The reaction norm of size and age at maturity under multiple predator risk


Correspondence author. E-mail:


1. Two major theories underpin our understanding of how predation risk shapes life history. The first is centred around predator induced changes in activity that subsequently reduce food intake and thus growth. The second is centred around size selective, predator induced changes in development.

2. Here, we challenge these theories using experiments and probabilistic models of maturation reaction norms to investigate predator induced life history in the water flea Daphnia pulex facing two different predators.

3. We combine this reaction norm investigation with an assessment of growth rate, development rate, moult number and moult duration to uncover the mechanisms controlling predator induced life history plasticity when D. pulex face either large or small size selective predators.

4. The probabilistic reaction norms reveal predator specific norms of reaction in size and age along a food gradient. Fish cues reduce age and size, with a bias in age, and do so by reducing moult number and duration. Midge cues increase age and size, with a bias in size, and do so by fine scale modulation of early growth rates.

5. These data contribute towards developing a unified view of how predation risk from multiple predators shapes life history evolution.


Prey species respond to predation risk in a multivariate manner. Traits that change in response to predation are behavioural such as habitat shifts, reductions in activity and altered feeding rates; life historical, such as achieving maturity at different times and or sizes; and morphological, where defensive structures, colours or shapes are employed at key stages in development (Tollrian & Harvell 1999; Touchon & Warkentin 2008). Current research increasingly respects the fact that prey actually combine a variety of these traits (Spitze 1992; Boersma, Spaak & De Meester 1998; Pigliucci 2003; DeWitt & Scheiner 2004; Mikolajewski & Johansson 2004; Relyea 2004).

Theory on how life history traits, such as size and age at first reproduction, respond to predation risk is well developed and predicts a wide range of outcomes often dependent on the relationship between mortality and growth rate at various ages/stages (Stearns and Koelle 1986, Abrams and Rowe 1996). Thus, the response to predation risk can be classified in part by the life cycles of the organisms under study (Stearns & Koella 1986). Paralleling the variety of predictions, there is a large difference, and rather little overlap, between research on how species with complex life cycles such as tadpoles, dragonflies and damselflies respond to risk vs. the response of simple life cycle species, such as fish, snails and daphnids.

For species with complex life cycles, where reproduction often takes place in a different environment than juvenile growth, the adaptive response to juvenile predation risk is most often characterized by a behavioural reduction in activity, leading to lower energy intake but reduced predation risk (Skelly 1992; Benard 2004; Noonburg & Nisbet 2005). This reduced activity is hypothesized to reduce food intake and cause a reduction in growth rate and development rate. The result is later age at maturity and smaller size at maturity – what Stearns identifies as the most common reaction norm (Stearns 1989). Such changes in size and age driven by reduced activity and feeding, are traditionally characterized as costs (Skelly 1992; Ball & Baker 1995, 1996) and this underpins the classic growth-survival trade-off (Werner & Anholt 1993). Essentially, reproductive output declines with decreasing size (cost) but survival increases with less activity (benefit).

However, in many non-metamorphic species, such as daphnids and fish, size and age responses are construed as adaptive per se (e.g. Tollrian 1995b) rather than being linked to activity or habitat use. Here, prey responses are influenced directly by chemical cues from predators acting independently on prey development rate and on growth rate – something we term decoupling (Beckerman, Wieski & Baird 2007). Under such decoupling, changes in size and age increase fitness by generating a life history/size refuge from predation.

The changes in the life history of such non-metamorphic taxa often take on predator specific endpoints. Large predators prefer large prey and select for small and early prey maturity. Small predators prefer and are limited to small prey and select for later and larger prey maturity (Gadgil & Bossert 1970; Taylor & Gabriel 1992, 1993; Riessen 1999; Gosline & Rodd 2008). Not surprisingly, these life history endpoints are also induced in the prey by chemical cues from the predators (Tollrian & Harvell 1999).

Here we investigated the mechanisms underpinning the life history response of the water flea Daphnia pulex facing predation risk from two common predators, the large size selective three-spined stickleback (Gasteroterus aculeatus) and the small size selective midge larvae Chaoborus flavicans. Our goal was to characterize mechanisms used by D. pulex to drive the predator specific patterns in size and age at maturity. Specifically, we examined reaction norms of age and size (Stearns & Koella 1986; Heino, Dieckmann & Godø 2002), the relationship between development rate and growth rate (Ball & Baker 1996) and changes in moult number and duration.

