently would respond to the optimizing selection to hold behaviour rates constant in the face of changes in body size. The

result would be a statistical association of sizedependent behaviour rates with patterns of bodysize variation: males from populations in which average body size is smaller should show higher rates of courtship displays than males of the same body size from populations in which the average body size is larger. Either hypothesis invokes a prominent role for locality-specific sexual selection, and thus the patterns themselves address the major questions presented above on the efficacy of adaptive evolution. In this paper we present behavioural data for six additional populations. We addressed the variation between these populations in the male size distributions, their range of size-dependent behaviour patterns and the repeatability of the ordered variation between populations in those behaviour patterns. In doing so, we have demonstrated that the patterns originally described are not general.

METHODS Behaviour Patterns in Male Mollies Males show three predominant mating behaviours. In a courtship display, used to elicit female cooperation in mating, the dorsal fin (enlarged in males) is erected and presented to the female (Parzefall 1969), often accompanied by a sigmoid curving of the body and a tilting of the body towards the female (Luckner 1979). A second behaviour, gonopodial thrusting, circumvents female cooperation in mating; the male orients himself behind a female, brings his gonopodium to a forward position and attempts to insert it forcefully into the female’s gonopore. A third behaviour, in which males make nasal or oral contact with the female’s gonopore, is termed gonoporal nibbling. Its function is unclear, but it appears to aid a male in determining a female’s reproductive condition (Farr & Travis 1986; Travis & Woodward 1989). Collections, Measurements and Observations of Behaviour We collected mollies from three populations, Mounds (MD), Pinhook (PH) and Live Oak (LO) (Wakulla County, Florida), in February 1993, and from three additional populations, Steve’s Ditch

Ptacek & Travis: Mollie mating behaviours (SD) (Franklin County, Florida), Fiddler’s Point (FP) and Wakulla Beach (WB) (Wakulla County, Florida), in July 1993. The Mounds and Pinhook populations inhabit brackish impoundments of variable temperature and salinities. The remaining four populations occur in brackish to saline salt marshes. We collected fish by repeatedly towing a 2.8#1.2-m seine across the entire area of the pond or creek being sampled. This method minimizes the potential for collecting a non-random sample of the size distribution of males and females from each population (Travis & Trexler 1987). Size distributions of males and females from each population closely matched the size distributions of fish collected from these localities in previous years (Farr et al. 1986; Travis & Trexler 1987; J. Travis, unpublished data). We collected 40 males and 40 females from each population except at Fiddler’s Point, where we collected 25 males and 20 females, and Wakulla Beach, from which we collected 30 males and 25 females. In the laboratory we maintained the fish from each population separately in 75-litre aquaria; several aquaria were used for each population, and approximately 10 males and 10 females were placed in each aquarium. Stock and observation aquaria were filled with water at 6 ppt salinity and kept at 25)C on a controlled 14:10 h light:dark cycle under fluorescent aquarium lights. Fish were fed several times daily with commercial flaked food, and filters were changed and approximately one-third of the water replaced weekly. Housing conditions were identical to those in all previous behavioural studies performed in our laboratory (Farr et al. 1986; Farr & Travis 1986; Travis & Woodward 1989; Sumner et al. 1994). We kept individual fish for at least three weeks before assaying male behaviour patterns. We measured the standard length of all of the fish, which is the straight-line distance from the tip of the snout or lips to the end of the last vertebra (base of the caudal fin; Trautman 1981). To estimate the pattern of size-dependent behavioural variation, we chose males nonrandomly, using the eight largest males, the eight smallest males and four intermediate-sized males. This choice increases the variance in male body size relative to a random sample and thus should decrease the standard error of the estimated slope of variation in each behavioural trait on body size.

