Variation in thermal sensitivity of performance among colour morphs of a pygmy grasshopper

Authors


Anders Forsman Department of Engineering & Natural Sciences, Växjö University, S-351 95 Växjö, Sweden. Tel: +46 (0)470 708955; fax: +46 (0)470 708756; e-mail: Anders.Forsman@itn.vxu.se

Abstract

Populations of pygmy grasshoppers, Tetrix subulata, display genetically coded discrete variation in colour pattern and there are differences among morphs in the capacity to achieve body heating. To determine whether colour morphs differ in thermal physiology, I assessed reaction distance and jumping performance of individuals belonging to different morphs at two different temperatures. Individuals allowed a potential predator to approach less closely and jumped longer distances at high than at low temperature. My analyses also uncovered variation among morphs in average reaction distance and jumping capacity, as well as in thermal sensitivity of these two traits. Matrix correlation analysis further revealed that pair-wise differences between morphs in thermal sensitivity of jumping performance (but not reaction distance) could be accurately predicted by differences in body temperatures preferred in a laboratory thermal gradient. These results support the view that morphology, behaviour and thermal physiology of ectotherms may evolve in concert. The relationship between reaction distance and jumping performance varied among colour morphs at high temperature, and the common within-morph relationship between these two traits deviated from the corresponding among-morph relationship. This suggests that the variation among morphs has partially arisen through active divergence, with selection having influenced both traits and modifications having occurred to different degrees in different morphs. My data further suggest that pale colour morphs, with a limited capacity to attain high body temperatures, may not necessarily be at a selective disadvantage, because their physiology may be adapted to lower body temperatures.

Introduction

Ectothermic animals, such as insects and reptiles, rely on external radiant energy to achieve body heating, and are able to maintain body temperatures that deviate significantly from those in the surrounding environment by adjusting their behaviour (e.g. Casey, 1981; Peterson et al., 1993; Forsman, 1995a; Lactin & Johnson, 1996). Despite the apparent efficacy of such behavioural thermoregulation, however, the range of body temperatures potentially available to individuals may be constrained by environmental factors (Huey et al., 1989; de Jong et al., 1996), as well as by properties of the individual itself. For instance, body size and coloration both affect heating rates and equilibrium body temperatures (e.g. Digby, 1955; Willmer & Unwin, 1981; Bakken et al., 1985; Stevenson, 1985; Forsman, 1995a, 1997; de Jong et al., 1996). In environments where animals cannot easily attain high body temperatures one may thus expect coloration, or any other factor that affects the ability to achieve high body temperature, to indirectly influence individual fitness (e.g. Christian & Tracy, 1981; Kingsolver & Watt, 1983; Huey & Kingsolver, 1989; Willmer, 1991; Adolph & Porter, 1993; de Jong et al., 1996; Gilchrist, 1996; Forsman & Appelqvist, 1998).

This expected association between body temperature and fitness in ectotherms is based on the fact that body temperature affects many aspects of the physiological repertoire of individuals, including growth, egg production, behaviour, mating success and locomotor performance (e.g. Rand, 1964; Bennett, 1984; Huey & Hertz, 1984; Huey & Kingsolver, 1989, 1993; Harrison et al., 1991; Willmer, 1991; Gilchrist, 1996; Carriere & Boivin, 1997). This thermal sensitivity can be described by a performance curve, a plot of biological activity as a function of temperature, where the slope of the curve describes the sensitivity of the organism to temperature change. Most traits show a curvilinear relationship, with a peak in performance at some intermediate body temperature (e.g. Huey & Kingsolver, 1989; Gilchrist, 1996; Carriere & Boivin, 1997).

