Jonas Ahnesjö, Department of Biology and Environmental Science, Kalmar University, SE-391 82 Kalmar, Sweden. Tel.: +46 (0)480 44 73 00; fax: +46 (0)480 44 73 05; e-mail: email@example.com
Theory posits that selection on functionally interrelated characters will promote physical and genetic integration resulting in evolution of favourable trait-value combinations. The pygmy grasshopper Tetrix undulata (Orthoptera: Tetrigidae) displays a genetically encoded polymorphism for colour pattern. Colour morphs differ in several traits, including behaviours, thermal biology and body size. To examine if these size differences may reflect phenotypic plasticity of growth and development in response to temperature we used a split brood-design and reared hatchlings from mothers belonging to different morphs in different thermal environments (warm or cold) until maturity. We found that time to maturity was longer in the cold compared with the warm treatment. In the warm (but not in the cold) treatment time to maturity also varied among individuals born to mothers belonging to different colour morphs. Although low temperature and long development time are normally accompanied by increased body size in ectotherms, our results revealed no difference in size at maturity between individuals reared in the two temperature treatments. There was also an increase (not a decrease) in adult body size with shortened time to maturity across families within each treatment. Taken together, this suggests that body size is canalized against environmental perturbations, and that early maturation does not necessarily trade off against a size-mediated decrease in fecundity. Heritability of body size was moderate in magnitude. Moreover, body size at maturity varied among individuals belonging to different morphs and was influenced also by maternal colour morph, suggesting that a genetic correlation exists between colour pattern and body size. These findings suggest that different characters have evolved in concert and that the various colour morphs represent different evolutionary strategies, i.e., alternative peaks in a multi-modal adaptive landscape.
Like many other grasshoppers (Rowell, 1971) T. undulata exhibits genetically encoded discrete variation in colour and pattern of the pronotum (Holst, 1986; Forsman et al., 2002, see below for details). Several lines of evidence suggest that colour pattern contributes to individual variation in fitness. For instance, a previous field experiment in which the phenotypes of free-ranging individuals were manipulated by painting revealed that colour pattern significantly influenced survival, suggesting that different morphs vary in susceptibility to visually oriented predators (Forsman & Appelqvist, 1998, 1999). Different colour morphs also vary in reproductive life history characteristics, including time interval between sequential clutches and adult body size (Forsman, 1999a, 2001). These size differences may reflect either a secondary response to evolution of alternative life-history strategies, or a morph-specific adjustment of size in response to selection for thermoregulatory capacity.
The previous comparison of body size was based on data for adult individuals collected from natural populations (Forsman, 1999a). Thus, although the observed size differences may reflect direct genetic effects, they may also result from indirect effects of coloration mediated via differential body temperature. Earlier studies of pygmy grasshoppers have shown that individuals belonging to alternative colour morphs differ in heating rates and thermoregulatory behaviour (Forsman, 1997; Forsman et al., 2002). Furthermore, female T. subulata experimentally maintained under warm conditions were more likely to oviposit, laid their first clutch earlier, produced more clutches and had shorter intervals between sequential clutches than did females in a colder environment (Forsman, 2001). However, it is not yet known whether temperature experienced during development influences performance and phenotypes of offspring. In the study reported here, full-sibling pygmy grasshopper individuals were reared in captivity under two different temperature regimes, using a split-brood design (Forsman et al., 2002). We estimate heritability of body size and examine whether time to maturity and adult body size is influenced by maternal colour morph and thermal environment experienced during development. A significant contribution of maternal colour morph on time to maturity and body size is expected under the hypothesis that the previously observed differences among colour morphs reflect a genetic correlation between loci influencing colour pattern, physiology and life history traits (Forsman, 1999a, b, 2001). In contrast, no effect of maternal colour morph is expected under the hypotheses that the differences in body size result primarily from plasticity of growth and development in response to morph-specific differences in body temperature. By rearing individuals under different temperature regimes we are able to evaluate the latter hypothesis: if thermal environment does not significantly influence body size, then the size-differences among colour morphs in natural populations probably are not attributable to indirect effects of coloration. Finally, we examine if time to maturity and size of individuals belonging to different morphs respond differently to temperature change, as a means of maintaining relative fitness in the face of environmental variation (Woltereck, 1909; Thompson, 1991; Nager et al., 2000).