61

This decrease will yield a more precise estimate of the pattern of size dependence. (This procedure exploits our prior knowledge that the relationships are linear on the double logarithmic scale; Draper & Smith 1981.) All males chosen for the behavioural assays were sexually mature. During maturation in male poeciliid fishes, the anal fin gradually fuses to form the gonopodium, the intromittent organ used in internal fertilization. Very young fish cannot be sexed, but most males longer than 25 mm show some level of anal fin metamorphosis and thus are not easily confused with immature females. We removed each test male from his stock aquarium, measured standard length to the nearest millimetre and kept the male in isolation for 24 h before testing. This protocol appears to motivate males of all body sizes in a similar fashion and yields the high rates of behaviour patterns typically observed in natural populations (Travis 1994b). Object females used in the tests were mature, visibly gravid females from the same population as the test male. Object females were not chosen from the same aquarium as the test male, so the test male had no immediate prior experience with the object female. We used gravid, non-receptive females for two reasons. First, female sailfin mollies are receptive for 1–2 days immediately after producing a brood and signal this receptivity to males (Farr et al. 1986; Farr & Travis 1986; Travis & Woodward 1989; Sumner et al. 1994), resulting in increased levels of sexual behaviour by males towards receptive females. Once females have been inseminated, males lower their rates of sexual behaviour, yet still show all three mating behaviours at rates that their body sizes would predict (Sumner et al. 1994). By using only gravid females, we eliminated variation in male behaviour due to variation in female condition. Second, gravid females were used by Farr et al. (1986) as object females in comparing male behavioural profiles, and we could compare our results with the three populations previously tested. To help standardize relative female size between all sizes of males tested and reflect features of natural populations, we also restricted object females to sizes within 10 mm of the test male’s standard length. In nature, females vary in size within each population because of age variation. Small males often encounter females that are

62

Animal Behaviour, 52, 1

larger than they, and large males may court smaller females. Previous behavioural observations have suggested a tendency towards sizeassortative mating, however (Travis 1994b); small males are less likely to approach very large females, and large males are certainly less likely to approach very small females. In addition, males prefer larger females in choice tests, and larger males have stronger preferences for them (see also McPeek 1992). The size restriction used in our study, which was tailored de facto to each population’s size distribution, allowed us to mimic the social conditions for each population and minimized the potential distortion of data that might have occurred if males from different populations responded differently to absolute female body size. We performed behavioural observations on a single male and a single female in a 37.5-litre aquarium. To minimize disturbance to the fish due to the presence of the observer, we covered three sides of the tank with aluminum foil and the front side with one-way film. We placed the test male in the tank and allowed 15 min for acclimation. We then added the object female to the test tank and allowed an additional 15 min for acclimation. We observed males for 10 min and scored three measures of sexual behaviour with a tape recorder: the number of courtship displays, the number of gonopodial thrusts and the number of gonoporal nibbles. Statistical Analyses We transformed male standard length and the rates for each behaviour to natural logs (ln) to obtain linearity in the behaviour rate–body size relationships and to make the estimates of average body size independent of the variance in body size. We used the angular transformation on the proportion of all observed behavioural acts that were courtship displays. We analysed the differences in average body sizes between populations by one-way analysis on the ln-transformed measurements. We examined average behaviour rates in a similar fashion. To examine whether populations differed in the relative variation in body size, we analysed the coefficients of variation on the natural logarithmic scale. This analysis is more appropriate than one that examines the intrademic variance in body size, because different average sizes of populations

