Mark Hassall, Centre for Ecology, Evolution and Conservation, School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, UK. Tel.: +44-16-03592541; fax: +44-16-03591327; e-mail: firstname.lastname@example.org
Analysis of size of offspring reared through three laboratory generations from populations of the field grasshopper Chorthippus brunneus from 27 sites around the British Isles showed that offspring were larger towards the cooler-wetter conditions in the western and northern limits of the range. This variation had a significant genetic component. There was a trade-off between clutch size and offspring size between and within populations. Under favourable thermal and feeding conditions maternal fitness was optimal when individuals produced the largest clutches of the smallest eggs, but under poor conditions maternal fitness was optimal when individuals produced small clutches of very large offspring. Calculation of geometric mean fitness over time indicated that having larger offspring near to the edge of the range could be advantageous as a conservative risk-spreading strategy. As well as geographic variation in egg size, significant environment–genotype interactions in egg size in relation to temperature were observed.
Identifying and explaining patterns of variability in offspring size is a central challenge in evolutionary ecology (Carrière & Roff, 1995). Differences in size of offspring may arise as phenotypic responses to differences in the environment (Fischer et al., 2003) but may also reflect adaptive genotypic responses due to selection in different environments acting either directly on offspring size (Messina & Fox, 2001; Roff, 2002) or indirectly on other life history traits that are correlated with offspring size.
One source of variability in offspring size is the differences that occur in different parts of the geographic range of a species. A consistent pattern across a number of taxa is that offspring tend to be larger at higher latitudes (Azevedo et al., 1996). Many aspects of the environment can change with latitude, including season length, predation pressure and often most conspicuously, temperature (Messina & Fox, 2001). Similar differences in offspring size have been observed with respect to altitude (Fox & Czesak, 2000) and also in populations that have evolved over several years at different temperatures in the laboratory (Azevedo et al., 1996). While these patterns are well established, what is less clear is why larger offspring are produced at lower temperatures.
In some cases responses to temperature appear to be phenotypic, as in some Coleoptera where different components of oogenesis have different sensitivities to temperature (Ernsting & Isaaks, 1997). In other cases, larger eggs at lower temperatures may confer a fitness advantage to the offspring by partly offsetting extended larval or nymphal development times caused by the lower temperatures (Dingle & Mousseau, 1994). In others, larger offspring may be correlated with larger adults (Atkinson & Begon, 1987), which are more fecund. Larger offspring may also have a survivorship advantage, either directly as a result of possessing a greater energy reserve, or indirectly due to having a potentially shorter nymphal development time (Azevedo et al., 1996).
To understand why larger offspring have evolved at the expense of a smaller clutch size in a particular environment, the fitness advantage to the offspring must be evaluated in respect to the fecundity cost of a smaller brood size to the maternal parent. Balancing selection between increased fitness of larger offspring and increased maternal fitness for larger broods of smaller offspring has been modelled for many systems (Smith & Fretwell, 1974) based on the trade-off between offspring size and clutch size observed in many semelparous organisms (Roff, 2002). The optimal size of offspring predicted by this model may vary with environmental conditions such as crowding (Begon & Parker, 1986), ration level (Roff, 2002) and potentially also temperature.
Such predictions are usually made in relation to stable environmental conditions, which may not pertain in the field, especially near the edge of the range of a species where yearly variation in meteorological conditions can have a very strong influence on the dynamics of populations. Under fluctuating environmental conditions selection may favour a bet-hedging strategy involving either conservative or diversifying risk-spreading tactics (Philippi & Seger, 1989; Hopper, 1999). Where levels of environmental variation are predictable, selection may also lead to adaptive phenotypic plasticity in life history traits (Nylin & Gotthard, 1998) to enable organisms to respond appropriately to temporal heterogeneity in their environments.
In Orthoptera, variation in egg size has been related to survivorship of emerging vermiform larvae in several species of crickets (Carrière et al., 1997); to moisture content of the soil or feeding conditions for the nymphs (Monk, 1985; Atkinson & Begon, 1987) and to the time within the season when the eggs were laid (Cherrill & Begon, 1989), which is closely correlated to their hatching date the next year (Cherrill, 2002). In the field grasshopper, Chorthippus brunneus (Thunberg), several life history traits including the pattern and rate of development, growth rate and adult size were found to vary across its geographical range in Europe (Telfer & Hassall, 1999; Walters, 2003), including hatchling size, smaller hatchings being found in areas with higher annual maximum temperatures (Telfer & Hassall, 1999).