Materials and methods
Study animals and source population
The pygmy grasshopper Tetrix undulata (Sow.) (Orthoptera, Tetrigidae) is a small (<15 mm body length) ground dwelling insect. It is common, widely distributed and inhabits many different habitats, although clear cuttings seem to be preferred (Holst, 1986). Adults and late nymphal instars over-winter and emerge in early spring when the reproductive season begins. Females produce multiple clutches of eggs.
Tetrix undulata displays a variability of discrete, genetically encoded colour morphs. Within a single population one may find individuals varying from black, through yellowish-brown to light grey, with some individuals being monochrome and others having a distinct pattern, such as a narrow light yellowish longitudinal stripe on the mid-line of the upper surface of the pronotum, or speckled in different shades of brown (Holst, 1986; Forsman et al., 2002). Colours and patterns are easily distinguishable in the second instar stage and remain unchanged during the rest of the life of the individual (personal observation, see also Nabours, 1929). Detailed long-term breeding experiments with laboratory colonies of several closely related species (Apotettix eurycephalus, Paratettix texanus, Telmatettix aztecus, Tetrigidea parvipennus) suggest that the polymorphism in these taxa is because of several closely linked dominant genes, and that these patterns are repeated in different genera and species (Nabours, 1929; Fisher, 1930, 1939). A single female may produce offspring belonging to several different colour morphs within the same litter (Forsman et al., 2002).
Influences of common environmental and maternal effects associated with genetic analyses can be assessed by using a half-sib design (Becker, 1984; Roff, 1997). However, this requires virgin females. Because our study species over-winter as adults or late nymphal instars and the breeding season commences in early spring it is thus necessary to maintain individuals in captivity during hibernation to obtain virgin females. However, a previous attempt to do this resulted in very high mortality. We therefore used a full-sib split-brood design instead.
Adult female grasshoppers were collected at two occasions in spring 2000 (26 April and 11 May) from a natural population inhabiting a burnt clear cutting located 50 km south of Växjö in south central Sweden. All animals were classified as belonging to either one of four colour morphs (black, brown, grey or striped). The black morph consisted of uniform black and black individuals with light brown markings on the lateral sides of the pronotum. The brown morph consisted of individuals in different brown nuances, some of which were speckled and others uniform. Grasshoppers classified as grey are black in ground colour but the upper surface of the pronotum is light grey. The striped morph is dark brown in ground colour with a yellowish longitudinal stripe running mid dorsally from the head to the apex of the pronotum. Additional colour morphs existed in the source population, but in low frequencies only and were therefore not used for the experiment. All females were assumed to have mated in the field prior to capture.
Body size was measured as the length of the pronotum to the nearest 0.01 mm using digital calipers (Mitutoyo Ltd, 500-181U, Andover, UK). The pronotum is a ‘one piece’ part of the exoskeleton and provides a good estimate of overall body size in pygmy grasshoppers (Forsman, 1997). Because measurements of this trait are not biased by soft body parts and joints, the measurement repeatability was very high (98.1%, F1,24 = 103.94, P < 0.0001), with only 1.9% of the variation among individuals being due to measurement error. This repeatability estimate was obtained from data from two repeated measurements of 24 individuals analysed using a one way anova as SB2/(SB2 + SW2), where SB2 = (MSB − MSW)/k, SW2 = MSW, k, the number of measurements per individual and MSW and MSB are the mean squares obtained within and between individuals, respectively (Sokal & Rohlf, 1981; Becker, 1984).