may result in different variances (Wilbur & Collins 1973). The analysis of coefficients of variation is complicated by their distributional properties, which preclude exact parametric testing (Sokal & Braumann 1980). To determine whether populations displayed different levels of relative variation, we followed the procedure of Farr et al. (1986) and used the q-statistic of O’Brien (1978). For each ln-transformed datum, we calculated a jackknife pseudovalue (q-value). The average of the pseudovalues for a population is an estimator of the coefficient of variation (provided the original data have been ln-transformed); the pseudovalues are then used in an analysis of variance like any other random variable. In some cases a male did not show one of the three behaviour patterns during the observation period. The typical practice of adding a small number to each observation to eliminate zeros before transformation can bias estimates of the slope between two transformed variables, if the zero values for the behaviour tend towards an extreme of the body-size distribution, as is the case in our data. We chose to add 0.05 to all behavioural observations. This value is small relative to the scale of the measurement, and examination of the regression residuals from several candidates indicated that this value produced the distribution of residuals that best conformed to the assumptions of the general linear model. To test for differences between populations in courtship display rates independently of body-size variation, we performed an analysis of covariance. This analysis could include only the four populations that showed a homogeneous slope in this relationship: Live Oak, Pinhook, Wakulla Beach and Steve’s Ditch. We excluded the Fiddler’s Point and Mounds populations from this analysis because of their significant slope heterogeneity. We extracted the effects of body size first as a covariate and then tested for population effects. Significant and substantial heterogeneity between populations in the slopes of rates of gonopodial thrusting and gonoporal nibbling on body size precluded the use of analyses of covariance for these behaviour patterns on any substantial subset of populations. We used Tukey’s HSD test as a post-hoc comparison of means between the six populations for male standard length and the rate of each behaviour. Comparisons for male standard length were based on averages calculated from all males

Ptacek & Travis: Mollie mating behaviours collected from each population. The comparisons between average behaviour rates were based on the 20 males tested from each population.

RESULTS Variation in Male Body Size Average male standard length varied from about 26 mm at Fiddler’s Point to almost 39 mm at Live Oak (Fig. 1). This range of variation encompasses that found in earlier, more extensive surveys throughout Florida (Travis & Trexler 1987). Populations differed in average male size (F5,209 =25.5, P<0.001), and population differences explained 38% of the variance observed in male standard length. Populations fell into two groups based on the Tukey test (Table I), one of predominantly large males (PH, LO, SD) and one of predominantly small males (FP, MD, WB). The coefficients of variation ranged between 11% (FP) and 23% (LO), which are typical values for Florida populations, but these differences were not significant (F5,19 =1.68, P>0.05). The differences in the size distribution between these populations (as indicated by significant differences in average male size) are therefore due to different positions on a general log-normal distribution. The Relationship Between Size and Behaviour Populations showed considerable heterogeneity in the relationships between size and rates of the three behaviour patterns (Figs 2–4). The regression slopes of ln behaviour rate on ln standard length varied in sign and magnitude between populations for each behaviour (Table II). The rate of courtship displays was positively related to male standard length in all six populations, but in only three, Fiddler’s Point, Pinhook and Live Oak, was the slope significantly greater than zero (Fig. 2). Fiddler’s Point was the only population to show a significant positively allometric relationship (a slope significantly greater than 1.0) between male size and courtship display rate. Mounds males showed no significant relationship between display rate and male size. The rate of gonopodial thrusting showed a significant negative relationship with male standard length for the Pinhook and Live Oak populations but no relationship for the remaining

63

populations. Live Oak was the only population to show significant negative allometry for this behaviour pattern. This lack of pattern in most populations is distinct from the significant negative relationships described by Farr et al. (1986). Gonoporal nibbling showed a significant positive allometric relationship with male size in the Wakulla Beach population but no relationship in any other population. Courtship display rates, adjusted for body size, differed significantly between the four populations whose slopes could be considered homogeneous (Table III). The Fiddler’s Point and Mounds populations could not be included in this analysis because of their significant slope heterogeneity, a violation of the assumptions of ANCOVA. Although two of these populations (Wakulla Beach and Steve’s Ditch) did not show a significant non-zero slope when tested alone (Table II), the analysis of covariance did not indicate significant heterogeneity of slopes, and a single slope fits all four populations. Differences between these populations accounted for 38% of the variance observed between all males in courtship display rates, much more than the 10% ascribable to variation in male body size. Males from Live Oak showed the highest adjusted display rates among these four populations, followed in rank order by males from Pinhook, Wakulla Beach and Steve’s Ditch (Table IV). Males from Steve’s Ditch had very low adjusted display rates, despite their large body sizes (Table I). The relatively lower adjusted display rates in males from Wakulla Beach, a population where males are distinctly smaller than at the other three, is not in accord with the expectations derived from Farr et al. (1986). The results of multiple comparison tests of body size and behaviour rates, unadjusted for size because of extensive slope heterogeneity, offer additional insights (Table I). The similarities between populations in male size did not predict which populations were most similar in rates of the three behaviours. Nor were populations that were similar in average rates of one behaviour necessarily similar in average rates of the other two. This result strongly suggests that, at the interdemic level, each behaviour pattern can vary independently of both male size and other behaviour patterns. Populations with predominantly large males as well as those with predominantly small males showed both high and low rates of courtship