Optimality models can provide a valuable tool with which to test adaptational hypotheses, however, such models need to be parameterized with empirical data if they are to yield any real insight into biological systems (Messina & Fox, 2001). A key trait associated with offspring size is survivorship, but size-dependent mortality is notoriously difficult to quantify and is often evident only under poor feeding conditions (Stearns, 1992). Few studies have demonstrated how progeny size alters fitness in different environments (Fox & Czesak, 2000), in particular why a larger egg size tends to be favoured at lower temperatures (Azevedo et al., 1996).
In this paper, we demonstrate a trade-off between clutch size and offspring size by comparing geographically distinct ecotypes and show that there are genetic effects on hatchling mass. We then test three hypotheses: (1) optimizing selection results in different optimal offspring sizes under different mean levels of environmental condition; (2) under different levels of variation in environmental conditions selection has favoured a bet-hedging strategy in which a large hatchling is selected as a conservative risk-spreading tactic; (3) that there is more phenotypic plasticity in offspring size nearer to the edge of the range. These three hypotheses are not mutually exclusive.
The field grasshopper C. brunneus (Orthoptera: Acrididae) is a medium-sized temperate species that has a wide geographical distribution covering much of Europe and temperate Asia, reaching its northern range limits in Scotland. It is typically found in open, dry and sunny habitats (Ragge, 1965). It has a univoltine life cycle, with an obligate diapause during the egg over-wintering stage. Eggs hatch from late May and the nymphs take 4–8 weeks to pass through four or five instars (Hassall & Grayson, 1987), reaching adult eclosion from late June to early August. It is iteroparous with clutches of up to 14 eggs laid in pods in the soil every few days. Adults typically start to die in late August, though some may survive as late as November.
Geographic variation survey and genetic effects
Three laboratory generations were cultured from 27 populations of C. brunneus from sites spread widely around Great Britain and Ireland as described by Telfer & Hassall (1999). Offspring data from that study are re-analysed here.
Clutch size-offspring size trade-offs
During the summer of 1992, 12 female and 8 male adults were collected from 14 of the original 27 sites over a wide geographical range. Egg pods from the 14 populations were transferred from cold storage to an incubator in Spring 1993 and reared through the first laboratory generation in aluminium cages at densities of 50 individuals per cage by supplementing losses from cages of spare animals.
Second filial generation (F2) cultures were established in 1994 for four cultures from Weeting, Littlehampton, Hayle and Dunbar. Dunbar (Scotland) and Littlehampton (south coast of England) populations had the lowest and highest mean adult female eclosion weights in the F1 generation, respectively, and represent geographical extremes of the collection area. The Hayle population from Cornwall is from the extreme oceanic southwest of England, and had the slowest development time in the F1 generation. Weeting in the Breckland district of East Anglia has a semi-continental climate contrasting to the other three sites.
The F2 generation was reared from egg pods laid between the 9th and 19th August 1993 and moved from cold storage to hatching pots in February 1994. Hatchlings were removed daily by gentle aspiration, sexed (Richards & Waloff, 1954; Hassall & Grayson, 1987) and individually weighed to ±0.01 mg before the nymphs had access to food.
The F3 generation individuals were reared from three populations chosen to represent a longitudinal gradient in climate continentality of founding sites, varying from relatively cool, wet and cloudy summers in southwest Ireland through semi-continental conditions in Breckland, East Anglia to relatively warm, dry and sunny summers in central Germany (Fig. 1).
The F2 generation egg stocks were kept at 4 °C in sealed pots of moist sterilized sand for at least 8 weeks to break diapause (Kelly-Stebbings & Hewitt, 1972) before transferring to 30/20 °C on a 14/10 h day/night cycle to induce hatching. Eggs hatched after 10–14 days. Random male–female pairs of weighed hatchlings were transferred to individual cylindrical containers consisting of nylon ‘bolting’ mesh cylinders (90 mm diameter, 300 mm high, 1 mm mesh) enclosed at the top with a tightly fitting Petri dish lid and fitted around pots (88 mm diameter × 88 m deep) of sown grass. After adult eclosion, a small Perspex pot of sterilized sand was placed within the cylindrical container for oviposition and egg pods removed every 2 days.