Rearing conditions for females
Wild-caught female pygmy grasshoppers were housed individually in plastic containers measuring 150 by 85 by 210 mm maintained in the laboratory at room temperature (22 ± 2 °C) and 75% relative humidity. Each cage contained a small aluminum cup (25 mm in diameter) of moist cotton for drinking and an identical cup filled with a mixture of equal amounts of moist peat and soil as food and oviposition media (Forsman, 1999a, 2001). Complementary food was a piece of fresh potato. The peat/soil containers were thoroughly examined for egg pods every second day and egg pods were carefully transferred to a piece of moist cotton inside a petri-dish (50 mm in diameter) for incubation at room temperature. Number of hatchlings was recorded.
Split-brood experimental design
To test for effects on survival, time to maturity and adult body size of rearing temperature and maternal colour morph, offspring from each female were reared in both cold and warm temperature regimes in a greenhouse, using a split-brood design. After hatching, all litters were split into two halves and each half was put in a separate 10 L white plastic bucket filled to about 1/5 with a 1 : 3 mixture of moist soil and peat as substrate. The buckets were covered with fibre-cloth (Lutrasil, Svalöf Weibull, Hammenhög, Sweden, 100% polypropene). No complementary food was provided to the hatchlings as they feed on humus and the naturally existing micro flora (e.g. algae and mosses) in the soil–peat mixture. The buckets were watered at regular intervals to prevent the substrate from drying and promote the growth of algae and mosses.
Buckets containing siblings were placed so that one half of each litter was raised in a ‘warm’ (south facing) compartment in a greenhouse and the other in a ‘cold’ (north facing) compartment. The automatic ventilation system was set differently in the two compartments to further exaggerate the difference in temperature regime between treatments. Tests for phenotypic plasticity in response to temperature are normally conducted under constant temperature conditions. However, such conditions bear little resemblance to the thermal conditions experienced by free-living organisms. Our experimental setup allows temperatures in both treatments to follow cyclical daily changes as well as long term changes over the entire duration of the experiment.
To quantify the thermal regime in each treatment, the temperature inside the buckets was recorded using digital thermometers placed in five different buckets in each of the two compartments. The thermometer probe was resting on the substrate surface inside the buckets and maximum and minimum temperatures were recorded every 2–5 days. Separate mean maximum and minimum values was computed for each compartment and recording date. Average minimum temperature differed significantly between the warm (mean ± SE, 12.9 ± 0.34, range 8.0–18.1 °C) and the cold treatment (11.0 ± 0.38, range 5.9–17.4 °C) (F1.86 = 12.96, P < 0.001)(Fig. 1). Also the average maximum temperature differed significantly, with higher average temperatures in the warm (mean ± SE, 32.2 ± 0.74, range 20.0–39.7 °C) compared with the cold treatment (27.4 ± 0.61, range 19.2–38.2 °C) (F1.86 = 24.67, P < 0.0001) (Fig. 1).
To estimate time to maturity and survival each bucket was thoroughly examined for surviving grasshoppers on two occasions during the experiment, in late summer (9–15 August) and in early autumn (5–8 September). On the first census, the grasshoppers were counted and classified by developmental stage (nymph or adult), colour morph (black, grey, striped, brown or ‘others’) and sex (adults only). The morph category ‘others’ consisted of offspring individuals belonging to colour morphs different from the categories used for the parental generation. On the second and final census, all adult individuals also were measured for body size (see above).
Estimation of survival
Survival was measured as the percentage of the released hatchlings in each bucket that was alive on the first census. The test for effects of treatment on survival was performed based on intra-family matched pairs comparisons. Because the frequency-distribution of survival data deviated from a normal-distribution, nonparametric statistics, i.e., the Mann–Whitney matched pair test, was used (Siegel & Castellan, 1988). Similarly, nonparametric anova (Kruskal–Wallis) was used to test for variation in survival among offspring born to mothers belonging to different colour morphs. Finally, a nonparametric anova was performed on the intra-family differences in proportion of survivors between treatments. This last test answers the question whether the effect of treatment on survival differed among offspring born to mothers belonging to different colour morphs, i.e., it tests for an effect of the interaction between temperature treatment and maternal colour morph on survival (Sokal & Rohlf, 1981, p. 328).
