Evolutionary theory predicts an interactive process whereby spatiotemporal environmental heterogeneity will maintain genetic variation, while genetic and phenotypic diversity will buffer populations against stress and allow for fast adaptive evolution in rapidly changing environments. Here, we study color polymorphism patterns in pygmy grasshoppers (Tetrix subulata) and show that the frequency of the melanistic (black) color variant was higher in areas that had been ravaged by fires the previous year than in nonburned habitats, that, in burned areas, the frequency of melanistic grasshoppers dropped from ca. 50% one year after a fire to 30% after four years, and that the variation in frequencies of melanistic individuals among and within populations was genetically based on and represented evolutionary modifications. Dark coloration may confer a selective benefit mediated by enhanced camouflage in recently fire-ravaged areas characterized by blackened visual backgrounds before vegetation has recovered. These findings provide rare evidence for unusually large, extremely rapid adaptive contemporary evolution in replicated natural populations in response to divergent and fluctuating selection associated with spatiotemporal environmental changes.

An improved knowledge of how environmental change affects natural populations of plants and animals is crucial for our understanding of evolution of biological diversity as well as for the development of successful plans for protection and management of biodiversity (Hanski 1998; Lande 1998; Bell 2010). The influence of natural selection on biodiversity in unstable and heterogeneous environments has been debated since the middle of the 20th century. Theory posits that overall, spatially divergent selection in combination with gene flow generally provides broad conditions for maintenance of genetic polymorphism and that fluctuating selection in temporally changing environments may maintain genetic variation under restricted conditions, but the consequences are scale dependent. The effect of spatial variation depends on if the environment is fine or coarse grained relative to the dispersal capacity of the organism, and whether temporally changing environments will promote or erode genetic variation depends on the frequency of change relative to the life span of the organism and on whether generations are overlapping or discrete (Haldane and Jayakar 1963; Levins 1968; Hedrick 1986, 2006; Seger and Brockmann 1987; Frank and Slatkin 1990; Roff 1992, 1997; Ellner 1996; Hanski 1998; Bell 2010). It is generally stated that gene flow may have a homogenizing effect and constrain adaptation to local conditions in heterogeneous environments (Kawecki and Ebert 2004). However, in combination with matching habitat choice, gene flow may contribute to population structuring and enhance evolution of local adaptation (Hanski 1998; Reznick and Ghalambor 2001; Edelaar et al. 2008), and it may do so more rapidly.

Populations may respond to environmental change in different ways, three of which result in a match between phenotype distributions and environmental conditions. First, individuals, or even entire populations, with certain phenotypic attributes may move and settle in areas where conditions are favorable (Edelaar et al. 2008). Second, the development and the resulting phenotype of individuals may show plasticity in response to environmental cues that reliably predict future selective regimes (West-Eberhard 2003). Third, the genetic architecture of populations may undergo micro-evolutionary response to natural selection (Endler 1986).

Evolution of local adaptation has commonly been considered a relatively slow process to accomplish, and there is therefore a broad concern that many populations and species are at risk of extinction in the face of natural and anthropogenic environmental fluctuations, habitat degradation, and fragmentation that result in strong directional selection (Hanski 1998; Lande 1998). Yet, a growing body of evidence indicates that evolution can be more rapid than previously thought (within decades—less than 10 generations) and occasionally very fast (between years or generations) (e.g., Hori 1993; Sinervo and Lively 1996; Reznick et al. 1997; Hendry and Kinnison 1999; Merilä et al. 2001; Reznick and Ghalambor 2001; Thrall and Burdon 2003; Millien 2006; Koskella and Lively 2007; Bell and Gonzales 2009; Gingerich 2009; Thompson 2009; Ozgo 2011). However, many studies measure phenotypic change without documenting the genetic and environmental component of that change and hence cannot rule out phenotypic plasticity (Hendry and Kinnison 1999; Merilä et al. 2001; Gienapp et al. 2008). This caveat set aside, reviews of contemporary adaptations in natural populations (e.g., Hendry and Kinnison 1999; Reznick and Ghalambor 2001; Gienapp et al. 2008; Gingerich 2009) suggest that rapid evolution is often associated with colonization and establishment in novel or modified environments, and with heterogeneous environments and a meta-population (Hanski 1998) structure. Both these settings are associated with changes in food resources, the competitor, and predator community composition, and in other aspects of the biophysical environment that results in strong directional selection and provides at least a short-term opportunity for population growth, suggesting that these ecological conditions may promote rapid evolution (Reznick and Ghalambor 2001; Gingerich 2009; Bell 2010). In this context, populations that are genetically and phenotypically more diverse are likely better able to colonize novel habitats and to withstand and adapt to changing environments (Lande 1998; Forsman et al. 2008; Hughes et al. 2008; Bell 2010).

