Translator Disclaimer
1 December 2008 The strength of temperature-mediated selection on body size in a wild insect population
Author Affiliations +

Body size is a crucial element of an organism's ecology and evolution, and is known to be influenced by natural selection. Here, we estimated the strength of temperature-mediated selection on body size in a wild population of the striped ground cricket, Allonemobius socius. We found that extremely high, but naturally occurring temperatures, selected for larger body sizes, as expected. The strength of this selection pressure was strong, with the univariate selection gradients being β = 0.591 ± 0.237; βσ = 0.281 ± 0.112; and βμ = 3.95 ± 1.58 (± s). These results suggest that periodically occurring temperature extremes can influence the evolution of body size in wild populations of A. socius and other orthopterans. Furthermore, these selection pressures may contribute to the often observed size cline known as the Converse of Bergmann's rule, and become more prevalent in the future as temperatures increase due to global warming.


Body size has a tremendous influence on the ecology and evolution of natural populations. Body size affects the biochemistry, physiology, morphology, life-history, behavior, distribution, demography, and competitive ability of most, if not all, known species (see reviews in Bonner 2006, Brown et al. 2004, Roff 2001, Sibly & Brown 2007). These effects, as well as body size's high levels of heritable genetic variation (Mousseau & Roff 1987, Roff & Mousseau 1987) have made it one of the most intensely studied traits in biology. And yet, biologists still struggle to identify the mechanisms that determine body size in natural populations (Atkinson & Sibly 1997, Mousseau 1997, van Voorhies 1996).

One mechanism known to have a tremendous effect on body size in wild populations is temperature-mediated selection (Bodie & Semlitsch 2000, Brown & Brown 1999, Grant & Grant 1995). Temperature-mediated selection pressures are particularly important to orthopterans and other ectotherms which lack physiological mechanisms for thermal regulation. For example, desiccation caused by extremely high temperature selects for larger individuals in many ectothermic species (Addo-Bediako et al. 2001, Chown & Gaston 1999, Levins 1969, Schoener & Janzen 1968). Larger individuals are better able to resist desiccation than smaller individuals because they have a greater capacity to store water (Chown 1993) and a smaller surface area-to-volume ratio, which reduces the rate of water loss (Chown et al. 1998). As such, natural selection tends to favor increased body size in populations of ectotherms subjected to extremely high temperatures.

The importance of estimating the strength of selection pressures under natural conditions has long been recognized (Endler 1986), and a large number of these estimates have been published (Hoekstra et al. 2001, Irschick et al. 2008, Kingsolver et al. 2001). However, these previous works have focused on a relatively small number of species. For example, while Kingsolver et al. (2001) presented over 1,500 estimates for the strength of selection, these estimates were only distributed across 61 species, of which 15 were invertebrates, including one orthopteran (Zuk 1988). This relatively modest sampling of species, despite the robust overall sample size, suggests that while our understanding of natural selection in the wild has increased greatly over the past several years, more work needs to be done.

In this study, we estimated the strength of temperature-mediated selection acting on body size in a wild population of the striped ground cricket, Allonemobius socius. These small-bodied orthopterans are found throughout the eastern United States (Marshall 2004) and are frequently used in ecological and evolutionary studies (Fedorka et al. 2007, Winterhalter & Mousseau 2007). We collected a population of these crickets from Newberry County, South Carolina, USA (lat 34°N, long 81°W) on August 1, 1999. On this particular date, temperatures in the area reached 40.5°C (Williams et al. 2006), which was more than one standard deviation above the average high temperature for that time of year and marked the beginning of several days of extreme heat in the area (Fig. 1). Adult crickets were captured individually (no nymphs were observed), using an insect net. Upon capture, individuals were placed in a shaded plastic collection bin (36.2 × 24.1 × 31.1 cm) for the remainder of the collection period (~1.5 h in total). While in this container, 59 of the 103 individuals died (57.3% mortality). Based on the general ability of large-bodied individuals to resist desiccation (Addo-Bediako et al. 2001, Chown & Gaston 1999, Levins 1969, Schoener & Janzen 1968), we predicted that larger individuals would have a higher probability of surviving this mortality event.