Recent studies of various ectothermic species have uncovered additive genetic variance in thermal performance curves (e.g. Gilchrist, 1996) and demonstrated that thermal sensitivity may show a rapid evolutionary response to temperature change (e.g. Bennett et al., 1992; Huey & Kingsolver, 1993; Partridge et al., 1995). Moreover, comparative studies across different species of lizards have revealed positive associations between preferred body temperatures and optimal temperatures for sprinting performance, suggesting that thermal preferences and thermal physiology may evolve in concert (Huey & Bennet, 1987; Bauwens et al., 1995). In line with this, data from a pygmy grasshopper (Tetrix subulata [L.] Orthoptera: Tetrigidae) provide evidence for parallel variation in behavioural thermoregulation and the ability to achieve high temperatures under insolation among different genetically coded colour morphs. Individuals belonging to dark morphs not only attain higher body temperatures when insolated but also prefer higher temperatures in a laboratory thermal gradient, compared with paler individuals (Forsman, 1997). The reasons for this variation in preferred temperature among colour morphs are unclear. One possible explanation, however, is that the physiology of pale morphs, with limited capacity to thermoregulate, may be adapted to lower body temperatures.

Here I examine whether thermal sensitivity of performance varies among different colour morphs. Tetrix subulata displays a strong thermal dependence of behaviour and locomotor performance; data from performance trials at four different temperatures (15, 20, 25 and 30 °C) have revealed that reaction distance, propensity to fly and jumping capacity increase with increasing ambient temperature (Forsman, 1999). These earlier results further showed an inconsistency in relative jumping performance of individuals across temperatures; the best jumpers at high temperature were not the best jumpers at low temperature. Jumping performance and reaction distance probably also influence individual fitness. Thus, laboratory experiments offer evidence that high jumping capacity enhances the probability of escape from attacking predators, and that visual predators impose correlational selection on colour pattern and behaviour (Forsman & Appelqvist, 1998).

In the experiment described below I measure reaction distance and jumping performance of Tetrix subulata at different temperatures and test for variation in thermal sensitivity among individuals belonging to different colour morphs. I examine whether differences in thermal sensitivity between morphs are correlated with estimates of differences in preferred body temperature, as one would expect if physiology and thermal preferences have evolved in concert (Huey & Bennet, 1987; Bauwens et al., 1995). Finally, I compare within- and among-morph relationships linking reaction distance to jumping performance. I expect a close resemblance between within- and among-morph relationships if the variation among colour morphs results primarily from random drift or reflects a correlated response to selection on only one of the characters (Lande, 1979; Lofsvold, 1988; Schluter, 1996). Conversely, I expect the within-morph relationship to deviate from the among-morph relationship if the variation is caused primarily by heterogeneous selection in the different morphs (Forsman & Shine, 1997).

Materials and methods

Biology of the insects

Tetrix subulata is a small (up to 14 mm body length, 0.07 g), short-limbed, diurnal, pygmy grasshopper, characterized by a long pronotum, and widely distributed in Europe (Holst, 1986). It lives on the surface of the soil where it feeds on moss, algae and humus. These grasshoppers overwinter as adults or late instars, emerge in early spring (April–May), reproduce and lay multiple clumps of eggs in late spring and summer, and develop into nymphs within 3–4 weeks, with the number of larval instars being five (males) or six (females) (Holst, 1986). Like many other grasshoppers (Nabours, 1929; Rowell, 1971; Holst, 1986), T. subulata exhibits discontinuous variation in colour and pattern of the pronotum. Within a single population individuals may vary from black, through yellowish-brown to light grey, with some individuals being monochrome and others having a distinct pattern, such as a light yellowish narrow midlongitudinal stripe on the upper surface of the pronotum (Holst, 1986; Forsman, 1997). Colours and patterns are distinct as early as the second instar and remain unchanged during the rest of the life of the individuals (A. Forsman, personal observation, see also Nabours, 1929). Long-term breeding experiments with several closely related species (Apotettix eurycephalus, Paratettix texanus, Telematettix aztecus, Tettigidea parvipennus) have shown that the polymorphism in elementary colour patterns in these taxa is due to several closely linked dominant factors, and that there is a strong tendency for the general patterns to be repeated in different genera and species (Nabours, 1929; Fisher, 1939).