Unfortunately, most reported cases of contemporary evolution in the wild are not replicated, limiting our ability to infer the importance of history, chance, natural selection, and sampling error. Natural and anthropogenic disturbance events—such as forest fires and industrial pollution—provide natural experiments that can be used to investigate the ecological and evolutionary consequences of environmental change (Kettlewell 1973; Majerus 1998; Reznick and Ghalambor 2001; Grant and Wiseman 2002). For instance, the blackened visual backgrounds and hot surface temperatures, due to high absorbtion of solar radiation, that characterize fire-ravaged areas before vegetation has recovered, may influence body temperatures, water balance and camouflage of differently colored individuals (Cott 1940; Hocking 1964; Rowell 1971; Forsman and Appelqvist 1999; Ahnesjö and Forsman 2006; Parkash et al. 2010; Thomas and McAlpine 2010). Here we present data from natural and captive populations of the color polymorphic pygmy grasshopper Tetrix subulata and demonstrate exceptionally rapid and replicated contemporary adaptation in response to postfire environmental change. Pygmy grasshoppers provide one of the classic examples of color polymorphism. For almost a century, intraspecific variation has been documented and genetically analyzed (Nabours 1929). Pygmy grasshoppers are small (up to 15-mm body length, average dry body mass ca. 0.07 g), diurnal, ground dwelling insects that inhabit biomes ranging from tropical rainforests to arctic regions of Europe, Asia, and America. Ground colors range from black, via various shades of brown to light gray, and some morphs are monochrome whereas others have patterning consisting of longitudinal stripes, or specks or spots of variable colors and widths. These alternative color variants of Tetrix also differ in morphology, physiology, behavior, and life-history traits (Forsman 2001; Forsman et al. 2002; Ahnesjö and Forsman 2006), so they are eco-morphs. Individuals with different trait-value combinations vary in their ability to cope with and exploit environmental resources, and shuttle between microhabitats to meet their specific demands regarding body temperature regulation, moisture, food, and protection from predators, such that they occupy on the average different niches (Forsman et al. 2002; Ahnesjö and Forsman 2006). Circumstantial evidence indicate an increased frequency of black and dark phenotypes in populations of pygmy grasshoppers and other insects that occupy recently fire-ravaged areas (Majerus 1998; Karlsson et al. 2008). Proposed mechanisms behind shifts in melansim frequencies include directional selection for camouflage, correlated responses to selection on traits that are associated with color pattern, developmental plasticity, or differential immigration, but no systematic comparison between several populations in burned versus nonburned areas has been previously conducted to evaluate the generality of the fire melanism hypothesis and the underlying causes.

We compared frequencies of black phenotypes in several T. subulata pygmy grasshopper populations that inhabited areas recently burned by fire with those in populations that inhabited nonburned areas. We also compared frequencies of black phenotypes among years to examine if the biotic and abiotic environmental changes associated with succession in postfire habitats (Thomas and McAlpine 2010) cause temporal shifts in the frequency of black phenotypes within populations. We performed common garden breeding experiments to determine if the differences seen among and within populations represented genetically determined evolutionary modifications, rather than developmental plasticity. Finally, we evaluate the hypothesis that the increased frequency of the melanistic form following fires is driven by differential immigration and matching habitat choice. Overall, our results provide rare evidence for extremely rapid adaptive contemporary evolution of large magnitude in replicated natural populations in response to fluctuating selective regimes.

Material and Methods


Tetrix subulata pygmy grasshoppers occupy damper microhabitats in relatively open areas (e.g., clear cuttings, shore meadows, areas burned by fire) where they feed on moss, algae, and humus. Adult and late instars nymphs hibernate during winter and emerge in April-May when reproduction ensues. Females survive at most one reproductive season, produce multiple pods of egg (< 35 eggs/clutch), and nymphs develop through five (males) or six (females) instars before eclosing. Tetrix subulata usually occurs in low densities but may become very numerous when conditions are favorable. The higher reproductive rate, shorter incubation time, and faster development rate associated with high temperatures may contribute to a faster population growth in burned areas with elevated surface temperatures (Forsman 2001; Ahnesjö and Forsman 2003, 2006).