Fig 1.

Daily high temperatures (°C) that occurred in Newberry County, SC in 1999 (points) relative to the average high temperatures (± s) for that time of year. The initial collection took place on August 1, 1999 and the second collection took place on August 3rd of the same year.



After the collection was completed, we brought all individuals (both living and dead) back to the laboratory. Here, digital images of the crickets' femurs were taken using a Leica LW Scientific MiniVID digital camera mounted to a Leica Zoom 2000 stereoscope. The length of each femur was then measured from these digital images using tpsDIG (Rohlf 2006). We used the average of the left and right femur as our estimate of body size. Because insects have determinant growth, length is generally a more consistent proxy of body size than mass. Furthermore, femur length has been used as a proxy for body size in this system (Mousseau & Roff 1989).

In order to determine the relationship between body size and mortality, we first compared the mean body size of the population before (i.e., all individuals) and after (i.e., just survivors) the selection event took place, using a t-test (Sokal & Rohlf 1995). We then estimated the strength of selection using the univariate variance-standardized selection gradient (βσ). This selection gradient is the most commonly reported and is obtained by regressing relative fitness (ωrelative = ωindividualaverage) onto the trait of interest (i.e., body size) in units of the population's standard deviation (Yσ = Yindividual/σ). The variance-standardized selection gradient describes the expected change in fitness after one standard deviation increase in the population's mean.

In addition to the univariate variance-standardized selection gradient (βσ), we also present the univariate unstandardized (β) and mean-standardized (βμ) selection gradients, as a convenience for those more comfortable with those formats. The unstandardized selection gradient represents the change in fitness expected after a one-unit increase in the population's mean. And the mean-standardized selection gradient represents the change in relative fitness expected after a 100% increase in the population's mean (Hereford et al. 2004). All three selection gradients can be used as an estimate of the strength of selection and only differ in terms of the measurement scale. The two standardized selection gradients (βσ and βμ) can be obtained by dividing the unstandardized selection gradient by either the population's standard deviation (βσ = β/σ) or its mean (βμ = β/μ).

We used survivorship as our estimate of fitness in this study. Each individual collected from the population was assigned either a 1 (i.e., alive) or a 0 (i.e., dead). Relative fitness was then estimated by dividing each of these scores by the population's mean fitness (ωaverage = 0.427). Because this measure was dichotomous, we used a logistic regression to test for the significance of our selection gradients (Janzen & Stern 1998). We visualized the selection gradients non-parametrically using cubic splines and following Schluter (1988). Standard errors were estimated by bootstrapping the dataset 500 times.

In order to determine if the mortality we observed in our collection was the result of natural selection or simply an artifact of our collection procedures, we resampled the population two days after the initial collection (August 3, 1999). On this date, daily high temperatures were still above the average high for that time of year (~32°C), but within one standard deviation of the mean (Fig. 1). No mortality was observed during the resampling and a total of 42 individuals were captured on this date (no nymphs were observed).

These crickets were brought back to the laboratory and their femurs measured in the same manner as for our original collection. Our hypothesis was that if the changes in body size that we observed during the initial collection resulted from our collection protocols, then those changes would not be reflected in the second sample. To test this hypothesis, we compared the mean femur length of the crickets collected on August 3rd to the means of the initial collection, both before (i.e., all individuals) and after (i.e., only survivors) the mortality event, using t-tests.

All statistical analyses were performed using SAS version 9.1 (SAS 2002) with the exception of the nonparametric regression which was performed using the program glmsWin 1.0 (Schluter & Nychka 2000).