Estimation of time to maturity
An index of time to maturity was computed as the proportion of surviving individuals in each bucket that had reached the adult stage at the time of the first census. This index can take any value between 0 (if none of the individuals had reached the adult stage) and 1.0 (if all individuals within a bucket had reached the adult stage). Because time to maturity data departed from a normal distribution, tests for effects of treatment, maternal colour morph and their interaction were performed using nonparametric statistics, as described above for the analysis of survival data.
Estimating heritability of body size
Heritability of body size was estimated using standard methods for heritability estimates of a quantitative trait, i.e., by means of least-squares linear regression of mean offspring body size against maternal body size (Becker, 1984). To allow comparisons of heritability of body size between sexes and the two rearing environments, separate heritability estimates were computed for each category. In addition, we computed estimates of heritability of body size using a full-sib design, obtaining the covariance between full-sibs by a one-way anova (Roff, 1997, p. 41).
Testing for effects of treatment, maternal colour morph and their interaction on body size
To test for effects of rearing temperature and maternal colour morph on body size at maturity of offspring, we used data from the split-brood experimental design to calculate mean body size of mature siblings for each family and treatment. Data were analysed based on pair-wise comparisons between siblings reared in different environments, using repeated measures analysis of variance, and treating the mean body size of siblings within each family reared in warm and cold treatment as a repeated measure (Forsman et al., 2002). Because males and females differ in body size, sex was included as an independent variable in the model. In this approach, the ‘within subject’ component tests for a main effect of treatment, i.e., for phenotypic plasticity of body size in response to temperature regime. A significant contribution of maternal colour morph is interpreted as evidence for a covariance between colour morph and body size that may reflect either maternal effects, pleiotropic effects of a single gene, physical linkage between adjacent loci influencing colour morph and body size, or linkage dis-equilibrium between more or less independently segregating loci (Lynch & Walsh, 1998). Colour morph is a categorical variable, but the relation among the four alternative morphs does not lend itself to an unambiguous ordinal scale. It is therefore not meaningful to use standard procedure for estimating the magnitude of the genetic correlation.
The maternal colour morph by rearing temperature interaction answers the question whether the mean difference in body size between sibling individuals reared in the warm and cold treatment is different depending on the colour morph of the mother. A significant contribution of this interaction thus is interpreted as evidence for a gene by environment interaction, meaning that body size of offspring born to mothers belonging to different colour morphs respond differently to rearing temperature, i.e., have different norms of reaction (Woltereck, 1909). The above analysis was performed using procedure GLM in SAS, and sigma (type III) sums of squares were used to test hypotheses (SAS Institute Inc., 1988).
Descriptive statistics is provided in Table 1. Of the 111 wild-caught females used as parental generation, 34 females produced egg pods with zero hatching success, and 75 individuals produced clutches that yielded at least six live hatchlings (mean number of hatchlings 20.4, SE = 0.82, range 7–32). Two additional females produced less than five hatchlings each, but these litters were excluded from the experiment and subsequent analyses because of small sample size. Eggs hatched after 14.8 days (range 9–20 days) on average. The mean number of hatchlings released into each bucket did not differ significantly between the warm (mean ± SE, 10.1 ± 0.46, n = 75) and the cold (9.7 ± 0.43, n = 75) treatment (t147 = 0.59, ns). There also was no difference in number of hatchlings released into the buckets from mothers belonging to the four different colour morphs, neither in the warm (Kruskal–Wallis, H3 = 2.99, ns) nor in the cold treatment (H3 = 5.56, P = 0.09, Table 1).
Table 1. Summary statistics. Number of hatchling individuals initially released from each maternal colour morph. Proportion of individuals alive (%) at the first census. Number of individuals that had reached the adult (mature) stage at the first census. Values are mean ± 1 SE. Numbers within brackets indicate sample size.
