Patterns of maternal provisioning
Larger females had a larger total yolk volume output and used this larger amount to produce more eggs. This is not surprising since larger females, as a result of increased size, have an increased amount of energy available for reproduction (Castellano et al., 2004). Larger females, however, did not produce larger eggs. Egg size is independent of female size and is therefore likely to have evolved in response to selection imposed by the offspring environment. There was a trade-off in egg size and number. This means that, regardless of female size and total yolk output, females that produced larger eggs produced fewer eggs. Variability in egg size within clutches was not related to female size, total yolk output, egg number or mean egg size. This supports the idea that variation in egg size within clutches being independent of these other parameters is an adaptive strategy in response to the offspring environment (Crump, 1981; Kaplan & Cooper, 1984; Koops et al., 2003). The egg size-number trade-off also applies to females producing clutches with variable egg sizes. A female producing more larger-sized eggs within a clutch produces fewer eggs in total and a female producing more smaller-sized eggs within a clutch produces a higher total egg number. Egg size therefore is independent of maternal phenotype. A single offspring size has not evolved. If the maintenance of small clutches of large eggs, larger clutches of small eggs and clutches of variable sized eggs in the population is an evolved response then egg size needs to affect offspring fitness and thus these strategies must result in maternal fitness consequences.
Since there was no relationship between egg size and variation in egg size within clutches our results do not support the imperfect information hypothesis (see Koops et al., 2003 for review) in this species. This hypothesis predicts that females should produce variable clutches to offset wrong decisions when trying to produce eggs that are optimal for predicted future conditions. Under this scenario, unequal allocation within clutches decreases with increasing mean egg size. This relationship is not present in C. georgiana.
Fitness of offspring from small and large eggs from different clutches
Several studies examining the effects of egg size have found that effects on offspring size (Crump, 1984; Tejedo & Reques, 1992; Loman, 2002) and the length of the larval period (Crump, 1984; Tejedo & Reques, 1992; Laugen et al., 2002) are evident at hatching or early in the larval period but disappear towards metamorphosis. Lack of fitness consequences of offspring size also occurs sometimes in other taxa (e.g. fish: Sehgal & Toor, 1991; birds: Smith et al., 1995; mammals: Oksanen et al., 2001). In C. georgiana, egg size has an important influence on size at metamorphosis and thus offspring fitness (Altwegg & Reyer, 2003). Furthermore egg size also influences the length of the larval period, which again has important fitness consequences in the temporary freshwater environments that many amphibians use for reproduction (Wassersug, 1975; Alford, 1999). Berven & Chadra (1988) found that at lower densities, offspring from small eggs in higher food treatments metamorphosed at a larger size than offspring from large eggs. This is similar to our result where we detected a significant interaction of egg size with food level, indicating that offspring from small eggs in the fed treatment were growing at a proportionally faster rate and closing on the size of the offspring from large eggs. However in C. georgiana the offspring from small eggs never surpassed the size of offspring from large eggs in the same food level treatment. Parichy & Kaplan (1992) also found this relationship although it disappeared in higher quality (more food) treatments, unlike our results where the size difference persisted in our fed treatments. Our study of C. georgiana is unique in that strong influences of egg size on size at metamorphosis and the length of the larval period persist in both a high quality (fed) and low quality (unfed) larval environment.
In some studies, egg size did not affect larval survival (Crump, 1984; Brooks et al., 1991; Tejedo & Reques, 1992; Smith et al., 1995). Parichy & Kaplan (1992) found that the probability of surviving to metamorphosis was lower for offspring from small eggs across both high quality (high food) and low quality (lower food) treatments. In our study egg size had an important effect on survival to metamorphosis, where greater maternal investment per offspring is needed to ensure survival in a poor quality offspring environment when food resources are not available.
Egg size had a strong influence on offspring fitness, therefore coupled with the egg size-number trade-off, should have strong influences on maternal fitness. Offspring from different egg sizes performed differently under different conditions, therefore the strategies of producing clutches of fewer large eggs or clutches of more numerous smaller eggs should result in different maternal fitness consequences under different quality offspring environments.
In effect this experiment simulated two offspring size categories on the Smith & Fretwell (1974) model of offspring provisioning in two different environments (fitness curves) that offspring may experience: poor quality (unfed treatments) and good quality (fed treatments).