Study animals and housing conditions

I conducted all experiments with adult female T. subulata collected from a population inhabiting the shorelines of a small pond, located 80 km east of Uppsala, Sweden, during 6 May–11 June 1997. I brought the animals to the laboratory and housed them individually in small plastic cups (measuring 70 mm in diameter and 80 mm in height) at room temperature (20–22 °C), with a slice of fresh potato as food and a small piece of moist cotton for drinking. I classified individuals as belonging to one of five different colour morph categories: uniform black; striped (black but with a narrow midlongitudinal yellowish stripe on the upper surface of the pronotum); light brown (uniform light brown dorsally and black laterally and ventrally); grey (uniform light grey dorsally and black laterally and ventrally), and others (various mottled morphs with black, dark brown, reddish brown or greenish ground colour) on the basis of their colour patterns. The first four morphs were all very distinct and easily classified. The various mottled morph category, however, was not a homogeneous group but consisted of several different morphs, some of which were quite similar and some of which occurred in very low frequencies.

Experimental procedures

To control for possible effects of acclimation to differences in body temperatures experienced in the wild, I maintained all animals under identical conditions in the laboratory (as described above) for 4–6 days prior to experimental testing. Performance trials were carried out at two different temperatures from low to high; 15 and 25 °C. This range of experimental temperatures was chosen because animals in the source population usually are not active below 15 °C and almost never experience ambient temperatures above 30 °C.

Studies of thermal sensitivity are constrained by the logistical trade-offs of making single or multiple measures, of few or many individuals, at two or several temperatures. Ideally, one would like to assess performance of individuals over six or more temperatures to allow quantification of various key parameters of the performance curves, such as performance breadth, performance optimum and maximum performance (Huey & Kingsolver, 1989; Gilchrist, 1996). Such a design, however, would require reducing the number of individuals and/or reducing the number of measurements per individual in each temperature. Preliminary trials indicated that measurement repeatability of jumping performance (estimated using data for 12 individuals that participated in jumping trials on four separate occasions at 25 °C) was significant but moderate (intraclass correlation coefficient = 0.54, F11,36 = 5.60, < 0.0001, Becker, 1984). I therefore decided to make multiple measurements of performance of each individual to obtain more precise estimates, with a corresponding increase in precision of thermal sensitivity estimates. Increasing the number of individuals at the expense of number of experimental temperatures was necessary to enhance the statistical power and the probability of detecting differences among colour morphs.

Jumping trials were conducted in a climatic chamber. I covered the floor of the chamber with white fabric to provide friction for take-off and to prevent the animals from sliding after landing. To avoid confounding effects of individual variation in acclimation ability, grasshoppers were allowed at least 8 h to acclimate to the new temperature before trials were initiated. I then removed a grasshopper from its cage, placed it in the middle of the floor on a marked spot, approached it slowly from behind with a hand-held pencil (meant to simulate a predator probing for prey) until the animal took off, and marked the place of landing. I repeated the approach procedure until each individual had conducted three consecutive jumps. After the third jump I placed the animal again in its cage, and measured the first reaction distance and the length of the three jumps to the nearest centimetre. When all individuals within a session had performed three jumps, I increased the room temperature to 25 °C, and gave the animals 8 h or more to acclimate before conducting a new set of trials. Results from an earlier experiment revealed no evidence for effects of habituation or prior experience on jumping performance, nor any effects of temperature sequence (i.e. low to high vs. high to low) on estimates of thermal sensitivity (Forsman, 1999).

At completion of the experiments I measured the animals for pronotum length using digital calipers to the nearest 0.01 mm and returned the animals to their cages. Individuals within a session were tested at random. Great care was taken to ensure that individuals belonging to different colour morphs were distributed evenly among sessions. I used data on mean jump length for each individual at each temperature in the statistical analyses (but all my results were similar when data on maximum jump length were used). Data on reaction distances were log(x + 1) transformed.