Maternal colour morph
9.9 ± 0.8 (n = 29)
10.7 ± 0.9 (n = 11)
8.1 ± 1.4 (n = 9)
10.6 ± 0.8 (n = 26)
9.2 ± 0.7 (n = 29)
10.2 ± 0.9 (n = 11)
7.6 ± 1.3 (n = 9)
10.7 ± 0.7 (n = 26)
72 ± 4.5 (n = 29)
85 ± 4.5 (n = 11)
74 ± 10.1 (n = 9)
78 ± 3.7 (n = 26)
74 ± 4.6 (n = 29)
71 ± 10.0 (n = 11)
58 ± 13.1 (n = 9)
75 ± 5.4 (n = 26)
6.0 ± 0.6 (n = 28)
7.2 ± 1.2 (n = 11)
4.1 ± 1.2 (n = 9)
7.1 ± 0.7 (n = 26)
3.5 ± 0.7 (n = 28)
2.2 ± 0.7 (n = 11)
4.4 ± 2.1 (n = 9)
5.2 ± 0.8 (n = 26)
As the number of hatchlings released varied among buckets, negative effects of high density (i.e. crowding) potentially may have confounded our estimates of survival, time to maturity and body size. Tests for confounding effects of crowding in the buckets were therefore performed using Spearmans rank correlation analysis, with the initial number of hatchlings in each bucket regarded as an independent variable. No negative effects of crowding on survival were evident in the warm treatment (rs = −0.13, n = 75, ns). In the cold treatment the proportion of individuals alive at the first census increased with increasing number of hatchlings released (rs = 0.74, n = 75, P < 0.0001), suggesting that crowding influenced survival positively rather than negatively. This should not confound our comparisons between treatments and colour morphs, however, because the density of hatchlings did not differ between these entities (see above). For time to maturity, no effect of crowding was evident in any of the two treatments (warm: rs = −0.04, n = 69, ns; cold: rs = −0.05, n = 69, ns). There also was no association between average body size at maturity and number of hatchlings initially released, neither in the warm (rs = −0.052, n = 69, ns) nor in the cold (rs = −0.21, n = 64, P = 0.10) treatment.
No effects of treatment or maternal colour morph on survival
There was no overall difference in survival, measured as the percentage of initially released hatchlings in each bucket that was alive at the time of the first census, between individuals reared in the warm and cold treatment (mean ± 1 SE, warm treatment: 76 ± 2.56, n = 75; cold treatment: 72 ± 3.35, n = 75, Mann–Whitney, W = 5798.5, ns). There was also no significant difference in survival among offspring born to mothers belonging to different colour morphs, neither in the warm (Kruskal–Wallis; H3 = 2.45, ns) nor in the cold (H3 = 0.64, ns) treatment (Table 1). Survival of offspring born to mothers belonging to different colour morphs also did not respond differently to rearing temperature, as evidenced by the nonsignificant interaction between maternal colour morph and treatment (Kruskall–Wallis; H3 = 1.85, ns). Analysis of survival until the second census yielded results (not shown) that were very similar to those reported above for survival until first census.
Effects of treatment and maternal colour morph on time to maturity
Time to maturity, measured as the proportion of individuals in each bucket that had reached the adult stage at the time of the first census, was lower on average in the cold (mean ± SE, 54 ± 4.52, n = 69) than in the warm (81 ± 3.10, n = 69) treatment (Mann–Whitney, W = 5788.0, P < 0.0001). Time to maturity varied significantly among offspring born to females belonging to different colour morphs in the warm treatment (Kruskal–Wallis; H3 = 7.88, P = 0.05, Table 1, Fig. 2). However, no significant effect of maternal colour morph on time to maturity was evident in the cold treatment (H3 = 2.87, P = 0.41) (Table 1, Fig. 2). The interaction between colour morph and treatment on time to maturity also was not significant (Kruskal–Wallis; H3 = 5.04, ns), however, suggesting that rearing temperature did not influence time to maturity of offspring differently depending on maternal colour morph (Fig. 2).