We used a new method to explore a reaction norm of size and age at maturity (Heino, Dieckmann & Godø 2002) and we combine this with an investigation of the relationship between development rate and growth rate, suggested by Ball & Baker (1995, 1996); see also Beckerman, Wieski & Baird 2007) to test the hypotheses that (a) fish cues generate small-early responses by altering age but not size, via moult number and or duration; and (b) that midge cues generate a late and large life history by changing juvenile growth rate per se at critical, risk sensitive stages in development (Beckerman, Wieski & Baird 2007).

An important aspect of our research is the use of these methods under a two-predator one prey system. To our knowledge, only Stibor & Lüning (1994) have investigated how cues from different predators translated into a prey’s life history responses, their focus being on testing theory about differential allocation to growth and reproduction under predation risk (Taylor & Gabriel 1992). Instead, most research has focused on documenting single predator specific responses, the ensuing trade-offs among predators and the related concepts of trait integration, co-specialization and compensation (Peckarsky & McIntosh 1998; Sih, Englund & Wooster 1998; DeWitt, Robinson & Wilson 2000; Decaestecker, De Meester & Ebert 2002; Langerhans & DeWitt 2002; DeWitt & Langerhans 2003; Relyea 2003; Vonesh & Osenberg 2003; Abjornsson, Hansson & Brönmark 2004; Mikolajewski & Rolff 2004; Van de Meutter, Stoks & DeMeester 2005; Boeing, Ramcharan & Riessen 2006; Mikolajewski et al. 2006).

Our approach augments this previous research by examining a bi-variate reaction norm of size and age under different predation regimes. We identify the mechanism by which risk generates predator specific reaction norms of size and age in D. pulex by combining probabilistic reaction norms with an analysis of the relationship between development and growth rate under predation risk and of the moulting process. Our analyses reveal predator specificity both in the shape of the reaction norms and the processes generating them.

Materials and methods

Study System

We examined the life history of D. pulex facing C. flavicans (phantom midge) or G. aculeatus (stickleback), common predators of daphnids in S. Yorkshire, UK (Hammill 2008). D. pulex are small (adults c. 2–2·5 mm) aquatic crustaceans, found in ponds and lakes, where both small and large daphnids face size selective predation. The empirical pattern of life history response to both fish and midge predation cues are well-established (Tollrian & Harvell 1999). Young of the year fish represent an early season predation pressure on daphnids, preying heavily on large juveniles and adults. As the fish grow and their preference shifts to larger prey, predation pressure shifts to the Chaoborus midge, which prey heavily on second – third instar juveniles. Thus, D. pulex can face predation risk from two size selective predators and each predator is size selective on a different stage of development. We used clone Cyril from Crabtree pond in Sheffield, UK. Cyril life history and morphology respond to midge and fish cues and comes from a pond that contains both midge and fish predators. More detail on Cyril is available in Hammill, Rogers & Beckerman (2008).

Reaction Norms and Development Rate vs. Growth Rate

Probabilistic reaction norms

To characterize the size-age reaction norm, we used the method of probabilistic reaction norms (PRN’s –Heino, Dieckmann & Godø 2002). PRNs are logistic regression models of maturation as a function of size and age. PRN’s generated by experimentally varying food along a defined gradient to manipulate growth rates are particularly useful at removing systematic bias that can occur when simply measuring size and age at maturity, and can reveal an accurate representation of the reaction norm (Heino, Dieckmann & Godø 2002). This systematic food variation (see below for experimental details) is replicated within the three predation risk scenarios to develop a comprehensive description of the probabilistic process giving rise to changes in the life history of D. pulex facing predation risk.

Development – Growth

Beckerman, Wieski & Baird (2007) extended the theory of Ball & Baker (1996) to describe multiple predator responses in life history. Under predator free conditions across a range of food levels, growth rate and development rate are predicted to positively co-vary. Along this line of co-variation, at high food levels, high growth is matched by a high development rate and reduced time to maturity – i.e. larger and early maturity. When predation acts to reduce activity, reduced growth rates and lower development rates arise from reduced food intake. This leads to increased age and decrease size at maturity.