Frequency

45

65

15

45

Pinhook

35

65

65

X = 38.40 2 s = 53.12 N = 40

55

25 35 45 55 Standard length (mm)

25

20

0

15

15

Figure 1. Size–frequency histograms for male Poecilia latipinna from six populations.

55

15

10

20

0

45

20

0

X = 28.05 2 s = 34.56 N = 40

0

35

65

X = 35.48 2 s = 46.82 N = 40

55

Mounds

0

Steve's Ditch

35

10

20

10

25

25

X = 25.92 2 s = 8.99 N = 25

10

15

15

Fiddler's Point

10

20

0

10

20

25

25

45

35

45

Live Oak

35

Wakulla Beach

65

55

65

X = 38.80 2 s = 80.99 N = 40

55

X = 30.53 2 s = 46.74 N = 30

64 Animal Behaviour, 52, 1

Ptacek & Travis: Mollie mating behaviours

65

Table I. Results of Tukey’s HSD test comparing mean male size and behaviour rates (during 10-min observations) between the six populations of male Poecilia latipinna Populations Large/high Standard length Courtship displays Gonopodial thrusts

PH LO MD

LO PH PH

Small/low SD MD SD

WB WB FP

MD SD LO

FP FP WB

Gonoporal nibbles

SD

PH

LO

FP

MD

WB

Total number of acts Proportion of acts that are courtship displays

SD

PH

LO

MD

FP

WB

WB

LO

MD

PH

FP

SD

Comparisons were made between means from ln-transformed data. Mean values were calculated from measurements of all males collected from each population for standard length. Means for the three behaviour rates were calculated from the 20 males tested from each population. Populations are coded as follows: FP=Fiddler’s Point; MD=Mounds; WB=Wakulla Beach; SD=Steve’s Ditch; PH=Pinhook; LO=Live Oak. Population means that are not significantly different are connected by an underline.

displays. This lack of pattern is characteristic of these between-population comparisons for the other two behaviour patterns as well; there was no order in the comparative rankings of body size and the rates of gonopodial thrusts or gonoporal nibbles. Three lines of evidence indicate that the divergence between populations in behavioural profiles was not simply produced by divergence in overall activity levels. First, the slopes of the regressions of rates of single behavioural acts on body size were not homogeneous between populations. Simple divergence in activity rates should not generate heterogeneous slopes. Second, the variation between populations in the rate at which all behaviour patterns were observed (the sum of the three enumerated behavioural acts) did not correspond to the variation in any other measured character (Table I). Third, although the proportion of all observed behaviour patterns that were courtship displays varied markedly between populations, the pattern of variation did not correspond to that in any other measured character (Table I). ‘Large male’ and ‘small male’ populations were interdigitated with respect to how much total activity was devoted to courtship displays. These patterns support our contention that variation between these populations is not a simple consequence of either overall body-size differentiation or divergence in overall activity levels.