Like many other insects some pygmy grasshopper species, including T. subulata, are wing dimorphic, viz there are both winged and wingless (or short-winged) individuals (Nabours 1929; Harrison 1980; Roff and Fairbairn 2001). The wings correspond in length, generally, with that of the pronotum that covers them (Nabours 1929). Only macropterous individuals with a long pronotum that extends well beyond the posterior point of the hind femur have fully developed functional wings (A. Forsman, pers. obs.). Performance trials have shown that individuals experimentally subjected to predation jump longer distances and are more likely to use their wings at high compared to low temperature, and that they move ca 50% farther when flapping their wings then when jumping without using their wings (Forsman 1999). Free-ranging undisturbed individuals usually walk slowly on the surface of the ground and occasionally perform short (< 0.2 m) jumps, but may jump longer distances or perform escape flights of up to 15 m when threatened. Mark–recapture data of free-ranging individuals indicate that T. subulata are sedentary animals that normally move only a few meters per day (average distance moved in four days was 4.6 m in males and 7.4 m for females) (Forsman and Appelqvist 1999; Caesar et al. 2007). However, long winged individuals experimentally released in unfavorable habitats may resort to active flights of at least 75 m (A. Forsman, pers. obs.), which may indicate a capacity for long-distance dispersal (Harrison 1980).


To quantify spatial and temporal patterns of variation in melanism frequency we collected 5058 T. subulata (1492 males and 3566 females) between 1995 and 2009 from 20 natural populations in either burned (N= 9) or nonburned (N= 11) areas in Sweden separated by almost 4o, or approximately 450 km north to south (Table 1). Sampling was performed from March–July when grasshoppers were reproductively active and vegetation allowed easy sampling. We searched for grasshoppers during days with suitable weather conditions (clear to overcast skies, air temperatures > 15 °C, low-to-moderate winds) while walking slowly through the area (Ahnesjö and Forsman 2006). Tetrix subulata does not have a uniform or random spatial distribution, and we therefore initially searched the entire areas and then concentrated our search and capture effort to those parts and microhabitats (i.e., humid bare soil, areas covered by mosses) that offered suitable conditions for pygmy grasshoppers (Ahnesjö and Forsman 2006). Because T. subulata predominantly move around on the ground surface and rarely climb vegetation they are difficult to capture with a bag net, and we therefore acted like visual predators and captured individuals that we could see (Forsman and Appelqvist 1999; Ahnesjö and Forsman 2006). We discuss the possible consequences of biased sampling due to differences in capture probability among color morphs for our results in the Discussion section. Individuals were classified by sex and color pattern as either melanistic (monochrome black and very dark individuals) or nonmelanistic (all other color morphs). A more detailed description and photographs of alternative color morphs are available elsewhere (Karlsson et al. 2008; Caesar et al. 2010).

Table 1.  Descriptive characteristics of Tetrix subulata populations sampled in recently burned or nonburned areas in Sweden and number of individuals captured (Ntot) and number of melanistic individuals capture (Nblack) during the period 1995 to 2009.
PopulationLat. Long. CoordinatesBurned or NonburnedType of habitatYear sampledYear postfireNtotNblack
AlmungeN59°53.049′, E18°03.813′BurnedClear cut (pine), managed fire, ca 20 ha19981171 68
FlyvägenN57°00.268′, E15°16.752′BurnedClear cut (spruce), managed fire, ca 5 ha20091136 32
HovmantorpN56°47.169′, E15°10.053′BurnedForest (spruce), natural fire, clear cut, ca 15 ha20091114 65
KostaN56°51.308′, E15°35.417′BurnedClear cut (pine), managed fire, ca 50 ha20052 98 46
=20063221 70
    20074 96 31
NässjönN56°53.739′, E15°18.775′BurnedForest (spruce), natural fire, clear cut, ca 20 ha20081 38 21
    20092 41 14
NorbergN60°03.309′, E15° 58.030′BurnedYoung forest (spruce), natural fire, ca 2 ha19951 20  9
SättrabyN59°50.427′, E18° 25.229′BurnedClear cut (pine), managed fire, ca 10 ha19952229 92
SävsjöströmN57°00.678′, E15°26.585′BurnedClear cut (pine), managed fire, ca 10 ha20032759300
    20043347 96
UttersbergN59°44.638′, E15°39.439′BurnedClear cut (pine), managed fire, ca 10 ha19951 50 29
    19962 45 22
=19973 12  5
ÅlemN56°56.019′, E16°21.808′NonburnedCultivated area adjacent to clear cut2009na264  0
AspelundN56°33.226′, E16°01.468′NonburnedPasture nearby small stream2008na 47  5
BredsätraN56°50.481′, E16°47.346′NonburnedPasture nearby man-made pond2008na122 16
=2009na192  1
DammenN59°50.298′, E18°25.275′NonburnedShorelines of man-made pond1996na178 28
=1997na229 43
Hägern, BoarumN57° 25.380′, E16°15.900′NonburnedPasture nearby burned area2009na104 14
In FredelnN59° 31.707′, E19°19.225′NonburnedCoastal meadow, small island1997na 36  6
JordtorpN56°40.622′, E16°33.303′NonburnedPasture and alkaline fen2007na 36  0
=2009na234  5
Läckeby, KvillaN56°43.953′, E15°10.546′NonburnedClear cut (spruce)2007na 42  4
LinnerydN56°44.496′, E15°10.868′NonburnedClear cut (spruce)2007na 26  4
SävsjönN56°32.268′, E15°48.606′NonburnedPasture and pond2009na106  0
VanserumbäckN56°40.457′, E16°38.338′NonburnedPasture2007na 46  8