Heritability of body size
Heritability estimates of body size for male and female offspring reared in the warm and cold treatment are summarized in Table 2. Overall, heritability of body size, as estimated from regression of mean offspring on mother, was moderate in magnitude but did not differ significantly from zero (because of the large SE), with the exception of females reared in the warm treatment (Table 2). A comparison between the two treatments revealed that heritability was about three times higher in the warm (estimates obtained for pooled sexes) compared with the cold treatment, but this difference was not statistically significant, again because of the large standard errors (Table 2). The heritability estimates of body size obtained using full-sib design were significantly greater than zero and higher than those obtained from parent–offspring regression, but of similar magnitude for offspring raised in the two temperature treatments (Table 2). The greater magnitude of the full-sib estimates probably reflects a contribution of common environment, such as food quality or quantity.
Table 2. Heritability [h2 ± 1SE, (n)] of body size (pronotum length) in Tetrix undulata as estimated from least-squares linear regressions of mean offspring size against maternal size and full-sib design, respectively (see text for details). Data for offspring born to wild-caught females and reared in either cold or warm temperature regime in captivity.
Method and sample
*Indicates that heritability was significantly (P < 0.05) greater than zero.
Regression of mean offspring on mother
0.45 ± 0.216 (69)*
0.36 ± 0.240 (60)
0.56 ± 0.362 (63)
0.18 ± 0.404 (51)
Females + males
0.48 ± 0.272 (60)
0.14 ± 0.274 (46)
0.75 ± 0.102 (69)*
0.82 ± 0.123 (64)*
Effects of maternal colour morph on body size
Analyses of data on pronotum length suggest that body size at maturity did not display phenotypic plasticity in response to rearing temperature, but that there is a genetic correlation between colour morph and body size at maturity. Thus, the repeated measures anova uncovered a significant main effect of sex (F1,96 = 393.89, P = 0.0001), with females being larger than males. There was no significant effect of treatment (F1,96 = 0.04, ns), suggesting that body size was not influenced by rearing temperature (Fig. 3). There was, however, a significant main effect of maternal colour morph on offspring body size (F3,96 = 6.47, P = 0.001, Fig. 3), consistent with the hypothesis that a genetic correlation exists between colour morph and body size. Finally, body size of offspring was not significantly influenced by the maternal colour morph by rearing temperature interaction (F3,96 = 0.18, ns), indicating that size of offspring born to mothers belonging to different colour morphs responded similarly to rearing temperature (Fig. 3).
Effects of own colour morph on body size
The analyses reported above revealed an effect of maternal colour morph on offspring body size. The existence of an association between colour morph and body size was further corroborated by analysis in which each individual offspring was regarded as an independent observation, and in which the individuals’ own colour morph (rather than maternal colour morph) was included as an independent variable. The results from a four-way anova with interactions, with own colour morph, sex, family and treatment as independent variables, uncovered significant main effects on body size of family, sex, colour morph and rearing temperature (Table 3). This analysis uses information from all families that produced any mature offspring (n = 69), unlike the repeated measures analysis, in which data was restricted to families that produced mature male and female offspring in both temperature regimes (n = 53). This may partially explain why effects of rearing temperature were seen on body size in these results but not in the repeated measures analysis. The variation in body size among individuals belonging to different colour morphs is shown for each sex and rearing temperature in Fig. 4. In addition to the main effects, there were significant effects of the interactions between own colour morph, sex and family (Table 2), reflecting that the variation in body size among offspring belonging to different colour morphs was different in males and females (Fig. 4), and variable among families. The nonsignificant interaction between morph and rearing temperature (Table 2, Fig. 4) argues against the hypothesis that body size of different colour morphs respond differently to rearing temperature, thus confirming the results from the repeated measures analysis. The significant interaction between family and rearing temperature nevertheless suggests that there were important genotype by environment interaction effects on body size, albeit not associated with colour morph.
Table 3. Results from four-way anova with interactions of body size at maturity when each individuals is regarded as an independent observation and own colour morph is included as an independent variable.