Alternately, predation risk could decouple development rate and growth rate: for a given growth rate, development rate might increase or decrease. This independent response of development time and growth rate to cues allows a variety of different life history endpoints, such as early-small or late-large, which are associated with classic size selective predation. Thus, challenging this theory can help illuminate predator specific biases in age or size changes that generate life history end-points (Ball & Baker 1996; Beckerman, Wieski & Baird 2007).


To evaluate the life history response to fish and Chaoborus, we carried out a replicated life table experiment along an environmental gradient of eight food levels, crossed with three predator treatments. Each replicate (five/treatment) housed a single daphnid in 160 ml jar containing at minimum 150 ml hard artificial pond water (Hard ASTM – see Beckerman, Wieski & Baird 2007; Hammill, Rogers & Beckerman 2008) and Chlorella vulgaris algae as food (Beckerman, Wieski & Baird 2007; Hammill, Rogers & Beckerman 2008). Daphnids were photographed (see below) and transferred daily to fresh experimental conditions.

The food variation generates standard variation in growth rates among, but independent of, any changes in growth rate that predation risk might cause. Development rate was estimated as 1/Age at Maturity. Growth rates were estimated as log ratio of size differences between birth and maturity divided by Age at Maturity. The number of moults to maturity and the mean inter-moult duration was also calculated for each food × treatment replicate.

Predation and food manipulations

The daphnids were exposed to either control, fish conditioned or Chaoborus conditioned water to generate the three predator treatments. Control conditions refer to daphnids kept in artificial pond water. Fish cue conditions were created by enriching the water via housing three-spined sticklebacks at a density of 0·5 fish l−1 for at least 24 hours at 15 °C in the artificial pond water. This water was filtered through a 20 μm filter (Whatman) and raised to 20 °C prior to use (Beckerman, Wieski & Baird 2007). Chaoborus cue conditions were created by adding concentrated purified Chaoborus extract (sensu Tollrian 1995a,b; Hammill, Rogers & Beckerman 2008) at a concentration of 1 μl ml−1 (c. 50 Chaoborus per litre).

The three predator treatments were crossed with a replicated (= 5) experimental gradient of eight C. vulgaris food treatments descending serially by 50% at each step from the highest concentration of 1·5 × 105 cells per ml. Size and age at maturity along this gradient of food, within each predator treatment defines the predator specific reaction norm. We use logistic regression (see below) to infer the probabilistic reaction norm from predation risk generated variability in growth and development rate under a common gradient of food.

Trials began with third generation, first instar offspring, which were measured and placed into individual jars. Each day the experimental conditions were refreshed and each animal was photographed (Leica MZ-9, Leica Microsystems GmbH, Wetzler, Germany; Nikon 4500, Nikon, Surrey, UK) and measured using the image analysis software NIH Image J (Rasband 1997–2009). Age and size at maturity were recorded as the day on which eggs first appeared in the brood pouch.


We analysed the reaction norm for size and age at maturity using probabilistic reaction norms (e.g. Heino, Dieckmann & Godø 2002; Morita & Fukuwaka 2006). Following Heino, Dieckmann & Godø (2002), we fitted logistic regression models to a binary response variable indicating maturation at size x and age t. The model took the form:


This model allows substantial flexibility in the shape of the reaction norm across all three treatments (Size × Age expands to Size + Age + Size:Age). We fitted this model using the quasi-binomial family to accommodate minor over-dispersion, using F-tests rather than χ2 tests to evaluate significance. The reaction norm is defined by the 50th percentile from the logistic regression for each predator treatment (Heino, Dieckmann & Godø 2002). For heuristic purposes, we also under-laid on each reaction norm linear approximations of growth associated with the eight common food treatments within each predation treatment. This provides a picture of the ‘envelope’ of growth trajectories driving the reaction norm (see fig. 3–5, Heino, Dieckmann & Godø 2002). We also analysed this model as a repeated measures model accounting for repeated measurements of size at age (before and after maturation). We found variation among individuals to be near zero, providing no difference in interpretation. For ease of visualization/prediction, we refer to the non-repeated measures model.

To assess the Ball & Baker (1995, 1996) model, we analysed the treatment specific relationship between development rate and growth rate using ancova. We specified the model testing for the effect of an interaction between predator treatment and growth rate on development rate.