To determine if the differences seen among and within populations represented genetically determined evolutionary modifications, rather than developmental plasticity, we performed a common garden experiment. We used a total of 2070 captive reared individuals, representing 163 families produced by wild-caught females from two populations in burned areas (Kosta 2005, Kosta 2006 and Sävsjöström 2004), and 104 families from four populations in nonburned areas (Ålem 2009, Bredsätra 2009, Hägern 2009 and Jordtorp 2009) (Table 1 and Table S1). Females were captured in spring, housed individually for egg-laying in the laboratory and egg pods were placed on moist cotton inside a plastic Petri dish for incubation. Newly hatched nymphs were housed by family in separate 10-L plastic buckets that contained a peat-soil mixture and common hair moss (Polytrichum commune), were maintained out-of-doors and watered at regular intervals (Karlsson et al. 2008). We examined the rearing buckets by the end of summer, counted and classified individuals for color morph as either melanistic or nonmelanistic. The mortality rate in the common garden experiment was about 50%, which is similar to what we have found in previous studies (Forsman et al. 2007; Caesar et al. 2010).


To evaluate if differential dispersal capacity, immigration, and matching habitat choice (Edelaar et al. 2008) may contribute to an increased frequency of the melanistic form in postfire environments, we used data from a natural population that inhabited an area outside Påryd in southeastern Sweden, that was burned off by a large (ca 125 ha) natural forest fire during summer (24–28 June) 2009. In theory, a higher immigration rate of melanistic than nonmelanistic phenotypes may arise if there is an excess of melanistic individuals in the source population, or if the melanistic phenotype is more prone to disperse and settle in recently burned environments. We first visited the area and searched for grasshoppers in September 1, 2009, two months after the fire, and found only two individuals (1 brown male, 1 gray female). We revisited the same area the following spring (May 27, 2010) and collected 85 adult male and female T. subulata from five different sampling sites. Individuals were classified by color morph (melanistic or nonmelanistic) and dispersal phenotype (either long pronotum and long hind wings or short pronotum and reduced wings or wingless) (Harrison 1980). We then compared the incidence of the long winged form between melanistic and nonmelanistic individuals, computed 95% confidence intervals and evaluated the null hypothesis that dispersal phenotype was independent of color morph using the Chi-square test (Zar 1974). If the melanistic phenotype is more prone than nonmelanistic phenotypes to disperse and settle in recently burned areas, we expect the incidence of long winged (potential immigrants) phenotypes to be higher in melanistic than in nonmelanistic individuals.


Because the proportion of male and female individuals that were melanistic was highly correlated across samples from different populations and years (correlation on arcsine-square root transformed proportions, r= 0.83, n= 31, P < 0.0001) we pooled data for the two sexes in subsequent analyses of data. To test if the percentage of the melanistic morph in the six populations that were sampled one year after a fire was different from that in the 11 populations in nonburned habitats we used a t-test (on arcsine-square root transformed proportions) after pooling data from different years for the three nonburned populations for which repeated samples were available (Table 1). To test if the percentage of the melanistic morph declined over time within populations during habitat succession associated with recovery from fire we performed a general linear mixed model implemented using procedure MIXED in SAS (Littell et al. 2006; Bolker et al. 2009). In this approach, we treated arcsine-square root transformed proportions melanistic individuals as the dependent variable, year-since-fire as a fixed effect and population as random effect. We used the Kenward–Roger method to approximate degrees of freedom (Bolker et al. 2009). To assess statistical significance of random effects, we used the log-likelihood ratio test with one degree of freedom per random effect (Littell et al. 2006; Bolker et al. 2009).

Data from the common garden experiment were used to test if fire regime of the source population could explain the distribution of melanistic and nonmelanistic of the captive reared individuals and if family explained a relevant part of the total variation. Analyses of the incidence of melanistic individuals among offspring in different families that descended from individuals collected in populations inhabiting burned versus nonburned areas were performed using generalized linear mixed models (GLMMs) implemented using procedure GLIMMIX in SAS (Littell et al. 2006; Bolker et al. 2009). The color pattern (melanistic or nonmelanistic) of individuals within families was treated as the dependent variable, fire regime (burned vs. nonburned habitat) was treated as a fixed effect, and family as a random effect. We used the Kenward–Roger method to approximate degrees of freedom (Bolker et al. 2009).