Type 3 SS
The overall model was significant, F321,489 = 19.83, P < 0.0001, R2 = 0.93.
Temperature × morph
Sex × morph
Family × morph
Temperature × family
Sex × family
Moph × sex × family
Association of time to maturity and adult body size
Both theory and empirical evidence suggest that a fast development trades off against small body size. However, our analyses uncovered a positive correlation (across families within treatments) between the average body size of mature offspring and the proportion of individuals that had reached the adult stage at the time of the first census (warm: rs = 0.39, n = 60, P = 0.01; cold: rs = 0.48, n = 46, P < 0.001; Fig. 5), suggesting that individuals that matured quickly also attained a larger body size at maturity. A similar association between time to maturity and size was evident across the four different colour morphs (Fig. 5).
Variation in time to maturity
The strong effect of rearing temperature on the proportion of hatchlings that had reached the adult stage at the time of the first census (Fig. 2) suggests that differences in body temperature experienced by individuals belonging to different colour morphs (Forsman, 1997; Forsman et al., 2002) may translate into morph-specific variation in age and time of maturity. Because of time constraints associated with this seasonal environment, selection is likely to favour genotypes (e.g. colour morphs) that mature quickly (Wiklund et al., 1991; Abrams et al., 1996; Johansson et al., 2001). According to life history theory, however, the fitness benefits associated with an early maturation should be partially offset by a smaller body size, and hence a reduced fecundity (e.g. Atkinson, 1994; Blanckenhorn, 1997; Roff, 2000). The association across families (within each treatment) between proportion mature individuals and average body size suggests that time to maturity was negatively (not positively) correlated with body size (Fig. 5). The fact that the shorter time to maturity in the warm treatment was not accompanied by a smaller adult body size also suggests that the potential fitness benefits associated with early maturation are not necessarily countered by a size-mediated decrease in fecundity in these grasshoppers.
In conclusion, these grasshoppers seem to display plasticity of growth and time to maturity in response to temperature. However, the differences in time to maturity among offspring born to mothers belonging to different colour morphs (in the warm treatment) cannot be attributed to environmental differences, because any variation in temperature or quality among buckets should be independent of maternal colour morph. Regardless whether differences in time to maturity among colour morphs are of direct or indirect genetic origin, they are likely to result in differential fitness among individuals belonging to different morphs, and hence influence the dynamics of the polymorphism.
Variation in body size
Temperature conditions experienced during growth generally have been found to translate into differences in adult body size in ectothermic organisms. In a review of 109 studies on various taxa including animals, bacterium, plants and protists, Atkinson (1994) reported that decreased temperature during growth resulted in increased body size at maturity in 83.5% of the cases and decreased size in 11.9% of cases, while the remaining 4.6% showed mixed responses. The lack of difference in body size between sibling grasshoppers reared under different thermal regimes in the present study thus is a noteworthy result. One possible explanation for the negative result is that the temperature difference between treatments was too small to elicit a detectable response. However we consider this unlikely given the strong influence of temperature on time to maturity. The similarity of adult body size between treatments also argues against the hypothesis that the differences in body size among colour morphs previously observed in natural populations of pygmy grasshoppers (Forsman & Appelqvist, 1999; Forsman, 1999a) are attributable to indirect effects of coloration mediated by differential body temperature. The lack of phenotypic plasticity in body size also fits well with the finding that morph-specific differences in size are repeatable across populations as well as among years (i.e. generations) within populations (Forsman, 1999a). Unlike time to maturity, adult body size in these grasshoppers seems to be canalized against environmental perturbations (Weinig, 2000; Debat & David, 2001), as previously suggested for some species of butterflies (Wiklund et al., 1991; Abrams et al., 1996). Thus, any differences in growth and development because of differences in thermal environment will be manifested as differences in age at maturity rather than as differences in adult body size. This suggests that the fitness costs associated with being of the wrong size (Blanckenhorn, 2000) exceed the potential benefits associated with a change in timing of maturation.