Traditionally, physiological traits are corrected for variation in body size prior to analyses. However, results from my previous experiment revealed no associations between morphological traits and performance in Tetrix subulata (Forsman, 1999). My data from the present study confirm this lack of relationship with morphology. Pearson correlation coefficients between body size (measured as length of the pronotum) and performance variables were low and, although two of the four correlations approached statistical significance, body size accounted for less than 2% of the variance in performance among individuals (size vs. reaction distance at 15 °C: = 0.13, = 0.06; reaction distance at 25 °C: = –0.10, = 0.14; jump length at 15 °C: = 0.12, = 0.09; jump length at 25 °C: = 0.01, = 0.84, all n = 212). I therefore did not include body size in subsequent analyses.

Statistical analyses

To test for effects of temperature and colour morph on reaction distance and jumping performance, I used the multivariate analysis of variance (MANOVA) method for repeated measures design (see O’Brian & Kaiser, 1985, for details), using procedure GLM (SAS Institute Inc. 1988). In this approach, dependent variables represent the change in linear dimensions between measurements, and because performance was measured in two temperatures there is one degree of freedom. The within-subjects part and the between-subjects part test for main effects of temperature and colour morph, respectively. The most important component for this 2 × 5 design (two temperatures, five colour morphs), however, is the morph by temperature interaction, which answers the question as to whether individuals belonging to different colour morphs undergo different mean changes in performance across temperatures (O’Brian & Kaiser, 1985). The interaction term thus tests the null-hypothesis of no differences in thermal sensitivity of performance among colour morphs. The above analysis was carried out separately on data on reaction distance and jumping performance. I also computed, for each individual, the change in reaction distance and jump length between trials conducted at high and low temperature. These data were used to test for overall divergence in thermal sensitivity among colour morphs, by treating the changes in reaction distance and jump length as two dependent variables in a MANOVA, and to construct the thermal sensitivity matrices described below.

To test whether the divergence in thermal sensitivity among pairs of colour morphs could be accurately predicted by differences in thermal preferences between morphs, I constructed a matrix of thermal sensitivity differences and compared this with a matrix of preferred body temperature differences. Data on preferred temperatures were obtained from a study of behavioural thermoregulation (A. Forsman, unpublished) where I estimated body temperatures of individuals belonging to different colour morphs in a laboratory thermal gradient. The methods and results of the preferred body temperature study will be reported elsewhere and I therefore only provide pertinent information here. I collected adult grasshoppers from a natural population and maintained them in the laboratory at 20–22 °C for 4 days prior to experimental testing. The thermal gradient consisted of an 800 × 350 × 3-mm copper plate, covered with a 4-mm layer of fine sand, and was divided into nine separate runways. A temperature gradient ranging from 14 to 50 °C was established by placing one end of the copper plate on a heat source and the other end on a bag with ice. Grasshoppers were placed individually in separate runways in the middle of the gradient, and allowed 20–30 min to acclimate before I recorded their position and temperature on six occasions at 10-min intervals. Analysis of data for 80 females revealed significant variation in mean preferred body temperatures among colour morphs (means ± SE, black 33.5 ± 1.09 °C, n = 18; brown 31.1 ± 1.63 °C, n = 18; grey 28.0 ± 1.52 °C, n = 13; striped 31.6 ± 0.99 °C, n = 10; others 28.4 ± 1.02 °C, n = 21, Kruskal–Wallis test, χ2 = 12.19, d.f. = 4, = 0.016, A. Forsman, unpublished). The elements within the constructed matrices are not independent. Nor can they be assumed to be normally distributed. I therefore estimated the association between elements in the thermal sensitivity matrix and the preferred body temperature matrix, and determined the significance of the correlation by comparison with the randomization distribution (5000 permutations) using Mantel's (1967) randomization test (see also Manly, 1991).