We found that extremely high, but naturally occurring, temperatures selected for increased body size in A. socius. The mean femur length of our study population increased from 6.69 ± 0.05 mm (± s) before selection to 6.82 ± 0.07 mm after selection had occurred (Fig. 2A). This change represented a 1.9% increase in the population's mean phenotype, but was not significant at the α = 0.05 level using a one-tailed t-test (t82=1.13, P = 0.13). However, femur length was positively and significantly related to survivorship based on our logistic regression (X2 = 6.18; P = 0.01). The selection gradients (± s) were estimated as: βσ = 0.281 ± 0.112; β = 0.591 ± 0.237; and βμ = 3.95 ± 1.58. The relationship between femur length and survivorship was roughly linear (Fig. 2B). By back-calculating our selection gradients, we found that for every 1 mm increase in femur length, a cricket's probability of survival increased by 37%.

Fig 2.

(A) The number of individuals at each femur length (mm) collected before selection (white bars) and after selection occurred (black bars). (B) The nonlinear selection gradient (± s) estimated by the cubic splines. Fitness is on a relative scale (i.e., ωmean = 1) and femur length is measured in terms of standard deviations. The raw data are also presented (closed circles).


The selection pressure we observed did not appear to be caused by our collection protocols. When we resampled the population two days after the initial collection (i.e., August 3rd), no mortality was observed and the average femur length of the population was 6.95 ± 0.08 mm (± s). This mean was 3.9% larger and significantly different (t143= 2.85, P = 0.0052) from the initial sample before the selection event; and 1.9% greater, but not significantly different (t82=1.13, P = 0.2610) after the selection had occurred (Fig. 3). In fact, because the mean femur length of the August 3rd collection was slightly higher (though not significantly different) from the survivors of our August 1st collection, our protocols may have led to an under-estimation of the strength of selection. As such, our selection gradients should be considered conservative.

Fig 3.

The mean ± s for femur length (mm) of crickets collected on August 1st before selection (open circle) and after selection (closed circle), as well as those collected on August 3rd (gray circles). Letters indicant significant differences at the α = 0.05 level.



The magnitude of selection gradients that we observed in this study (Fig. 2) suggests that the strength of temperature-mediated selection in this system is quite strong. The variance-standardized selection gradient (βσ) was 0.281 ± 0.112 (s), which is 87% greater than the median strength of selection on body size reported in previous studies (body size median βσ = 0.15, Kingsolver & Pfennig 2007). The strength of selection we observed was also 27% greater than the median values reported across all traits (all traits median βσ = 0.22, Kingsolver & Pfennig 2007).

One factor that may have influenced our estimates of the selection gradient was the fact that we used a univariate measure of fitness as opposed to a multivariate technique (Blows 2007). If the extreme temperatures observed in this study also acted on traits that were genetically correlated with body size, then our estimates would include both direct and indirect selection pressures (Lande & Arnold 1983). In their review, Kingsolver et al. (2001) found a strong correspondence in the variance-standardized selection gradients between univariate and multivariate techniques. This trend suggests that the univariate methodology that we employed did not bias our estimate of βσ, although we could not test this hypothesis directly because only femur length was measured in this study.

In addition to providing information about the general strength of natural selection, selection gradients help us predict the evolutionary response of a population in the next generation. The change in a population's mean phenotype (in standard deviation units) that is expected in the next generation (Rσ) is equal to the product of the variance-standardized selection gradient and the trait's narrow-sense heritability (Rσ = βσh2). Narrow-sense heritability represents the proportion of the total phenotypic variation that can be attributed to the additive effect of genes (Falconer & Mackay 1996). Based on our observed selection gradients (Table 1) and a previously published estimate of the heritability for femur length in A. socius (h2 = 0.45 ± 0.04 s, Mousseau & Roff 1989), we predict that femur length will increase by 0.13 standard deviations in the next generation — approximately 0.06 mm or 0.8%. If these selection pressures were applied consistently and the heritability remained stable (Roff & Mousseau 1999), then the mean femur length of the population would increase by 50% in 339 generations.

Our data suggest that increases in body size brought about by exposure to extremely high temperatures can affect the body size of A. socius in the next generation. This observation has an important implication for broad-scale patterns of geographic variation in body size among ectotherms. Although endotherms (Ashton et al. 2000, Zink & Remsen 1986) and many ectotherms (Ashton & Feldman 2003, Litzgus et al. 2004) tend to exhibit larger body sizes at higher latitudes (a pattern known as Bergmann's rule), other ectotherms (Ashton & Feldman 2003, Huey et al. 2000) including many orthopterans (Masaki 1967, Masaki 1978, Mousseau 1997) tend to be larger at lower latitudes (i.e., they follow the converse of Bergmann's rule).