Grasshoppers were reared in growth cabinets at 25, 30, 35 and 40 °C, daytime temperatures on a 14/10 h day/night light cycle with 15 °C night-time temperatures for all treatments. The positions of cylindrical containers were rotated to reduce the effect of small temperature deviations within the cabinet.
High-quality food treatments comprised a three species grass-mix of Festuca ovina agg., Agrostis capillaris L. and Poa pratensis L., sown 6 weeks prior to use in a heated greenhouse in Fison's ‘Levington’ multipurpose compost and fed ad libitum. Low-quality food treatments comprised the same three species grown in soil from cores taken from Lakenheath in Breckland (Hassall & Dangerfield, 1989,1997) and heated to 55 °C to extract soil animals (Hassall et al. 1988). The soil on this site is a calcareous brown earth of the Methwold Series (Corbett, 1973).
Analysis of genetic effects
Variation in life histories between generations was investigated for all 27 cultures using linear regression. Population means of females in the F2 generation were regressed against the population means in their parental generation. Similarly, F3 data were regressed on F2 data, and on the grandparental (F1) data. Significant regressions between successive generations indicate genetic effects for the traits, or significant maternal effects.
Individuals on the high-quality food diet were ranked in order of hatchling mass within each temperature treatment and then divided into five size classes containing 11–13 individuals. On the low food quality treatment, ranked individuals were divided into seven size classes containing 11–13 individuals. On the low-quality food, males were also ranked and divided into three size classes containing 14–16 individuals. Survivorship of individuals within a given size class were regressed against the mean hatchling mass of the size class.
Fitness was calculated using a model of grasshopper life histories (Grant et al., 1993) with parameter values for over-winter egg mortality (M), adult mortality rate (μ2) and season length (S) kept constant as set by Grant et al. (1993). Time from hatching to first ovipositioning (t1) was determined empirically in each of the temperature and food quality treatments. Survivorship from hatching to adult eclosion for a given hatchling size was incorporated into the model (Table 1) by deriving juvenile mortality rate (μ1) from −ln(1−survivorship)/time to adult eclosion, and rearranging eqn 1.
Table 1. Parameters used to calculate fitness using the Grant et al. (1993) model of grasshopper life history.
High food quality
Low food quality
t1, time from hatching to first oviposition (days); μ1, juvenile mortality rate (determined from survivorship data to adult eclosion as a function of hatchling size); t2, interclutch interval (days); μ2, adult mortality rate; n, eggs per clutch; S, season length (days); M, over-winter egg mortality. Parameters μ1 and n are calculated as a function of hatchling size (see methods).
μ1 = −ln[f(x)]/development time
n = clutch mass/hatchling mass
To calculate offspring fitness for a given offspring size, the number of eggs per clutch (n) was kept constant within a treatment and determined empirically as 3.32 and 7.34 for 25 and 35 °C, respectively, for low-quality food and 4.69, 6.52, 8.85 and 10.24 for 25, 30, 35 and 40 °C, respectively, for high-quality food. To calculate maternal fitness, clutch size (n) was adjusted for changes in offspring size by dividing an empirically derived constant clutch egg mass within a given treatment by a hatchling mass. Since only dry clutch egg mass was measured experimentally, total live hatchling mass was estimated from dry clutch egg mass divided by 0.3, as dry egg mass was approximately 30% of live hatchling mass (unpublished data).
The effect of varying performance among generations under different environmental conditions was assessed by calculating geometric mean fitness for two different hatchling sizes: 4.34 and 6.00 mg, corresponding to the masses of the continental and oceanic ecotypes, respectively (Walters, 2003). The effects of experiencing a harsh survival year (equivalent to 25 °C on low food quality) on resultant geometric mean fitness was assessed for two alternative ecotype strategies of ‘small’ vs. ‘large’ offspring size. The experiment was run for two alternative, more favourable, background environments; one near to ‘optimal conditions’ for fitness (equivalent to 40 °C on high food quality) and one sub-optimal but still more favourable than the poor years (equivalent to 30 °C on high food quality). The conditions of the model do not reflect actual variability in growth conditions that occur in the field; rather they provide a hypothetical framework for assessing the relative selective pressures on hatchling size in an environment that varies year to year.