DISCUSSION Patterns in the relationship between behaviour rates and male size were similar to those observed by Farr et al. (1986) in the three ‘large male’ populations. The three ‘small male’ populations, however, showed marked deviations in one or more behaviour patterns. Differences between populations in average rates of behaviours were not ordered with respect to male size, suggesting that behavioural rates can vary between populations without being constrained by the direction of body size differences. For example, differences in average courtship display rates revealed two groups of populations, those with high average display rates (Mounds, Live Oak, Pinhook and Wakulla Beach) and those with low average display rates (Steve’s Ditch and Fiddler’s Point; Table I); however, populations having similar display rates were not ordered with respect to male size. The ordering of populations with respect to rates of gonopodial thrusting, while showing a more complex pattern, was also independent of male size. One group of populations showed high thrust rates (Mounds, Pinhook, Steve’s Ditch and Fiddler’s Point) and graded into a group with low thrust rates (Live Oak and Wakulla Beach; Table I). Behaviour rates show evidence of varying independently of one another as well. For example, males from Pinhook showed high rates of both

Ln courtship displays

4.2 Pinhook

3.8

3.4 3.8 Ln standard length

3.4

Mounds

4.2

4.2

–3 3.0

–1

1

3

5

–3 3.0

–1

1

3

5

3.4

3.4 Live Oak

3.8

3.8

Wakulla Beach

Figure 2. Scattergrams of number of courtship displays plotted against standard length for male Poecilia latipinna from six populations.

–3 3.0

4.2

–3 3.0 3.8

–1

–1

3.4

1

1

5

–3 3.0

3

Steve's Ditch

3.8

3

5

3.4

–1

–1

–3 3.0

1

1

5

3

Fiddler's Point

3

5

4.2

4.2

66 Animal Behaviour, 52, 1

Ln gonopodial thrusts

4.2 Pinhook

3.8

3.4 3.8 Ln standard length

3.4

Mounds

4.2

4.2

–3 3.0

–1

1

3

5

–3 3.0

–1

1

3

5

3.4

3.4 Live Oak

3.8

3.8

Wakulla Beach

Figure 3. Scattergrams of number of gonopodial thrusts plotted against standard length for male Poecilia latipinna from six populations.

–3 3.0

4.2

–3 3.0 3.8

–1

–1

3.4

1

1

5

–3 3.0

3

Steve's Ditch

3.8

3

5

3.4

–1

–1

–3 3.0

1

1

5

3

Fiddler's Point

3

5

4.2

4.2

Ptacek & Travis: Mollie mating behaviours 67

Ln gonoporal nibbles

4.2 Pinhook

3.8

3.4 3.8 Ln standard length

3.4

Mounds

4.2

4.2

–3 3.0

–1

1

3

5

–3 3.0

–1

1

3

5

3.4

3.4 Live Oak

3.8

3.8

Wakulla Beach

Figure 4. Scattergrams of number of gonoporal nibbles plotted against standard length for male Poecilia latipinna from six populations.

–3 3.0

4.2

–3 3.0 3.8

–1

–1

3.4

1

1

5

–3 3.0

3

Steve's Ditch

3.8

3

5

3.4

–1

–1

–3 3.0

1

1

5

3

Fiddler's Point

3

5

4.2

4.2

68 Animal Behaviour, 52, 1

Ptacek & Travis: Mollie mating behaviours

69

Table II. Values for slopes of the regressions of ln behaviour rate on ln standard length for each behaviour for the six populations of male Poecilia latipinna Standard length (mm) Population Fiddler’s Point Mounds Wakulla Beach Steve’s Ditch Pinhook Live Oak Overall slopes

Displays

Thrusts

Nibbles

Mean

Range

C.V.

Mean

Slope

Mean

Slope

Mean

Slope

26 29 32 37 40 39

23–33 22–42 23–53 27–55 24–62 24–63

0.12 0.21 0.22 0.19 0.19 0.23

11 27 17 8 32 36

9.98* 0.14 1.41 1.11 1.58* 1.05* 1.77*

11 17 4 23 16 10

1.12 "2.07 3.10 "0.67 "1.64* "3.44* "0.88

5 3 3 23 7 8

"1.37 0.19 4.40* "0.08 1.82 2.10 1.93

Twenty males from each population were tested. Average numbers of behaviours were calculated from 10-min observation periods. C.V.=coefficient of variation. Asterisks indicate slopes significantly different from zero at P<0.05. Table III. ANCOVAs for the effects of body size and population of origin on courtship display rates of male Poecilia latipinna for four populations: Live Oak, Pinhook, Wakulla Beach and Steve’s Ditch Source