As expected, the percentage of individuals that belonged to the black color morph was much higher on average in populations in recently burned areas (data for first year after the fire, mean = 46%, range: 23–58%, n= 6 populations) than in populations in nonburned areas (mean = 9%, range: 0–17%, n= 11 populations) (t-test on arcsine-square root transformed proportions, t= 5.87, df = 15, P < 0.0001) (Fig. 1). This result indicates rapid, independent, parallel evolution of melanism in postfire environments.

Figure 1.

The incidence of black pygmy grasshoppers is higher in burned than in nonburned areas and changes faster than in the classical example of industrial melanism in the Peppered Moth. (A) Mean percentage of individuals that are melanistic for 11 populations in nonburned areas and for nine populations in burned areas sampled at different times after the fire event. Black dots connected with lines indicate samples from successive years in the same population. Red squares represent mean values across populations within each year category, red line denotes fitted linear regression of yearly means. (B) Changes in frequency of melanistic phenotypes associated with fire melanism in Tetrix subulata (this study) compared with changes associated with industrial melanism in Biston betularia in Manchester and Liverpool in the United Kingdom (data in Grant et al. 1995) and in Michigan USA (data in Grant and Wiseman 2002). For T. subulata year was arbitrarily set to 2003–2009; see Table 1 for authentic year of sampling. See text for references to previously published data on B. betularia. Picture of B. betularia taken by Olaf Leillinger at 2006–06-13, License CC-BY-SA-2.5 and GNU FDL. Picture of T. subulata taken by A. Forsman.


Our data for populations in burned areas consisted of samples collected one to four years after the fire event (Table 1). As expected, the percentage of individuals belonging to the black morph rapidly and consistently declined from about 50% on average one year after a fire to about 33% on average four years after a fire (linear mixed model ANCOVA of arcsine-square root transformed proportions, effect of years since fire (fixed): F1,9.94= 8.81, P= 0.014; effect of source population (random): χ2= 2.1, df = 1, 0.1 < P < 0.25, Z= 1.29, P= 0.098)(Fig. 1). No such consistent temporal change in frequency of the melanistic morph was evident in the three populations from nonburned areas that were sampled in two different years (mean change =−2.5%, range to 2.1 to −12.59%, n= 3 comparisons, paired t-test based on arcsine-square root transformed proportions, t= 0.49, P= 0.67).


Because the results above are based on data from grasshoppers collected in the wild, a critical question is whether the higher incidence of melanism seen in burned versus nonburned areas represented genetically based micro-evolutionary response (Merilä et al. 2001). Data on color pattern for the 2070 captive reared individuals included in our common garden experiment (Table S1) confirmed that the observed differences among populations are likely to be of genetic origin; the incidence of melanistic individuals was lower among laboratory-born captive reared descendents from four populations in nonburned areas (mean percentage of melanistic individuals across 104 families = 3.3%, range: 0–66%n= 1007 individuals) as compared with descendents from three samples from populations in burned areas (mean percentage of melanistic individuals across 163 families = 27%, range: 0–100%, n= 1063 individuals; effect of fire-regime when data were analyzed using GLMM implemented using procedure GLIMMIX and treating individuals rather than family means as datapoints; F1,5.15= 19.06, P= 0.0068; Fig. 2). It is noteworthy that the incidence of the melanistic morph was higher among descendants from populations in burned areas despite their mothers having been collected three or four years after the fire and well after the peak of melanism frequency (Fig. 1).

Figure 2.

The incidence of melanistic individuals is higher among captive reared individuals descending from populations inhabiting burned versus nonburned areas. Data for captive born individuals originating from different populations reared in a common environment from the time of hatching until maturity. Populations from nonburned environments were Ålem, Bredsätra, Jordtorp and Hägern. Populations from burned areas were Kosta 3 years post fire, Kosta 4 years post fire, and Sävsjöström 4 years post fire. Figure is based on data for family means. The thick and thin lines within the box indicate median and mean, the boundaries of the box indicate 25th and 75th percentiles, whiskers below and above indicate 10th and 90th percentiles and dots indicate outlying observations. Numbers below indicate number of families.

A resemblance among relatives consistent with a genetic effect on the expression of melanism was evident also at the level of populations. There was a strong association across populations between the incidence of the melanistic morph among samples of captive reared individuals and wild-caught individuals from the parental generation in the corresponding populations (least-squares linear regression: F1, 5= 21.76, P= 0.005, r= 0.90, n= 7, P= 0.005) (Fig. 3).