The numerical experiment provides a method for assessing the relative influence of strong size-dependent mortality, which may occur infrequently under particularly poor growth conditions, on populations that typically experience more favourable conditions. The comparison of two hypothetical populations, one that typically experiences ‘sub-optimal’ growth conditions vs. another that typically experiences ‘very favourable’ growth conditions provides a way of evaluating the relative influence of poor growth conditions on those populations that are near to the edge of the range vs. those that are close to the centre of the range, respectively.
Geographic variability in and genetic effects on offspring size
Offspring size increased with higher annual rainfall (r =0.566, n = 26, P < 0.001) but decreased with climate continentality (Fig. 2a), summarized by July minus January mean temperatures, an index of climatic conditions (Lamb, 1972) from the relatively cool and wet summers of the northwest Scotland and Ireland to the warm and dry summers of the southeast of England, offspring size decreased with mean July temperatures (Fig. 2b) and July maximum temperatures (r = −0.670, n = 26, P < 0.001), but increased with July daily temperature range (r = 0.529, n = 26, P < 0.001) and NW–SE range (r = 0.557, n = 26, P < 0.001). Clutch size (F2 generation) increased with climate continentality (r = 0.435, n = 26, P < 0.05) and decreased with increased annual rainfall (r = 0.447, n = 26, P < 0.05).
The hatchling mass of successive generations of grasshoppers was significantly positively related to the hatchling mass of the parental and grandparental generations (Table 2). Viability of eggs under favourable laboratory conditions varied across the geographic range of parent populations, increasing with maximum July temperatures (r = 0.595, n = 27, P < 0.001) July daily temperature range (r = 0.620, n = 27, P < 0.001) and decreasing with increased annual rainfall (r = −0.530, n = 27, P < 0.001). Thus egg survivorship in the laboratory was greatest for the populations in the southeast, which had the largest clutches of the smallest eggs.
Table 2. Statistics for hatchling mass and mature mass regressed between generations.
F2 on F1
F3 on F2
F3 on F1
F2 on F1
Trade-off between offspring size and clutch size
There was a negative correlation between clutch size and offspring size both between populations from the 27 sites (r = −0.4055, n = 27, P < 0.05) and for the five cultures from the north, south, east and west edges of the sampling area (r = 0.4331, n = 131, P < 0.001).
Effects of egg size on juvenile performance
Hatchling size was negatively correlated with the time from hatching to second moult (i.e. duration of instars I and II) (r = −0.4420, n = 24, P < 0.05), and positively correlated with mass after second ecdysis for females (r = 0.4965, n = 21, P < 0.05). The proportion of females inserting the IIa instar was negatively correlated with hatchling size (r = −0.5154, n = 27, P < 0.01). Thus larger hatchlings develop faster through the first two instars, retain their larger size until at least the third nymphal stage and tend to insert the IIa instar less frequently.
For female grasshoppers, hatchling weight was not significantly correlated with relative growth rate to adult eclosion, nymphal development period, adult eclosion mass or any measure of mature size. However, for males, larger hatchlings did have a larger mass at adult eclosion (ln-transformed to reduce skewness) (r = 0.4405, n =27, P < 0.05).
Larger offspring survived better from hatching to adult eclosion on both the high- and low-quality foods (Fig. 3a, b, Table 3). To clarify the relationship between survivorship and hatchling mass on the high-quality foods, the effect of temperature was removed using a general linear model with survivorship for a given hatchling mass as a function of rearing temperature (F20 = 6.40, P < 0.01) and plotting unstandardized residuals against hatchling mass (y = 0.0763x − 0.3939: r = 0.656, n = 11, P < 0.01).
Table 3. Survivorship ‘s’ from hatching to adult eclosion as a function of hatchling mass ‘x’ (mg).
s = f(x)
The equation coefficient for survivorship on a high-quality food diet was derived by regressing the unstandardized residuals of survivorship, after the effect of temperature has been removed, against hatchling mass.