Comparing within- and among-morph slope coefficients

Finally, to test whether the divergence in thermal sensitivity among colour morphs can be explained by neutral rather than adaptive models of evolution, I compared the slope of within-morph linear relationships between performance variables with the slope of the among-morph relationships. I expect a close resemblance between within- and among-morph relationships if divergence results primarily from drift or correlated response (Lande, 1979; Lofsvold, 1988; Schluter, 1996). Conversely, I expect within-morph relationships to vary among the different colour morphs and to deviate from the among-morph relationship if divergence is caused largely by heterogeneous selection (Forsman & Shine, 1997).

Bivariate slope coefficients were calculated from least-squares linear regressions. Variables were standardized to mean zero within each colour morph prior to analyses. I estimated within-morph slope coefficients from separate regressions through individual data points within each colour morph and then tested for variability in slope coefficients among colour morphs using a heterogeneity of slopes model in GLM (SAS Institute, 1988). I also computed common within-morph relationships from regressions through individual data points using pooled data for all morphs. For this purpose I standardized the data to mean zero within each morph prior to analysis. Among-morph slope coefficients were estimated from regressions through morph means, and their 95% confidence intervals were estimated using bootstrap.

The comparisons of within- and among-morph relationships are based on the assumption that phenotypic correlations between performance variables are similar to the underlying genetic correlations. Unfortunately I do not have the data necessary to test this assumption. However, several studies have found that there are marked similarities between the genetic and phenotypic covariance patterns, such that estimates of phenotypic correlations may, in general, be suitable substitutes for the genetic correlations, at least for coarser kinds of comparisons and predictions made here (e.g. Cheverud, 1988; Roff, 1995; Schluter, 1996; but see Willis et al., 1991).

Results

General

I obtained data on reaction distance and jumping performance for 212 individuals (46 black, 40 brown, 42 grey, 30 striped, 54 ‘others’). Significant correlations were evident between the two kinds of performance variables within each temperature, and within each performance variable between the two different temperatures (Table 1). Thus, individuals with long reaction distances also tended to have a high jumping performance. Overall, ambient temperature exerted a significant effect on both reaction distance and jumping performance, as described in greater detail below. My analyses also revealed significant overall variation in performance among the different colour morphs (MANOVA on reaction distance and jumping perfor-mance measured at both temperatures, Wilk's λ = 0.86, F16,623.9 = 2.01, = 0.011).

Table 1.  Pearson correlations (above the diagonal) and partial correlations (below the diagonal) between reaction distance and jumping performance in Tetrix subulata measured at 15 °C and at 25 °C. Partial correlation coefficients were obtained from the corrected sums of squares and cross-products matrix and measure the association between any pair of traits when the effects of the other two traits are held constant. All variables were standardized to mean zero within each colour morph prior to analyses. The analyses are based on data for 212 individuals. Thumbnail image of

Reaction distance

Mean reaction distance increased significantly with increasing ambient temperature (MANOVA for repeated measures, effect of temperature, Wilk's λ = 0.89, F1,207 = 24.69, < 0.0001). Changes in temperature had the same effect on reaction distance in all colour morphs (effect of interaction, Wilk's λ = 0.99, F4,207 = 0.59, = 0.67), but average reaction distance varied among individuals belonging to different morphs (F4,207 = 3.34, = 0.011) (Fig. 1a). Univariate results from separate ANOVAs of data at high and low temperature revealed significant variation in reaction distance among colour morphs at 15 °C (F4,207 = 2.60, = 0.037) but not at 25 °C (F4,207 = 2.18, = 0.072). When ambient temperature was increased from 15 to 25 °C, average reaction distance (pooled morphs) increased by 34%, from 21.6 cm to 28.9 cm (Fig. 1a).

Figure 1.

Variation in thermal sensitivity of reaction distance and jumping performance in different colour morphs of the pygmy grasshopper Tetrix subulata. Data from estimates of performance trials at low, 15 °C, and high, 25 °C, temperature in a climatic chamber.