Converse Bergmann's size clines are thought to be caused primarily by the time constraints that season length places on an organism's growth (Roff 1980). Given similar growth rates, larger body sizes require longer periods of time to produce than smaller body sizes. Because season lengths tend to be longer at lower latitudes, organisms that follow the converse of Bergmann's rule are thought to have more time for growth and therefore are able to reach larger sizes.

While this explanation is generally accepted (Blanckenhorn & Demont 2004, Chown & Klok 2003, Mousseau 1997, Stillwell et al. 2007), our data suggest that periodically occurring extreme heating events may also contribute to the converse of Bergmann's rule. The high temperatures that we observed in 1999 were atypical for the area. In Newberry County, daily high temperatures of 40.5°C or greater have only occurred twelve times between 1948 and 2001 (including twice in 1999) or on average of once every 4.5 years (Williams et al. 2006). If the rate at which these high temperatures are encountered increases with decreasing latitudes (as we would expect) then selection for larger body sizes should be stronger at low, relative to high, latitudes. Our results suggest that selection pressure associated with periodically occurring extreme heating events may contribute to the converse of Bergmann's rule. Furthermore, if these extreme heating events occur with greater frequency as a result of projected global warming (IPCC 2001), we anticipate that the number of ectothermic species that follow the converse of Bergmann's rule will increase over time.

Finally, although we have assumed that the mortality we observed was caused by desiccation, it may have been due to heat shock (Lindquist 1986). In order to distinguish between these two possibilities, populations would have to be reared in the laboratory and subjected to similarly high temperatures (~40.5°C) under both moist and dry conditions. If the mortality observed in this study was due to desiccation, then we would expect lower mortality rates in the moist treatment.


The authors thank K. Fedorka and an anonymous reviewer for their insightful comments. Funding was provided by the National Science Foundation (NSF0234510) and the Department of Energy (DEFG0207ER64455).


  1. A. Addo-Bediako, S. L. Chown, and K. J. Gaston . 2001. Revisiting water loss in insects: a large scale view. Journal of Insect Physiology 47:1377–1388. Google Scholar

  2. K. G. Ashton and C. R. Feldman . 2003. Bergmann's rule in nonavian reptiles: turtles follow it, lizards and snakes reverse it. Evolution 57:1151–1163. Google Scholar

  3. K. G. Ashton, M. C. Tracy, and A. de Queiroz . 2000. Is Bergmann's rule valid for mammals. American Naturalist 156:390–415. Google Scholar

  4. D. Atkinson and R. M. Sibly . 1997. Why are organisms usually bigger in colder environments? Making sense of a life history puzzle. Trends in Ecology & Evolution 12:235–239. Google Scholar

  5. W. U. Blanckenhorn and M. Demont . 2004. Bergmann and converse Bergmann latitudinal clines in arthropods: two ends of a continuum. Integrative and Comparative Biology 44:413–424. Google Scholar

  6. M. W. Blows 2007. A tale of two matrices: multivariate approaches in evolutionary biology. Journal of Evolutionary Biology 20:1–8. Google Scholar

  7. J. R. Bodie and R. D. Semlitsch . 2000. Size-specific mortality and natural selection in freshwater turtles. Copeia 732–739. Google Scholar

  8. J. T. Bonner 2006. Why Size Matters. Princeton University Press. Princeton, NJ. Google Scholar

  9. C. R. Brown and M. B. Brown . 1999. Natural selection on tail and bill morphology in barn swallows Hirundo rustica during severe weather. Ibis 141:652–659. Google Scholar

  10. J. H. Brown, J. F. Gillooly, A. P. Allen, V. M. Savage, and G. B. West . 2004. Toward a metabolic theory of ecology. Ecology 85:1771–1789. Google Scholar