Figure 3.

The incidence of melanistic individuals among offspring from different populations reared in a common environment is associated with that in the parental generation. Data for captive born individuals originating from different populations reared in a common environment from the time of hatching until maturity and for wild-caught individuals from the same populations. Data for wild-caught individuals include all males and females in the sample, not only the parents of the captive reared animals. Data for populations in burned (gray filled symbols) and nonburned (open symbols) environments.


To evaluate if the incredibly fast increase seen in the frequency of the melanistic form following fires was caused by differential immigration and matching habitat choice, we compared the frequency of long winged phenotypes that may represent long-distance immigrants (Harrison 1980) in melanistic versus nonmelanistic color morphs in our sample collected in spring from a population that inhabited an area that was burned off by a natural fire the previous summer. About 40% (33 of 85) of the individuals were melanistic, and about half (55%, 47 of 85) belonged to the long winged phenotype. The incidence of long winged phenotypes did not differ between melanistic and nonmelanistic individuals (χ2= 0.61, df = 1, P= 0.43) (Fig. 4), indicating that differential immigration was not an important driver of changes in morph frequencies.

Figure 4.

The percentage long winged individuals within a population inhabiting a newly burned area is similar in melanistic and nonmelanistic colour morphs. Data for 85 wild-caught adult male and female T. subulata collected in spring (May 27, 2010) from five different sites in an area that was burned off by natural fire the previous summer (24–26 June, 2009). Error bars indicate 95% confidence intervals. Numbers below indicate number of individuals.


Do changes in the environment and selection regime associated with recovery from fire promote rapid evolutionary shifts of color morph frequencies, as predicted by theory? Across 20 populations of T. subulata, we found that the percentage of individuals that belonged to the melanistic (black) color morph was much higher on average in populations in recently burned areas than in populations in nonburned areas and, in burned areas, the frequency of melanistic grasshoppers dropped from ca. 50% one year after a fire to 30% after four years. A common garden breeding experiment confirmed that the variation in the frequency of melanistic individuals among families and populations had a genetic component, indicating that the observed population divergence reflects rapid evolutionary adaptation.


That the phenotypic expression of the melanistic form is at least partially heritable is supported by comparisons within as well as among populations. We have previously demonstrated that within populations, melanistic T. subulata mothers produce a higher percentage of black offspring than do nonmelanistic mothers (Karlsson et al. 2008; Karlsson and Forsman 2010). Our present analyses based on comparisons within the common garden samples also uncovered variation in the incidence of melanism among families. In addition, there was a strong association between the incidence of melanism in wild caught and captive reared individuals across different populations (Fig. 3). The design of our common garden experiment does not enable us to control for or estimate effects of maternal or paternal environmental conditions experienced prior to egg-laying or during early ontogeny that may have influenced the phenotypic expression of melanism. It therefore remains uncertain whether the variation in melanism frequency among families and populations reflects direct genetic determination. However, melanism does not seem to be induced by developmental plasticity in response to environmental cues after hatching (Karlsson et al. 2009; Karlsson and Forsman 2010).

Data for individuals that were experimentally raised using a split-brood design on either crushed charcoal with chemical and physical properties associated with fire-ravaged areas versus white aquarium gravel indicate that there is no plasticity of either color pattern or of overall darkness of coloration in response to rearing substrate, but a strong resemblance between maternal and offspring color patterns (Karlsson et al. 2009). Hochkirch et al. (2008) arrived at the opposite conclusion that there is a plasticity effect of substrate darkness on coloration in T. subulata. However, the experimental protocol used by Hochkirch et al. (2008) allows for several alternative explanations and therefore provides no firm evidence for plasticity, and their results could not be replicated in a later experiment that was specifically designed to eliminate other sources of variation (Karlsson et al. 2009). Color pattern in T. subulata also is not influenced by rearing density (Karlsson and Forsman 2010), and studies of other pygmy grasshopper species have found no evidence of plasticity of coloration in response to other environmental cues (Nabours 1929; Ahnesjö and Forsman 2003; Forsman 2011). Together, these findings add support to the interpretation that the higher incidence of melanistic individuals seen in recently burned areas, as well as the subsequent declining incidence of the melanistic morph over time in post-fire environments (Fig. 1), reflect fast, large, and adaptive evolutionary modifications of the genetic composition of populations in response to strong divergent and fluctuating selection.