Low food quality
0.1831x − 0.3711
0.1217x − 0.2902
High food quality
0.076x + 0.3825
0.076x + 0.2646
0.076x + 0.1821
0.076x + 0.0509
Offspring and maternal fitness
Offspring fitness isoclines for each temperature/food quality treatment (Fig. 4a, c), suggest that a larger individual offspring will always have greater fitness irrespective of conditions, although as the relationship is hyperbolic the potential benefits of an increase in hatchling mass decrease as fitness approaches an asymptote. In many organisms, the cost to the maternal parent of producing larger offspring is a reduction in offspring number, resulting in maternal fitness being greatest at an intermediate optimal offspring size. The results for C. brunneus (Fig. 4b, d) indicate that maternal fitness is optimized when their offspring size is either very small under favourable thermal and food conditions or very large, under poor thermal and food quality conditions. From these results it can be predicted that there will be a significant selective advantage in having large numbers of small offspring when both the thermal and food quality conditions are favourable, and a significant selective advantage in having small numbers of large offspring when thermal and food quality conditions are at their harshest.
Under intermediate conditions there will be relatively little selective pressure on offspring size so it can be predicted that there will be a higher level of variation in offspring size in the centre of the geographical range. Analysis of coefficients of variation in mean offspring size for the five populations with the smallest offspring (CV = 9.14 ± 0.94), the five with the largest offspring (CV = 9.45 ± 0.8) and the 17 populations with intermediate offspring size (CV = 11.97 ± 0.47) support this prediction by showing that variation was significantly (Fe,26 = 3.965, P < 0.05) higher for these populations from the middle of the range.
Large offspring could, however, also be selected as a bet-hedging strategy in environments where harsh conditions occur only irregularly. This was assessed in a numerical experiment by calculating geometric mean fitness over time in respect to an increase in the frequency of experiencing just ‘harsh’ years (25 °C, low food quality). Two alternative offspring sizes were modelled, a large 6.00 mg vs. a small 4.34 mg hatchling mass. These hatchling masses are equivalent to the mean hatchling mass measured for two ecotypes (Table 4) that have been selected in a relatively ‘harsh’ (cool wet oceanic) and ‘favourable’ (warm dry continental) environment, respectively. The relative fitness advantages of the two offspring sizes were assessed in two background environments: (1) very favourable where survivorship is equivalent to the 40 °C high food quality treatment and (2) less favourable but still much better than the harsh environment, equivalent to the 30 °C high food quality treatment. In contrast to when conditions were constant when small offspring size was optimal (Fig. 4c), predictions of geometric mean fitness when some years were harsh (Fig. 5) suggest that large offspring size would be selected for if the frequency of experiencing a poor year was greater than 1 in 29 for the very favourable background conditions and 1 in 100 years for the less favourable background conditions, the fitness advantage becoming substantial when the frequency of experiencing such a poor year increases to more than 1 in 10 years.
Table 4. Geographic variation in hatchling mass in relation to climate continentality along a longitudinal cline.
The results of this numerical experiment thus suggest that large offspring size is more likely to be selected for close to the northern and western range limits where thermal conditions are more frequently unfavourable.
Thermal reaction norms for offspring and clutch size for three ecotypes selected under different climates are given in Fig. 6. The oceanic ecotype laid larger eggs than either the semi-continental or continental ecotypes across the whole temperature range (Tukey post-hoc test: P < 0.001). There were corresponding differences in clutch size; the oceanic ecotype producing smaller clutches than either the semi-continental (Tukey post-hoc test: P < 0.001) or continental ecotypes (Tukey post-hoc test: P = 0.015). Differences in the gradients of reaction norms for offspring size across a range of temperatures suggests that there is also geographical variation in egg size plasticity in response to temperature (Fig. 6a). There was a genotype-environment interaction in egg size between temperature and ecotypes (Fig. 6a), most significantly for oceanic and semi-continental ecotypes between 35 and 40 °C (F1,38 = 6.142, P < 0.05).