To test the hypothesis that moult number and or duration were targeted by fish cues, but not by midge cues, we further analysed moult number and moult duration using ancova, with food density (log transformed to accommodate non-linearities) hypothesized to interact with predation treatment. To test the hypothesis that growth rate per se was transformed by midge cues but not fish, we again used ancova to investigate the effect of food level and predation treatment on growth rate. We measured early growth rate as the log ratio of size differences in the first four instars divided by time to instar four, with initial body size as a covariate.


The reaction norm (Fig. 1) for this D. pulex clone is a complex response in size and age to the different predator treatments (Fpredation = 37·4, d.f. = 2, < 0·01; Fage×size = 13·8, d.f. = 1, < 0·01). The concave curve in (Fig. 1) (all panels) characterizes maturation across the food gradient (envelope of growth; high food generates qualitatively earlier maturation and larger size). Fish chemical cues reduced both size and age at maturity, but generated a much larger decrease in age (Fig. 1a vs. b) – note compression of reaction norm towards early age). Midge chemical cues induced increases in size and age at maturity, but generated a much larger increase in size (Fig. 1c vs. b) – note upward bending and movement of reaction norm). (Fig. 1d) provides a 3 × 3 cross-section of the reaction norms. The probability of maturation is presented as a function of three sizes (panels: 1·5, 1·6 and 1·7 mm) and three ages (7, 10 and 12 days) for each predator treatment (50th percentile ± 95%CI), showing how fish cues (dashed line) consistently generate a higher probability of maturation at small sizes and ages in contrast to the midge effect of reduced probability of maturation at any age until large size.

Figure 1.

 Probabilistic size and age reaction norm under fish predator (a), control (b) and midge predator (c) conditions. A probabilistic reaction norm is defined as the 50th percentile from a logistic regression predicting the probability of maturation as a function of age and size. We plot the 10th, 25th, 50th, 75th and 90th percentiles (black to white shading). Underlaid on these data are linear approximations (dashed grey lines) to the growth trajectories generated by the food treatments to create variation in size and age within each predation treatment. Note that these do not figure explicitly in the analysis. The dashed lines move from low (light) to high (dark) food and represent the average trajectory for the same food treatments in each predator treatment with a solid line representing average growth trajectory in each predator treatment. The range of growth rates is the ‘envelope’ in which maturation occurs. (d) A decomposition of the reaction norms. The probability of maturation is presented as a function of three sizes (panels: 1·5, 1·6 and 1·7 mm) and three ages (7, 10 and 12 days), showing how fish cues (dashed line) consistently generate a higher probability of maturation. Data in (d) are fitted 50th percentiles ± 95% CI.

We rejected the hypothesis that the effect of growth rate on development rate depends on the predator treatment (Fig. 2; Fpredation×growth rate = 2·16, d.f. = 2, > 0·05). Having rejected that hypothesis, we re-fit the model without the interaction, evaluating the simpler hypothesis that there are differences in the intercepts for each treatment, as per the original theory (Ball & Baker 1996; Beckerman, Wieski & Baird 2007). As expected, under control, predator free conditions, development rate increased linearly with growth rate (tcontrol.slope = 36·393, < 0·01). Any variation in food moves daphnids along this line resulting in the expected early and large size under high food and late and small size under low food.

Figure 2.

 Data and fitted regression lines from the ancova testing whether the effect of growth rate on development rate depends on predator treatment (e.g. growth rate × predator treatment interaction). An additive effect of growth rate and predator treatment would indicate some ‘decoupling’ of growth and development because development rate would change as a function of predator regime, independent of growth rate. Our data confirm this effect for fish (elevated development rate), but indicate no change for midge. See (Table 1) for Tukey contrasts on intercepts.

Table 1 presents Tukey contrasts showing that fish predator cues increased development rate independent of growth, but that the effect of chaoborus cues were not distinguished from control conditions.