At the breeding site visited in this study, C. georgiana tadpoles are found in ponds ranging from only a few centimetres in diameter to up to 2 m across. Depths of these ponds also vary from 0.5 cm up to 15 cm (Seymour et al., 2000; Dziminski, unpublished data). Therefore there is spatial heterogeneity in pond size. Food resources are likely to vary between these ponds and can be very low in some (Doughty & Roberts, 2003), but even if food resources are equally distributed among ponds per unit of water or per area of pond substrate [since C. georgiana tadpoles are benthic (Main, 1957)], then density effects, if the female was to oviposit in one of the smaller ponds, would reduce the amount of food available per offspring in these crowded ponds. Therefore offspring may encounter high or low food conditions, and in the smallest ponds tadpoles may even experience a total lack of food. If a female produces small eggs and happens to oviposit in a large (food available) pond, this would result in a higher maternal fitness value. However, if she happened to oviposit these eggs in a small (low or no food available) pond, this would then result in a lower maternal fitness value (Fig. 3a). The reciprocal maternal fitness values would result if the female produced large eggs (Fig. 3a). It is unlikely that C. georgiana females choose ponds; rather, they choose males (Smith & Roberts, 2003), and in many cases females are intercepted by sneak or satellite males (Byrne, 2004; Byrne & Roberts, 2004) eliminating the opportunity of mating with a particular male and thus ovipositing in a particular pond. The opportunity for male choice of calling site based on pond quality (Howard, 1978) may exist but this interaction would be complex involving the male attracting a female that is carrying the right sized eggs for the pond he is calling from. Data on oxygen tension at oviposition sites suggest egg deposition is not driven by site quality (Byrne & Roberts, 1999; Seymour et al., 2000). More likely is the possibility that larvae may disperse into nearby ponds that are appropriate for their size. This is possible during periods of high rainfall when ponds become connected by a large amount of surface water (Dziminski, personal observation). This spatial heterogeneity in the quality of the offspring environment coupled with density effects and the opportunity for offspring to disperse into appropriate habitats creates the requirements for the selection of variable maternal investment between clutches (McGinley et al., 1987).
As well as being spatially variable in quality, the ponds C. georgiana use for reproduction can vary temporally in quality within the breeding season. These ponds can dry and fill several times within the breeding season and the length of time a pond contains water can vary unpredictably from persisting only a few days to persisting the entire breeding season (Seymour et al., 2000; Doughty & Roberts, 2003; Dziminski, unpublished data). To escape desiccation larvae must metamorphose. Production of large eggs results in the highest maternal fitness in the shorter duration ponds regardless of whether food is available (Fig. 3b). When food is available, production of large eggs results in the highest maternal fitness in shorter duration ponds (poor offspring environment) because offspring from small eggs cannot metamorphose in time and perish (Fig. 3b). In the longer duration ponds (good offspring environment) the production of small eggs results in the highest maternal fitness (Fig. 3b). The mixed strategy has intermediate maternal fitness to the small and large strategy, but has the benefit of ensuring some offspring surviving in the shorter duration ponds (Fig. 3b): a form of bet hedging (Crump, 1981; Philippi & Seger, 1989). Within-season temporal variation in the quality of the offspring environment is thought to lead to selection for variation in offspring size if parents can assess prevailing conditions and produce offspring of the appropriate size (McGinley et al., 1987). However, it is unlikely that C. georgiana females would assess the quality of the offspring environment taking into account factors noted above plus: (1) the high spatial variation in the quality of ponds in terms of food resources, (2) the temporal within-season variation in pond persistence and the unpredictability of this variation, for example the pond may persist, or dry early depending on the future weather conditions and (3) the interaction of (2) above and spatial variation in the temporal within-season variation of ponds. This complex set of interactions, that can be additive or cumulative, coupled with unpredictability may create the condition where selection favours a bet-hedging strategy such as producing mixed egg sizes within a clutch (Fig. 3b). This relationship though, would depend intrinsically on the proportion of energy invested into small and large offspring, the frequency of good offspring conditions, and the magnitude of the difference in the quality of the offspring environments, spatially and temporally (McGinley et al., 1987; Einum & Fleming, 2004). As the data are for eggs collected from individuals in the wild, variation within and among females may also be induced or partly induced by the environment (Reznick & Yang, 1993; Einum & Fleming, 2004) rather than present adaptive strategies. Long-term laboratory studies, although difficult with amphibians, would test this. Regardless of the actual trigger causing variation in egg size, the fact that there are such strong potential fitness consequences means that variation both within and between clutches would be selected for or against, depending upon the environment. In insects, selection on offspring by the environment has been shown to lead to differences in maternal provisioning strategies (Fox & Mousseau, 1996; Fox et al., 2001; Kudo, 2001).
Our example of the variable strategy involved using half the available energy for small eggs, half for large eggs. Although such extreme discrete bimodal distributions of egg sizes did not occur in natural clutches (Fig. 1a), rather more platykurtic distributions, our theoretical example is still a representative estimate of an alternate provisioning strategy. These more platykurtic allocation strategies would result in different maternal fitness values. These fitness values, however, would still lie in between the upper and lower bounds of producing just large or just small eggs (Fig. 3b). The resulting curves adjoining these strategies in the two different environments would vary in slope or position between the upper and lower bounds as a result of different values of maternal fitness. To measure the exact maternal fitness of these strategies would require measuring fitness of offspring from a range of egg sizes between these upper and lower bounds. Then, using this relationship, the calculation of the fitness of every offspring in the clutch and thus exact maternal fitness would be possible.
In this study we have shown that offspring provisioning is independent of female size or other clutch parameters and variation in offspring size within clutches is also independent. A trade-off occurred between offspring size and number. Maternal provisioning has strong influences on size at metamorphosis and the length of the larval period as well as survival to metamorphosis. This translates to strong influences on offspring fitness and thus maternal fitness. Under the spatially and temporally variable and unpredictable offspring environment that this species uses, the conditions that select for the evolution of variable maternal provisioning both between and within clutches can exist. Since there are positive consequences of alternative offspring provisioning strategies under combinations of these conditions, the high variation in maternal provisioning both within and between clutches in this species is very likely to be an evolved response to the offspring environment.