Because we sampled populations in different years during a 15-year period (1995–2009) (Table 1), the initial rapid increase and subsequent decline in melanism frequency in recently burned areas are best explained by parallel microevolutionary modification driven by changes in the selective regime associated with postfire environments, rather than by yearly variation in other environmental factors such as weather conditions. With only two populations, environments are likely to differ in so many respects that it is impossible to identify the factor(s) responsible for an observed change or difference and to eliminate sampling error. A major strength of our study in this context is that the changes were repeated among and within several replicated populations. This enables us to infer a causal link between postfire environmental conditions and evolutionary modifications of melanism frequencies (Kawecki and Ebert 2004).


It is theoretically possible that the incredibly rapid increases seen in the frequency of the melanistic form immediately following fires were partly driven by differential immigration and matching habitat choice (Edelaar et al. 2008). However, for phenotype-dependent movements to play an important role in structuring populations in recently burned areas, the melanistic color form must be overrepresented among long winged individuals that comprise potential immigrants (Harrison 1980). That the incidence of long winged phenotypes did not differ between melanistic and nonmelanistic individuals shortly after a fire event (Fig. 4) therefore argues against an important effect of differential immigration. The observed changes in melanism frequencies also are not readily explained solely by continuous immigration followed by differential morph-specific survival, rather than by within-population changes, because such an immigration/differential-mortality balance requires that there exists adjacent source populations with relatively high melanism frequencies. Given the low melanism frequencies in nonburned areas (Table 1) this seems unlikely. For the same reason, it is unlikely that melanism frequencies increase rapidly in recently burned areas because individuals on the edge of habitat boundaries preferentially disperse into matching habitats.

It is theoretically possible, but unlikely, that spatial and temporal variation in melanism frequencies were caused by biased sampling. Estimates of capture probabilities based on recapture histories of marked T. subulata individuals that had been painted gray, black, or striped in a population that inhabited an area that had been burned by fire two years earlier (Sävsjöström, Table 1) have shown that gray individuals were captured at a higher rate than striped or black individuals (Forsman and Appelqvist 1999). Because black individuals are more difficult to detect and capture in burned environments, a possible consequence of biased sampling is that melanism frequencies in the most recently burned environments were higher (not lower) than indicated by our results, and that we therefore may have underestimated both the rate and the magnitude of the changes in melanism frequencies shown in Fig. 1.


Of the potential drivers of evolutionary modifications (selection, gene flow, drift, and mutation), only selection can explain the consistent direction, fast rate, and large magnitude of the changes in melanism frequency observed in these replicated pygmy grasshopper populations. We consider directional selection for camouflage mediated by differential predation (Isley 1938; Forsman and Appelqvist 1999; Mullen et al. 2009) to be one of the most plausible selective agents. Their small size and locally high population densities render grasshoppers susceptible to visual predators such as birds (Isley 1938; Bock et al. 1992) and lizards (Civantos et al. 2004). Furthermore, Tsurui et al. (2010), using humans as dummy predators, demonstrated that differences in relative crypsis among alternative T. japonica pygmy grasshopper color morphs change across different visual backgrounds (sand vs. grass). There are also experimental demonstrations of phenotype-dependent predation in pygmy grasshoppers—including our own previous work based on manipulation of color patterns and its effect on subsequent survival of free-ranging T. subulata individuals under natural conditions in postfire environments (population Sävsjöström, Table 1) as well as staged predator–prey encounters in the laboratory—that color pattern influences survival and that the protective value of a given color pattern may depend on the sex and movement patterns, as well as on the visual properties of the microhabitats used by individuals (Forsman and Appelqvist 1998, 1999; Civantos et al. 2004). We therefore propose that the rise and fall of the melanistic color variant is driven at least in part by differential predation and selection for camouflage, and that the protective value of black coloration against the blackened background that characterizes burned environments gradually changes when vegetation recovers and alters the visual properties of the habitat.

Selective mechanisms in addition to camouflage might have contributed to the increased and decreased melanism frequencies in fire ravaged areas. Correlated responses to selection on traits that are genetically, developmentally or functionally associated with color pattern (True 2003; McKinnon and Pierotti 2010) is one possibility. Alternative pygmy grasshopper color morphs differ in a suite of morphological, physiological, behavioral, and life-history traits (Forsman and Appelqvist 1999; Forsman 2001; Forsman et al. 2002; Ahnesjö and Forsman 2003, 2006). The changes in melanism frequencies may therefore have been influenced also by correlated responses to selection on traits other than coloration per se, such as enhanced resistance to hot temperatures (Forsman 2011). Due to high absorbance of solar radiation, surface temperatures in areas covered with burnt-off material can be very high (Ahnesjö and Forsman 2006; Thomas and McAlpine 2010) and small ground-dwelling insects shuttle between microhabitats with different thermal properties to avoid overheating (Heinrich 1993; Forsman et al. 2002; Ahnesjö and Forsman 2006). In this context, their different thermal physiology and morphology may render melanistic individuals less vulnerable to heat stress and desiccation (Forsman et al. 2002; Parkash et al. 2010; Forsman 2011) (but see Civantos et al. 2005b). It has also been suggested that melanistic individuals benefit from an improved physical barrier against infection, wound healing, cellular innate immunity, and parasite resistance (Protas and Patel 2008; Mikkola and Rantala 2010), but these mechanisms do not seem to be important in pygmy grasshoppers (Civantos et al. 2005a).