  11. S. L. Chown 1993. Desiccation resistance in six sub-Antarctic weevils (Coleoptera, Curculionidae) — humidity as an abiotic factor influencing assemblage structure. Functional Ecology 7:318–325. Google Scholar

  12. S. L. Chown and K. J. Gaston . 1999. Exploring links between physiology and ecology at macro-scales: the role of respiratory metabolism in insects. Biological Reviews 74:87–120. Google Scholar

  13. S. L. Chown and C. J. Klok . 2003. Altitudinal body size clines: latitudinal effects associated with changing seasonality. Ecography 26:445–455. Google Scholar

  14. S. L. Chown, P. Pistorius, and C. H. Scholtz . 1998. Morphological correlates of flightlessness in southern African Scarabaeinae (Coleoptera: Scarabaeidae): testing a condition of the water-conservation hypothesis. Canadian Journal of Zoology 76:1123–1133. Google Scholar

  15. J. A. Endler 1986. Natural Selection in the Wild. Princeton University Press. Princeton. Google Scholar

  16. D. S. Falconer and T. F. C. Mackay . 1996. Introduction to Quantitative Genetics. Prentice Hall. Essex. Google Scholar

  17. K. M. Fedorka, W. E. Winterhalter, and T. A. Mousseau . 2007. The evolutionary genetics of sexual size dimorphism in the cricket Allonemobius socius. Heredity 99:218–223. Google Scholar

  18. P. R. Grant and B. R. Grant . 1995. Predicting microevolutionary responses to directional selection on heritable variation. Evolution 49:241–251. Google Scholar

  19. J. Hereford, T. F. Hansen, and D. Houle . 2004. Comparing strengths of directional selection: how strong is strong. Evolution 58:2133–2143. Google Scholar

  20. H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S. N. Vignieri, A. Hoang, C. E. Hill, P. Beerli, and J. G. Kingsolver . 2001. Strength and tempo of directional selection in the wild. Proceedings National Academy of Sciences 98:9157–9160. Google Scholar

  21. R. B. Huey, G. W. Gilchrist, M. L. Carlson, D. Berrigan, and L. Serra . 2000. Rapid evolution of a geographic cline in size in an introduced fly. Science 287:308–309. Google Scholar

  22. IPCC 2001. Climate Change 2001: Synthesis Report. (Intergovernmental Panel on Climate Change). ><. Google Scholar

  23. D. J. Irschick, J. J. Meyers, J. F. Husak, and J. F. Le Galliard . 2008. How does selection operate on whole-organism functional performance capacities? A review and synthesis. Evolutionary Ecology Research 10:177–196. Google Scholar

  24. F. J. Janzen and H. S. Stern . 1998. Logistic regression for empirical studies of multivariate selection. Evolution 52:1564–1571. Google Scholar

  25. J. G. Kingsolver, H. E. Hoekstra, J. M. Hoekstra, D. Berrigan, S. N. Vignieri, C. E. Hill, A. Hoang, P. Gibert, and P. Beerli . 2001. The strength of phenotypic selection in natural populations. American Naturalist 157:245–261. Google Scholar

  26. R. Lande and S. J. Arnold . 1983. The measurement of selection on correlated characters. Evolution 37:1210–1226. Google Scholar

  27. R. Levins 1969. Thermal acclimation and heat resistance in Drosophila species. American Naturalist 103:483–499. Google Scholar

  28. S. Lindquist 1986. The heat-shock response. Annual Review of Biochemistry 55:1151–1191. Google Scholar

  29. J. D. Litzgus, S. E. DuRant, and T. A. Mousseau . 2004. Clinal variation in body and cell size in a widely distributed vertebrate ectotherm. Oecologia 140:551–558. Google Scholar

  30. J. L. Marshall 2004. The Allonemobius-Wolbachia host-endosymbiont system: evidence for rapid speciation and against reproductive isolation driven by cytoplasmic incompatibility. Evolution 58:2409–2425. Google Scholar