Jumping performance

Analysis of data on jumping performance revealed significant heterogeneity in thermal sensitivity among colour morphs (MANOVA for repeated measures, effect of interaction, Wilk's λ = 0.96, F4,247 = 2.60, = 0.037). Thus, differences in relative performance among morphs were not consistent across temperatures, as indicated by the crossing performance gradients depicted in Fig. 1(b). Morphs that performed poorly at low temperature performed relatively well at high temperature. Conversely, morphs that performed well at low temperature showed relatively poor performance at high temperature. Univariate ANOVAs on data from high and low temperature uncovered close to significant variation in performance among colour morphs at 15 °C (F4,247 = 2.36, = 0.054) but not at 25 °C (F4,247 = 0.71, = 0.59). When ambient temperature was changed from 15 to 25 °C, average jump length (pooled morphs) increased by 15%, from 53.4 cm to 61.6 cm (Fig. 1b).

Analysing changes in reaction distance and jumping performance

Significant variation in thermal sensitivity among colour morphs was evident also when data on changes in reaction distance and jumping performance between low and high temperature were analysed together (MANOVA, Wilk's λ = 0.92, F8,412 = 2.24, = 0.024). This divergence was due primarily to the striped morph, which showed a relatively small change in reaction distance, and a relatively large change in jumping performance following an increase in ambient temperature, compared with remaining colour morphs (Fig. 2).

Figure 2.

  Divergence in thermal sensitivity among colour morphs in Tetrix subulata, from comparisons of changes in reaction distance and jumping performance associated with an increase in ambient temperature from 15 °C to 25 °C.

Thermal preference was a reliable predictor of thermal sensitivity of jumping performance but not of reaction distance. Thus, results from Mantel's tests showed that the differences in preferred body temperatures between pairs of colour morphs accounted for 10.9% of the divergence in thermal sensitivity of reaction distance, but the matrix correlation was not significant (= 0.33, = 0.16). In contrast, differences in preferred body temperatures accounted for 22% of the divergence in thermal sensitivity of jumping performance, as evidenced by a strong and significant negative matrix correlation (= –0.47, = 0.042). The conflicting results for reaction distance and jumping ability were due to the fact that differences between morphs in thermal sensitivity of reaction distance were not associated with differences in thermal sensitivity of jumping performance (matrix correlation, = 0.04, = 0.40). This lack of association between reaction distance and jumping performance further ensures that the comparison between the jumping performance matrix and the thermal preference matrix was not confounded by reaction distance, and vice versa.

Comparing within- and among-morph relationships

All but one of the within-morph slope coefficients between reaction distance and jumping performance measured in high and low temperature were positive (Table 2). None of the four kinds of slope coefficients varied significantly among the five different colour morphs, but the test results for heterogeneity of slopes of the relationships linking reaction distance at 25 °C to jump length at 25 °C, and for the relationship linking jump length at 15 °C to jump length at 25 °C were close to statistical significance; = 0.052 and = 0.092, respectively (Table 2). Calculations of common within-morph slope coefficients for the combined morphs (based on data that were standardized to mean zero within morphs) yielded estimates that were positive and significantly different from zero in all four cases (Table 3).

Table 2.  Linear slope coefficients between reaction distance (R) and jumping performance (J) measured at two different temperatures (15 and 25 °C) within five colour morphs of Tetrix subulata. Data were standardized to mean zero within colour morphs prior to analyses. Results from tests for heterogeneity of slopes among colour morphs are shown. Thumbnail image of
Table 3.  Common within- and among-morph slope coefficients between reaction distance (R) and jumping performance (J) measured at two different temperatures (15 and 25 °C) in Tetrix subulata. Data were standardized to mean zero within colour morphs before computing common within-morph coefficients from simple least-squares linear regressions through individual data points. Among-morph slope coefficients were estimated from regressions through mean values for each colour morph, and their 95% confidence intervals were estimated using bootstrap. Thumbnail image of