df

Sums of squares

Body size Population Residual Interaction Error Total

1 3 75 3 72 79

10.61 37.93 53.79 0.24 53.55 102.33

courtship displays and gonopodial thrusting, but males from Live Oak and Wakulla Beach showed high rates of courtship displays and low rates of gonopodial thrusting. Males from Steve’s Ditch showed the opposite pattern. Differences between populations in some of the slopes for rates of courtship displays and gonopodial thrusting indicate that the level of behavioural allometry differs between populations (Table I). The slope estimates for the three ‘small male’ populations must, however, be interpreted with caution, especially in the Fiddler’s Point population. Smaller variances in male body size can lead to instability in slope estimates. Despite this statistical caveat, males from Mounds appeared to show no relationship between body size and courtship display rate, but males from other populations did, and the data give no indication that a larger sample size of males from Mounds would change this conclusion. Similarly, although no population showed a significant

F

P

Coefficient of determination

14.33 17.64 — 0.11 —

0.001 0.0001 — 0.95 —

0.10 0.38 — — —

positive relationship between gonopodial thrusting and body size, it is reasonable to conclude that some populations do show such a relationship, but others do not. In the simplest of mechanisms through which real variation in the level of behavioural allometry can arise, an ecological factor might favour increased effort for each male’s most effective behaviour. For example, higher densities might select for larger males to perform more displays to capture female attention, and smaller males would devote less time to ineffective displays and resort to other mechanisms for achieving mating success. The resultant increased range in courtship displays would create a steeper slope for the regression of courtship displays on body size. Variation between localities in characteristic population sizes would help to create variation in the levels of allometry. A more complex set of mechanisms would invoke an interplay of local variation in natural

Animal Behaviour, 52, 1

70

Table IV. Adjusted behaviour rates, back-transformed from natural logarithmic scale (during 10-min observations), for different-sized male Poecilia latipinna from six populations Behaviour pattern Population Courtship displays Fiddler’s Point Mounds Wakulla Beach Steve’s Ditch Pinhook Live Oak Gonopodial thrusts Fiddler’s Point Mounds Wakulla Beach Steve’s Ditch Pinhook Live Oak

Male standard length (mm) 25 40 55

2 24 10 3 13 20

333 26 20 6 29 34

6438 27 30 8 46 47

6 17 1 9 26 11

11 6 3 7 11 2

15 3 8 5 7 1

selection on body size and directional sexual selection for behaviour rates. Natural selection for smaller size (e.g. Trexler et al. 1994) might reduce size variation, but a directional preference of females for larger male size could produce a response in traits that increase apparent body size, such as fin characters or higher courtship display rates. Natural selection reduces the range of body sizes, but sexual selection increases the range of courtship displays or relative fin sizes. The result could be differential behavioural or morphological allometry. Although there is some evidence for differential allometry, we found even stronger evidence for substantial differences between populations in the size-specific rates of behaviour patterns. Males of the same body size from different populations showed up to six-fold variation in rates of courtship displays and four- to five-fold variation in rates of gonopodial thrusting. These values extend considerably the range of such variation that had been reported earlier (Farr et al. 1986). It is conceivable that the observed variation between males from different populations arose artefactually through differential acclimation to the laboratory regime by males from different populations or differential acclimation by females and a consequent shift in behaviour by males. Three lines of evidence argue against differential acclimation. First, experimentally induced variation in ontogenetic experience produced no effect