  31. S. Masaki 1967. Geographic variation and climatic adaptation in a field cricket (Orthoptera - Gryllidae). Evolution 21:725–741. Google Scholar

  32. S. Masaki 1978. Climatic adaptation and species status in a lawn ground cricket. 2. Body size. Oecologia 35:343–356. Google Scholar

  33. T. A. Mousseau 1997. Ectotherms follow the converse to Bergmann's rule. Evolution 51:630–632. Google Scholar

  34. T. A. Mousseau and D. A. Roff . 1987. Natural selection and the heritability of fitness components. Heredity 59:181–197. Google Scholar

  35. T. A. Mousseau and D. A. Roff . 1989. Adaptation to seasonality in a cricket — patterns of phenotypic and genotypic variation in body size and diapause expression along a cline in season length. Evolution 43:1483–1496. Google Scholar

  36. D. Roff 1980. Optimizing development time in a seasonal environment — the ups and downs of clinal variation. Oecologia 45:202–208. Google Scholar

  37. D. Roff 2001. Life History Evolution. Sinauer Associates. Sunderland, MA. Google Scholar

  38. D. A. Roff and T. A. Mousseau . 1987. Quantitative genetics and fitness — lessons from Drosophila. Heredity 58:103–118. Google Scholar

  39. D. A. Roff and T. A. Mousseau . 1999. Does natural selection alter genetic architecture? An evaluation of quantitative genetic variation among populations of Allonemobius socius and A. fasciatus. Journal of Evolutionary Biology 12:361–369. Google Scholar

  40. F. J. Rohlf 2006. Digitize landmarks and outlines from image files, scanner, or video: tpsDIG1. ><. Google Scholar

  41. SAS 2002. SAS 9.1. SAS Institute Inc. Cary, NC. Google Scholar

  42. D. Schluter 1988. Estimating the form of natural selection on a quantitative trait. Evolution 42:849–861. Google Scholar

  43. D. Schluter and D. Nychka . 2000. Program for estimating fitness functions using the cubic spline: glmsWin 1.0. ><. Google Scholar

  44. T. W. Schoener and D. H. Janzen . 1968. Notes on environmental determinants of tropical versus temperate insect size patterns. American Naturalist 102:207–224. Google Scholar

  45. R. M. Sibly and J. H. Brown . 2007. Effects of body size and lifestyle on evolution of mammal life histories. Proceedings National Academy of Sciences 104:17707–17712. Google Scholar

  46. R. R. Sokal and F. J. Rohlf . 1995. Biometry. W.H. Freeman and Co. New York. Google Scholar

  47. R. C. Stillwell, G. E. Morse, and C. W. Fox . 2007. Geographic variation in body size and sexual size dimorphism of a seed-feeding beetle. American Naturalist 170:358–369. Google Scholar

  48. W. A. van Voorhies 1996. Bergmann size clines: a simple explanation for their occurrence in ectotherms. Evolution 50:1259–1264. Google Scholar

  49. C. N. Williams Jr., M. J. Menne, and D. R. Easterling . 2006. United States Historical Climatology Network Daily Temperature, Precipitation, and Snow Data. ><. Google Scholar

  50. W. E. Winterhalter and T. A. Mousseau . 2007. Patterns of phenotypic and genetic variation for the plasticity of diapause incidence. Evolution 61:1520–1531. Google Scholar

  51. R. M. Zink and J. V. Remsen . 1986. Evolutionary processes and patterns of geographic variation in birds. pp 1–69. In R. F. Johnston , editor. (Ed.). Current Ornithology. Plenum Press. New York. Google Scholar

  52. M. Zuk 1988. Parasite load, body size, and age of wild-caught male field crickets (Orthoptera, Gryllidae) — effects on sexual selection. Evolution 42:969–976. Google Scholar

Wade E. Winterhalter and Timothy A. Mousseau "The strength of temperature-mediated selection on body size in a wild insect population," Journal of Orthoptera Research 17(2), 347-351, (1 December 2008).
Accepted: 1 June 2008; Published: 1 December 2008

Back to Top