Our present findings add further support to the growing body of evidence that evolution may be much faster than previously recognized (Reznick and Ghalambor 2001; Millien 2006; Gingerich 2009). Our comparisons across populations suggest a fivefold increase from about 9% to 50% in the frequency of the melanistic morph between consecutive generations, and comparisons within populations demonstrate a drop from about 50% to 30% on average in a 3-year time interval (Fig. 1A). To this end, it is interesting to put our findings in the context of the classical example of microevolution of industrial melanism in the Peppered Moth, Biston betularia. The evolutionary shifts of melanism frequencies in these grasshopper populations were of smaller magnitude, but occurred at a much faster rate than the rise and fall of the melanistic morph of B. betularia associated with atmospheric pollution and lichen succession in the UK and USA (Kettlewell 1973; Grant et al. 1995; Majerus 1998; Grant and Wiseman 2002; Fig. 1B). Another difference is that the changes in pygmy grasshoppers were replicated in several populations sampled during a 15-year period, and there therefore remains little doubt that these evolutionary shifts have been driven by changes in the selective environment created by recent fire.

The evolutionary shifts in the frequency of the melanistic form following fires in these grasshoppers are so rapid that they may at first seem difficult to explain by local selection. However, burned matter and charcoal covering the ground in post fire environments changes the visual background into almost uniform black, probably resulting in extremely strong directional selection in favor of the well-camouflaged melanistic phenotype. Furthermore, a single T. subulata female may produce ca 50 nymphs in only two weeks (25 eggs/clutch × 70% hatching rate × 3 clutches in 15 days) (Forsman 2001) and parent offspring resemblance is high (ca 50% of offspring produced by melanistic mothers are melanistic) (Karlsson et al. 2008). It is therefore theoretically possible that a few mostly melanistic survivors (or immigrants) in a newly burned off area may give rise to hundreds of mostly black descendants the following year. It has recently been argued (Bell 2010) that because selection intensities in the wild are sometimes much higher than assumed in most population genetic models of the maintenance of polymorphism, it may be necessary to question the validity and generality of the conclusions emanating from such models. Our present results adhere to this notion and indicate that pygmy grasshoppers may offer a suitable model system for evaluating these models in the future.


In conclusion, our findings provide evidence for extraordinarily fast (between successive years rather than within decades or centuries; Reznick and Ghalambor 2001; Bell 2010) contemporary adaptive evolution of melanism frequencies in wild replicated populations of pygmy grasshoppers in response to a fluctuating selective regime associated with fire events. These results highlight the continued value of natural and anthropogenic disturbances as natural experiments in enhancing our understanding of evolution and the maintenance of biological diversity, especially in respect to animal color patterns (e.g., Majerus 1998; Protas and Patel 2008; Mullen et al. 2009). Fire melanism in pygmy grasshoppers provides a candidate representative of a meta-population context mosaic nature of adaptation (Hanski 1998; Reznick and Ghalambor 2001), whereby subdivided populations are exposed to multiscale heterogeneous environmental conditions imposed by natural and anthropogenic fires that rapidly modify habitat patches and thereby create opportunities for repeated colonizations within the existing range distribution, rapid population growth, and intense directional selection on color pattern or on traits associated with coloration. Our findings are in agreement with the notion (Lande 1998; Forsman et al. 2008; Hughes et al. 2008) that populations that are genetically and phenotypically more diverse may contain preadapted phenotypes and therefore are likely to better withstand and adapt to rapidly changing environments, (re-)colonize novel or modified habitats and expand their distribution ranges.

Associate Editor: B. C. Sheldon


We are indebted to S. Appelqvist, J. Ahnesjö, S. Caesar, E. Civantos, C. Kindblom, M. Forsman, S. Forsman, and P. Tibblin for assistance in the field and laboratory. J. Merilä, J. Thompson, and two anonymous reviewers commented on the manuscript. The study was supported by The Swedish Science Council, The Swedish Research Council Formas (grants to AF), University of Kalmar and Linnaeus University.