In contrast to within-morph relationships, among-morph slope coefficients (estimated from regressions through morph means) varied in both direction and magnitude (Fig. 3, Table 3). However, 95% confidence intervals were large and only the inverse relationship between reaction distance at 25 °C and jump length at 25 °C (estimated slope =–0.70) differed significantly from zero (Table 3). Pair-wise comparisons between among- and common within-morph relationships also revealed a significant difference only for the relationship between reaction distance and jump length in high temperature (Fig. 3, Table 3). All of the within-morph slope coefficients were larger than the corresponding among-morph values (20 of 20), and in 7 of 20 cases they were outside the 95% confidence interval of the among-morph slope estimate (Tables 2 and 3).

Figure 3.

  Relationships between reaction distance and jumping performance measured at two different temperatures. The figures are based on data for five different colour morphs and show means and standard errors. Thick lines represent among-morph relationships based on linear regressions through the mean values for each colour morph. Thin lines represent common within-morph relationships and are based on regressions through individual data points for 212 individuals (data were standardized prior to analysis so that all colour morphs had the same means).

Discussion

My analyses revealed significant differences among Tetrix subulata colour morphs in reaction distance and jumping capacity, as well as in their thermal sensitivity, and suggest that this variation has partially arisen through some process of active divergence, with modifications having occurred to different degrees in different morphs. The results from comparisons of slope coefficients thus suggest that some, but not all, of the divergence in reaction distance and jumping performance among colour morphs has occurred along the genetic lines of least resistance (Schluter, 1996). There was a general similarity among the different colour morphs with respect to the within-morph slope coefficients, and most (three of four) of the among-morph relationships were similar to the common within-morph relationships (2, 3Fig. 3, Tables 2 and 3). Assuming that phenotypic correlations are reliable indicators of genetic correlations (e.g. Roff, 1995; Gilchrist, 1996; Schluter, 1996; but see Willis et al., 1991), these patterns indicate that some of the differentiation in reaction distance and jumping capacity among colour morphs reflects a divergence in both characters along a slope determined by the additive genetic regression between the two traits (Lande, 1979; Lofsvold, 1988; Schluter, 1996; but see Willis et al., 1991). Despite this overall similarity, however, one important difference was also apparent. Thus, colour morphs displayed different relationships between jumping capacity and reaction distance at high temperature (Table 2), and the common within-morph relationship between these traits deviated significantly from the corresponding among-morph relationship (Fig. 3, Table 2). This suggests that although genetic constraints may have retarded the rate of evolutionary change (Lande, 1980; Arnold, 1992), they clearly have not altogether prevented modifications of thermal sensitivity of behaviour and performance in the different colour morphs.

The results from the matrix correlation analysis show that differences between colour morphs in thermal sensitivity of jumping performance were associated with differences in preferred body temperatures. No such correlation was evident between reaction distance and preferred temperature, however. The reason for the conflicting results for these two traits is unclear. A possible explanation, however, is that reaction distance and jumping performance are affected by different sets of genes (Roff, 1996). This interpretation is further supported by the finding that differences between morphs in thermal sensitivity of reaction distance and jumping performance were not correlated with each other. The underpinnings of the multiple-trait differentiation among colour morphs in Tetrix subulata have not been identified. It is possible, however, that the large number of closely linked gene loci that controls the colour and pattern of the pronotum in these grasshoppers (Nabours, 1929; Fisher, 1939) also influences other traits, such as details of morphology, behaviour and thermal physiology (Lank et al., 1995).