The evolution of phenotypic plasticity in life histories is considered one of the most exciting topics in life history biology (Gotthard & Nylin, 1995); in particular more effort is required to understand the evolution of reaction norms for progeny size (Fox & Czesak, 2000). The thermal reaction norms for all three populations from the longitudinal cline in continentality in this study showed a steep increase in size of offspring between 25 and 30 °C, but this may simply reflect physiological constraints in provisioning eggs at the lower temperature which is close to the minimum at which this species can successfully complete its life cycle (Willott & Hassall, 1998). Between 30 and 40 °C, the continental and semi-continental populations did not show any significant response in offspring size in relation to temperature but the size of offspring of the oceanic ecotype did decrease, resulting in a significant genotype–environment interaction.
Having larger offspring in colder years could be adaptive for the oceanic ecotype because it is subject to consistently harsher thermal conditions than either of the other two ecotypes (Fig. 1). For larger offspring to be adaptive, adverse temperatures in the year during oogenesis should correlate with adverse conditions during other stages of the life cycle the following year. Cold temperatures in the year that the eggs are laid could result in the eggs being laid relatively late in the season, due to the imposed delay on time to first oviposition (Walters, 2003). Since the hatching date of the eggs in the following year is related to the date that the eggs were oviposited in the current year (Cherrill, 2002), nymphs hatching late would be likely to experience a relatively short and limiting season length. The steeper reaction norm over the 30–40 °C range for the western population may then be an adaptive response to a reduction in season length the following season. These thermal reaction norms thus provide support for the phenotypic plasticity hypothesis in that it could be of adaptive value for an ecotype that has evolved under less predictable conditions near the limit of the range, to have more plasticity in offspring size in relation to temperature.
For many arthropods there are intraspecific latitudinal clines from small eggs produced at lower latitudes to larger eggs at higher latitudes (reviewed by Fox & Czesak, 2000), but it is not normally clear how the relationship between progeny size and fitness varies between different local environments (Messina & Fox, 2001). Despite adult female C. brunneus being smaller nearer to the cooler cloudier northern and western limits of its range (Telfer & Hassall, 1999), they also lay significantly smaller clutches of larger eggs nearer to these range limits. This consistent pattern was not entirely phenotypic as it persisted through three laboratory generations reared under laboratory conditions when a significant proportion of the variability in offspring size was found to be caused by genetic effects. Why then have females of C. brunneus near to the western and northern edges of their range been selected to reduce their fecundity in order to produce larger offspring?
Four possible fitness benefits that could result from a larger egg size in C. brunneus are: (1) increased over-wintering survivorship of the eggs, (2) a shorter nymphal development period, (3) the production of larger and more fecund adults and (4) a higher nymphal survivorship. The relationship between offspring size and egg viability in 10% moist sand was not significant but when comparing populations of both C. brunnneus and C. paralleflus, Monk (1985) concluded that larger offspring survived better in wetter soils.
For the four populations from different parts of the range there was some evidence of larger offspring having shorter development times as larger hatchlings reached their second ecdysis more quickly than small ones, but this could have been influenced by the higher frequencies of insertion of the IIa instar in the south. There was not a direct relationship between hatchling mass and development period but there was a significant negative correlation between development period and northing in the main geographical study (Telfer & Hassall, 1999). Comparing the three populations from both the longitudinal cline and the full 27 site survey there was no significant relationship between hatchling size and adult female mass at eclosion but such a relationship may have been obscured because adult mass is greater when an additional IIa instar is inserted, as happens more often for the smaller hatchlings from the south east of the range (Telfer & Hassall, 1999).
Large eggs can lead to higher nymphal survival. Larger newly hatched larval crickets have a clear advantage when crawling from the oviposition site to the surface of the soil (Carrière et al., 1997). Similarly a larger initial hatchling size of C. brunneus increases nymphal survivorship to adult eclosion, even under the most favourable thermal conditions with an excess of high-quality food (Fig. 3). Survivorship was highest at 40 °C, close to the optimal internal body temperature for this species, and decreased significantly with decreasing temperature down to 25 °C. The influence of hatchling size on survivorship was much stronger when the food, consisting of the same three species of grass, was grown on soil from one of the field sites resulting in a nitrogen content which is much closer to that found in the field (Grayson, 1984). On this lower quality food, the slopes of the relationships between survivorship and offspring size were substantially steeper than on the higher quality food grown on nutrient rich potting compost, suggesting that under these poor conditions a larger offspring would have a greater fitness advantage.