Table 1.   Tukey independent contrasts from ancova of development rate vs. growth rate and predator treatments. The analysis indicated additive effects of predator environment and growth rate on development rate (ancova, see text). The analysis shows significant increases in development time for fish exposure relative to control, but not for midge exposure
ContrastEstimateSEt valuePr(>|t|)
Fish – control0·00650·00193·490·0021
Midge – control−0·00280·002−1·410·34
Midge – fish−0·00940·0018−5·06<1e-04

We found evidence for an additive effect of food and predation risk treatment on moult number (Fig. 3a, Table 2) and moult duration (Fig. 3b, Table 2). Both number and duration decreased with increasing food (Table 2). There was no difference between moult number or duration under midge vs. control conditions (Table 2). There was a significant reduction in moult number and duration under fish predation risk (Table 2).

Figure 3.

 Moult number (a) and moult duration (b) both respond additively to the effects of increasing food and predation risk cues from midge and fish. Fish cues reduce both number and moult duration, while the effect of midge cue on moult number and duration cannot be distinguished from control conditions. Data are plotted with fitted lines from an ancova.

Table 2. anova and coefficients tables for analysis of instar number and duration. (A) Instar number and duration are altered by predation treatment and food abundance in an additive manner (all interactions > 0·1). (B) Contrasts showing significant decrease in instar number and duration with food and a significant decrease in instar number and duration with fish cues but not midge cues
 d.f.Sum of sqRSSAkaike Information CriterionF valuePr(F)
Instar number
Predation treatment221·1766·47−28·5721·03<0·001
Log(algae abundance)198·75144·0546·12196·17<0·001
Instar duration
Predation treatment20·572·42−340·1613·92<0·001
Log(algae abundance)12·714·55−278·60132·05<0·001
 EstimateSEt valuePr(>|t|)
  1. F-statistic: 49·25 on 3 and 90 d.f., P-value < 2·2e-16.

Instar number
Fish vs. control−1·060·17−6·11<0·001
Midge vs. control−0·240·19−1·220·23
Log(algae abundance)−0·680·05−14·01<0·001
F-statistic: 72·78 on 3 and 90 d.f., P-value < 2·2e-16
Instar duration
Fish vs. control−0·180·04−5·13<0·001
Midge vs. control−0·060·04−1·530·13
Log(algae abundance)−0·110·01−11·49<0·001

Controlling for the effects of initial size (F = 11·26, d.f. = 1, = 0·001), the effects of food on early growth rate (instar 1–4) depended on predation treatment (Fig. 4; Falgae×predation risk = 3·32, d.f. = 2, = 0·04). There was no evidence that fish cues altered the relationship between growth rate and food (ΔIntercept = 0·042, t = 1·29, = 0·2; ΔSlope = −0·0009, t = 0·29, = 0·77). However midge cue generated a significantly steeper relationship between food and growth rate than control (ΔSlope = 0·007, t = 1·98, = 0·05) but no change in the intercept (ΔIntercept = −0·07, t = −1·86, = 0·06).

Figure 4.

 Growth rates early in life (instar 1–4) increase with increasing food (ancova with initial body size as an additional covariate). Midge cues generate a steeper gradient between growth rate and food than control and fish. Fish cues do not alter significantly the relationship between growth and food.


Our goal was to characterize mechanisms used by D. pulex to drive the predator specific patterns in size and age at maturity. Specifically, we examined reaction norms of age and size at maturity, development rate relative to growth rate and changes in moult number and duration.

Our reaction norm analysis indicates that fish generate a sharp bias towards earlier maturity and a small reduction is size at maturity (Fig. 1a vs. b). The ancova analysis of growth and development (Fig. 2) clearly indicates that large size selective predators such as fish can decouple development and growth (see also Beckerman, Wieski & Baird 2007). We also found evidence that fish cues generate a decrease in the number and duration of moults (Fig. 3), but had little effect on the relationship between growth and food (Fig. 4).

In contrast, the reaction norm data show that midge cues generate a marked increase in size at maturity and a lesser increase in age at maturity (Fig. 1c vs. b). The analyses of growth and development rate (Fig. 2) provided no evidence for a decoupling of development and growth. Furthermore, we found little evidence that midge cues change in moult number or duration (Fig. 3). Finally, however, we found that early midge predation cues elevated growth rates, compared with control or fish cues, as food became available (Fig. 4).