Our results are consistent with the hypothesis of a genetic contribution to variation in adult body size. Thus, heritability of body size (albeit not significantly greater than zero in the case of the parent offspring regression estimate, see Table 2) was within the range normally obtained for morphological characters, indicating that there may exist additive genetic variance for body size at maturity, such that an evolutionary response to selection is possible (e.g. Mousseau & Roff, 1987). The significant influence of maternal colour morph on body size of mature offspring further is in accordance with expectations from a genetic coupling between loci influencing colour pattern and body size, although we cannot dismiss the possibility that the association reflects epigenetic maternal effects (Mousseau & Dingle, 1991). The conclusion that the association is of genetic origin is further corroborated by the significant effect of the offspring's own colour morph on body size (Table 3). Interestingly, the size differences among colour morphs in T. undulata documented here, with the grey morph being the smallest and the brown morph the largest, concur with those previously observed in the con-generic species T. subulata (Forsman & Appelqvist, 1999; Forsman, 1999a). This suggests either that the genetic correlation between colour pattern and size was manifested before divergence of the two species, or that selection has favoured similar combinations of trait-values in both species.
Multiple-trait coevolution, correlational selection and adaptive landscapes
Previous studies of pygmy grasshoppers have revealed morph-specific differences also in clutch size and inter-clutch interval (Forsman, 1999a, 2001), body temperature preferences and sensitivity of performance to temperature change (Forsman, 1999b, 2000), thermoregulatory behaviour (Forsman et al., 2002), anti-predator behaviour (Forsman & Appelqvist, 1998) and microhabitat choice (J. Ahnesjö & A. Forsman, unpublished). For most traits, the association with colour pattern is based on phenotypic correlations only, but available evidence indicate a genetic contribution to the inter-dependence of colour morph and thermoregulatory behaviour (Forsman et al., 2002), and of colour morph and body size (this study). Collectively, this suggests that colour pattern has evolved in concert with physiological, morphological, behavioural and life-history traits and that the functional relatedness of these characters have resulted in integration of genes coding for these traits (see Olson & Miller, 1958).
Functional integration of and genetic coupling between different traits is thought to result from correlational selection, that is, when two or more different traits interactively influence individual performance and fitness (e.g. Clark & Sheppard, 1969; Cheverud, 1982; Lande & Arnold, 1983; Lande, 1984; Endler, 1986, 1995; Zeng, 1988; Brodie, 1989, 1992). The present study does not enable us to identify what mechanisms or selective agent(s) may have been responsible for the covariance between coloration and size, but the following alternatives are possible. Because body size is correlated with reproductive characteristics, morph-specific differences in size may reflect a secondary response to evolution of alternative life-history strategies (Forsman, 1999a, 2001). Alternatively, the variation in size among morphs may be attributable to selection for thermoregulatory capacity, with the relatively small size of the grey morph compensating for the poor ability to achieve body heating associated with pale coloration (Digby, 1955; Stewart & Dixon, 1989; Forsman, 1997; Forsman et al., 2002). Finally, it is possible that the protective value (i.e. relative crypsis) of the different colour patterns is dependent upon body size (Forsman & Appelqvist, 1999, see also Cott, 1940; King, 1992).
The colour polymorphism in these grasshoppers may be symbolized by an adaptive landscape (Wright, 1963, 1988), with the alternative fitness peaks representing optimal combinations of coloration and body size (or some other trait of interest). Under this scenario, the population as a whole will experience disruptive selection, whereas each colour morph will be subject to stabilizing selection. The insensitivity of body size to temperature conditions experienced during development, together with the covariance between colour morph and body size, suggests that selection has promoted morph-specific canalization of size against environmental perturbations (Weinig, 2000; Debat & David, 2001).
We are grateful to M. Björklund, E. Svensson and two anonymous referees for helpful comments on an earlier version of the manuscript, and to G. Karlsson for statistical advice. The study was supported financially by The Swedish Natural Science Research Council (grant to AF), The Swedish Research Council (AF) and Växjö university.