on size-specific behaviour patterns of males (Farr & Travis 1989). Second, short-term variation in the social milieu produced negligible variation in size-specific behaviour patterns (Travis & Woodward 1989), and the major variation that it induced appears to have been a laboratory artefact (Travis 1994b). Third, the patterns observable in one natural population are well matched by the laboratory observations (Travis 1994b). Laboratory conditions might induce significant shifts in male behaviour if the six populations possess distinct norms of reaction for male behaviour patterns (Via & Lande 1985; Stearns & Koella 1986; Carroll & Corneli, in press). Our laboratory regime might inadvertently recreate one of the limited conditions under which population variation will appear; under other conditions there would be no variation. Although we cannot refute this possibility, our laboratory conditions are well within the range of conditions that these estuarine populations experience (Trexler et al. 1990). Also, divergent patterns of plasticity between populations to any social or environmental gradient will be precluded by moderate amounts of gene flow (Via & Lande 1985; Gomulkiewicz & Kirkpatrick 1992), and these mollie populations appear to exchange migrants at a very high rate (Trexler 1988; Trexler et al. 1990). Results of these behavioural observations indicate that populations differ substantially in the relationship between rates of certain mating behaviours and male body size, and we have refuted earlier work that indicated that behavioural variation between populations was ordered with respect to male size. Our results also fail to support a hypothesis of any unidirectional evolutionary pathway (e.g. that sexual selection always increases display rates, as previous work has suggested). At a minimum, our results show that condition-dependent behaviour can vary far more than has been generally assumed. The direction of behavioural evolution may also be far more unpredictable. Regardless of causation, this study demonstrates the potential for divergence in these traits that exceeds prior expectation. Further study will be required to uncover the directions of evolution as well as its causes. ACKNOWLEDGMENTS We thank C. Johnson, J. Leips, C. Baer and M. Woodward for field assistance in collecting fish.

Ptacek & Travis: Mollie mating behaviours We also thank M. Woodward and J. Lewis for assistance in behavioural observations. The manuscript benefitted from comments by A. Thistle, F. Dyer and two anonymous referees. This research was sponsored in part by NSF grant number DEB 92-20849 to J.T. REFERENCES Carroll, S. & Corneli, P. In press. Norms of reaction in behavior: the evolution of mating systems in soapberry bugs. In: Geographic Variation in Behavior: Evolution and Ecology (Ed. by S. Foster & J. Endler). Chicago: University of Chicago Press. Dingle, H. 1994. Genetic analysis of animal migration. In: Quantitative Genetic Approaches to Animal Behavior (Ed. by C. M. Boake), pp. 145–164. Chicago: University of Chicago Press. Draper, N. R. & Smith, H. 1981. Applied Regression Analysis, 2nd edn. New York: Wiley. Farr, J. A. & Travis, J. 1986. Fertility advertisement by female sailfin mollies, Poecilia latipinna (Pisces: Poeciliidae). Copeia, 1986, 467–472. Farr, J. A. & Travis, J. 1989. The effect of ontogenetic experience on variation in growth, maturation, and sexual behavior in the sailfin molly, Poecilia latipinna (Pisces: Poeciliidae). Environ. Biol. Fish., 26, 39–48. Farr, J. A., Travis, J. & Trexler, J. C. 1986. Behavioural allometry and interdemic variation in sexual behaviour of the sailfin molly, Poecilia latipinna (Pisces: Poeciliidae). Anim. Behav., 34, 497–509. Garland, T., Jr. & Adolph, S. C. 1991. Physiological differentiation of vertebrate populations. A. Rev. Ecol. Syst., 22, 193–228. Gomulkiewicz, R. & Kirkpatrick, M. 1992. Quantitative genetics and the evolution of reaction norms. Evolution, 46, 390–411. Gross, M. R. 1985. Disruptive selection for alternative life histories in salmon. Nature, Lond., 313, 47–48. James, F. C. 1991. Complementary descriptive and experimental studies of clinal variation in birds. Am. Zool., 31, 694–706. Luckner, C. L. 1979. Morphological and behavioral polymorphism in Poecilia latipinna males (Pisces: Poeciliidae). Ph.D. dissertation, Louisiana State University. Lynch, C. B. 1994. Evolutionary inferences from genetic analyses of cold adaptation in laboratory and wild populations of the house mouse, Mus domesticus. In: Quantitative Genetic Approaches to Animal Behavior (Ed. by C. M. Boake), pp. 278–301. Chicago: University of Chicago Press. McPeek, M. A. 1992. Mechanisms of sexual selection operating on body size in the mosquitofish (Gambusia holbrooki). Behav. Ecol., 3, 1–12. O’Brien, R. G. 1978. Robust techniques for testing heterogeneity of variance effects in factorial designs. Psychometrika, 43, 327–342. Parzefall, J. 1969. Zur vergeichenden Ethologie verschiedener Mollienesia-Arten einschliesslich