The variation in reaction distance and jumping performance apparent among colour morphs may partially result from selection on colour pattern and behaviour imposed by visual predators (Forsman & Appelqvist, 1998, 1999). Thus, the variation in thermal sensitivity among morphs depicted in Fig. 2 suggests that the striped morph, characterized by a relatively small change in reaction distance and a large change in jumping capacity following an increment in temperature, has diverged in a direction different from remaining morphs. I also found a positive association between reaction distance and jumping performance within temperatures (Table 1). A similar association is evident in comparisons of average performance of individuals across four different temperatures (Forsman, 1999), and supports the view that individuals may compensate for reduced escape capacity by modifying their antipredator behaviour (e.g. Bauwens & Thoen, 1981; Koufopanou & Bell, 1984; Schwarzkopf & Shine, 1991). Attracting the attention of potential predators by initiating escape at an early stage is less likely to be a detrimental strategy when locomotor performance is high (Forsman & Appelqvist, 1998; but see Rand, 1964, for an opposite view).

This raises the question as to why the comparison across morph means (at high temperature) revealed a significant inverse relationship between reaction distance and jumping ability (Table 3, Fig. 3). One possible explanation is that selection imposed by visual predators acts on the combination of trait values, rather than on prey coloration or behaviour per se (Brodie, 1992; Forsman, 1995b). Thus, in a previous study, we experimentally manipulated colour pattern and behaviour of Tetrix subulata and exposed them to predation from domestic chicks. We found that a striped colour pattern enhanced survival, compared with that of a uniformly black colour pattern, when reaction distance was short and jumping performance poor. Conversely, the striped pattern decreased survival when reaction distance was long and performance poor (Forsman & Appelqvist, 1998). Such correlational selection results in functional integration of and genetic coupling between different traits (e.g. Cheverud, 1982; Lande & Arnold, 1983; Endler, 1995), and may have promoted different combinations of reaction distance and jumping capacity in the different colour morphs.

An important insight from this study is that although coloration may have profound effects on the capacity for temperature regulation (see above), this may not necessarily translate into morph-specific differences in fitness because the physiology of individuals belonging to pale colour variants, with limited capacity to attain high body temperatures, may be adapted to lower body temperatures. Previous experiments have shown that differences in preferred body temperature between colour morphs are in perfect accordance with differences in the ability to achieve high temperatures under insolation (Forsman, 1997, and unpublished data). That finding, together with the association between thermal sensitivity of reaction distance and preferred body temperature apparent in the present study, supports the view that morphology, behaviour and thermal physiology evolve in concert (Huey & Bennett, 1987; Bauwens et al., 1995).

Further experiments are necessary to more fully resolve the issue of evolutionary changes in thermal sensitivity in Tetrix subulata colour morphs, however. For instance, estimates of performance under a larger number of temperatures are required to evaluate the relative roles of the ‘warmer is better’ and the ‘jack-of-all-temperatures is a master of none’ hypotheses in this system (Huey & Kingsoler, 1989; Bauwens et al., 1995). My finding of a significant morph by temperature interaction shows that relative jumping performance of colour morphs differed between temperatures (Fig. 1b). This result lends some support for the ‘jack-of-all-trades’ hypothesis, and is interesting also because it differs from that of Huey & Hertz (1984) who measured sprint speed of lizards at different temperatures and found that the fastest sprinters at optimal temperatures tended to be fast across all temperatures. However, my present data do not provide direct evidence for a trade-off between performance at high and low temperature (but see Gilchrist, 1996). The ‘warmer is better’ hypothesis is based on the finding that the catalytic efficiency of enzymes generally is higher at high than at low temperature, and predicts a positive correlation between maximum rate of performance and temperature at which performance is at its maximum. Tests of this hypothesis based on comparisons across species and populations have yielded conflicting results, however (reviewed in Gilchrist, 1996; Carriere & Boivin, 1997). Measures of performance of different colour morphs at several experimental temperatures would provide a powerful test of this hypothesis at the intrapopulation level, and answer the question of whether colour morphs differ in performance when allowed to thermoregulate at their preferred body temperature.

Acknowledgments

I thank M. Björklund and S. Ulfstrand for their helpful comments on earlier versions of the manuscript. S. Appelqvist assisted in the field and laboratory. The study was supported by grants from The Swedish Natural Science Research Council.

Ancillary