An analysis of maternal fitness suggests that when C. brunneus is reared on high-quality food under favourable thermal conditions (>30 °C) there is always a selective advantage to maximizing offspring number over offspring size. A minimum viable egg size was not evident based upon these analyses of intrinsic nymphal survivorship, as even the smallest hatchlings (<3 mg) survived to adult eclosion, mated successfully and went on to produce clutches of eggs. However, in the field a minimum viable offspring size may for example be determined by survivorship of the initial vermiform larvae, as is the case for some crickets (Carrière et al., 1997). In contrast, when thermal conditions are ‘poor’, particularly when feeding conditions (nitrogen content) are also poor there is a selective advantage to maximizing offspring size over offspring number, reinforcing the suggestion that costs of survivorship may only be evident under poor food conditions (Stearns, 1992). These results thus do support the first, optimizing selection, hypothesis in that the optimal size of offspring varies under different stable environmental conditions, with smaller offspring selected under more favourable conditions such as occur in the south and east of the range and larger offspring under the less favourable conditions near the north and west limits of the range.
In the field, thermal conditions may not be as stable as assumed in the above optimality analysis but will vary considerably from year to year. Thus optimal clutch size may be influenced by different selection pressures for small or large egg sizes depending on whether the year was relatively ‘good’ or ‘poor’ for survival. To account for the effect of variation in thermal conditions geometric mean fitness can be calculated over time (Philippi & Seger, 1989) as is illustrated in the numerical experiment (Fig. 5). The results suggest that when ‘poor’ years are periodically experienced in an otherwise favourable environment, which would, if stable, select for small offspring, a large offspring size may be selected for instead. The results show that this effect is more pronounced in cooler environments, such as near to range limits, where fitness is typically lower and that the larger sized offspring typical of the oceanic population would have a substantial fitness advantage if unfavourable conditions occurred more than 1 year in 10. Furthermore a population with smaller offspring, of the size of the continental ecotype, would have a fitness value of less than one at higher frequencies of experiencing such a poor year. These analyses therefore support the bet-hedging hypothesis as when favourability of environmental conditions varies temporally a large offspring could represent an adaptive conservative risk-spreading tactic that increases geometric mean fitness (Hopper, 1999).
If a large offspring size has been selected as a bet-hedging strategy then it can be predicted that the level of variation in offspring size would be the lowest in those populations that experience the greatest directional selection pressure on offspring size under (a) the most unfavourable conditions, such as at the edge of range or (b) under the most favourable conditions. Under these extremes, there is likely to be a significant fitness advantage in having either a large or a small offspring size, respectively. Where the probability of experiencing a poor year is neither frequent nor rare, there is likely to be weaker directional selection pressure on offspring size. The coefficient of variation was lowest in those populations that laid the largest and smallest eggs and significantly higher for intermediate populations which provides additional empirical evidence from the field in support of the bet-hedging hypothesis.
To summarize, these results provide support for all three of the hypotheses proposed. The larger offspring of C. brunneus observed closer to the limits of its range could have resulted from optimising selection operating due to larger offspring being adaptive under the harsher nymphal survivorship conditions in the north and west of the range. Under more variable favourable environmental conditions at the edge of a range, having few but large offspring can also be adaptive as a conservative bet-hedging strategy (Haccou & Iwasa, 1995). Furthermore in more temporally heterogeneous environments, greater phenotypic plasticity in relation to changes in temperature can be adaptive. We conclude therefore that all three of these different selective pressures could potentially act simultaneously on offspring size near the limits of a range.
We thank Dr. D. Jackson and Mr. R. Bryant for assistance in the laboratory and M. Schädler for advice and assistance in locating a suitable continental field site. We are very grateful to Professor G. Hewitt for assistance in planning this project and Professor J. Palutikof for advice in obtaining meteorological data. This work was supported by Natural Environment Research Council studentships to R.W. and M.T. and NERC small grant (NER/BS/2002/00370). Meteorological data were provided by Met Éireann, the British Atmospheric Data Centre and UFZ – Environmental Research Leipzig-Halle, Department of Soil Sciences.