71

einer Höhlenform von M. sphenops. Behaviour, 33, 1–37. Ryan, M. J. & Causey, B. J. 1989. ‘Alternative’ mating behavior in the swordtails Xiphophorus nigrensis and Xiphophorus pygmaeus. Behav. Ecol. Sociobiol., 24, 341–348. Ryan, M. J., Pease, C. M. & Morris, M. R. 1992. A genetic polymorphism in the swordtail Xiphophorus nigrensis: testing the prediction of equal fitnesses. Am. Nat., 139, 21–31. Sokal, R. R. & Braumann, C. A. 1980. Significance tests for coefficients of variation are variability profiles. Syst. Zool., 29, 50–66. Stearns, S. C. & Koella, J. C. 1986. The evolution of phenotypic plasticity in life-history traits: predictions of reaction norms for age and size at maturity. Evolution, 40, 893–913. Sumner, I. T., Travis, J. & Wilson, C. D. 1994. Methods of female fertility advertisement and variation among males in responsiveness in the sailfin molly (Poecilia latipinna). Copeia, 1994, 27–34. Trautman, M. B. 1981. The Fishes of Ohio. Columbus, Ohio: Ohio State University Press. Travis, J. 1994a. The interplay of life-history variation and sexual selection in sailfin mollies. In: Ecological Genetics (Ed. by L. A. Real), pp. 205–232. Chapel Hill: University of North Carolina Press. Travis, J. 1994b. Size-dependent behavioral variation and its genetic control within and among populations. In: Quantitative Genetic Approaches to Animal Behavior (Ed. by C. M. Boake), pp. 165–187. Chicago: University of Chicago Press. Travis, J. & Trexler, J. C. 1987. Regional variation in habitat requirements of the sailfin molly, with special reference to the Florida Keys. Florida Game and Fresh Water Fish Comm. Nongame Wildl. Program Tech. Rep. No. 3. 47 pp.+iii. Travis, J. & Woodward, B. D. 1989. Social context and courtship flexibility in male sailfin mollies, Poecilia latipinna (Pisces: Poeciliidae). Anim. Behav., 38, 1001– 1011. Trexler, J. C. 1988. Hierarchical organization of genetic variation in the sailfin molly, Poecilia latipinna (Pisces: Poeciliidae). Evolution, 42, 1006–1017. Trexler, J. C., Travis, J. & Trexler, M. 1990. Phenotypic plasticity in the sailfin molly, Poecilia latipinna (Pisces: Poeciliidae). II. Laboratory experiment. Evolution, 44, 157–167. Trexler, J. C., Tempe, R. C. & Travis, J. 1994. Sizeselective predation of sailfin mollies by two species of heron. Oikos, 69, 250–258. Verrell, P. A. 1988. Stabilizing selection, sexual selection, and speciation: a view of specific-mate recognition systems. Syst. Zool., 37, 209–215. Via, S. & Lande, R. 1985. Genotype-environment interactions and the evolution of phenotypic plasticity. Evolution, 39, 505–522. Warner, R. R., Robertson, D. R. & Leigh, E. G. 1975. Sex change and sexual selection. Science, 190, 633–638. Wilbur, H. M. & Collins, J. P. 1973. Ecological aspects of amphibian metamorphosis. Science, 182, 1305–1314.