Thus, we see predator specificity in the component of life history (size vs. age), predator specificity in the capacity to de-couple growth and development via changes in moult number and duration, and predator specificity in the effects of predation risk on juvenile growth (see Stibor & Lüning 1994). We have shown that chemical cues from large size selective, vertebrate predators appear to generate a small – early life history by decoupling development and growth rate, underpinned by a reduction in moult number and duration. In the same prey species, chemical cues from small size selective, invertebrate predators, suggest fine control of growth rate per se at early stages, where risk is highest.

All of our analyses (see also Beckerman, Wieski & Baird 2007) confirm current theory and empirical evidence that daphnids likely obtain a refuge from vertebrate predation risk generated by early maturation (Tollrian & Dodson 1999). Furthermore, our data support the hypothesis that there is a premium on increasing growth rate to decrease predation risk (Spitze, Burnson & Lynch 1991; Stibor 1992; Tollrian 1993; Stibor & Lüning 1994), minimizing time in the most vulnerable developmental stages (Tollrian 1993, 1995a, b). As midge larvae select small developmental stages as prey items, there is likely no pressure to reduce growth rate at later developmental stages. Given that predation risk is centred on juveniles, we actually expect little pressure on development rate because only small individuals are at risk.

Most importantly, our analyses disentangle growth and maturation (probabilistic reaction norms), visualizes the role of development rate and growth rate under predation risk (Ball & Baker 1995, 1996), and isolates plasticity in moulting to highlight the mechanisms by which D. pulex prey handle predation risk from two different predators using similar ‘tools’. Our data emphasize that large size selection represents maturation or post-maturation pressure, leading to selection for rapid development. In contrast, small size selection acts on juveniles, promoting growth, but not necessarily reflecting any maturation rate pressure (but see Spitze 1991). Thus, under fish predation, growth rates are increased as a result of increased development rate, while under midge predation pressure, somatic growth rates per se are increased to reach a size refuge with limited pressure on the age at which to mature.

Our data reflect some predictions from classical theory (Stearns & Koella 1986; Stearns 1989). Midge predation generates a strong relationship between mortality and juvenile growth rate and thus between mortality and age. Both of these biological details result in downward sloping sigmoid curves as suggested by Stearns & Koella (1986). Our data suggest that a strong link between daphnid growth rate and juvenile mortality (e.g. Fig. 4.) may drive the overall patterns (Fig. 1a–c). We find no evidence that the potentially high extrinsic adult mortality associated with fish predation moves the reaction norm to the right. Instead, it is the relative contrast of fish to midge that describes the rightward moving reaction norm. These patterns suggest that the mortality – growth – age relationships associated with chaoborus may be more significant than the mortality – age relationships associated with fish.

More generally, our data complement more recent research (Stibor 1992; Stibor & Lüning 1994) and recent modelling efforts where natural history and physiology are specified to define clear constraints and generate predictions about expected response to predation risk (e.g. predator gape limitation can elevate prey growth rates Noonburg & Nisbet 2005; Urban 2007). We now have a more comprehensive understanding of the proximate control of life history under predation risk in D. pulex facing multiple predators, within the context of theory about allocation to growth and development (Stibor 1992; Stibor & Lüning 1994).

Size and Age Reaction Norms and the Evolution of Plasticity

Reaction norms describe the phenotype of a given genotype in different environments. Quantifying reaction norms for co-varying traits continues to be a challenge for evolutionary biologists (Stearns & Koella 1986; DeWitt, Sih & Hucko 1999; Mikolajewski & Johansson 2004). Probabilistic reactions norms for size and age have been used throughout fisheries research to examine and propose hypotheses about evolutionary changes size and age resulting from commercial fishing (Heino, Dieckmann & Godø 2002; Gårdmark, Dieckmann & Lundberg 2003; Grift et al. 2003; Barot et al. 2005; Gårdmark & Dieckmann 2006; de Roos, Boukal & Persson 2006; Morita, Tsuboi & Nagasawa 2009; Sharpe & Hendry 2009). They are useful because the method helps limit the confounding effects of changes in growth and survival on maturation by modeling maturation probability conditional on individuals having reached a particular age and size (Olsen et al. 2004).

Here we present a reaction norm of size and age across a gradient of food, but in three different experimentally controlled environments. Each panel in (Fig. 1a–c) can be considered a reaction norm itself, along the underlying food gradient. But our data also show predator specificity in the expression of this reaction norm. Thus, the change in the maturation reaction norm across the three environments (Fig. 1a–c) is a reaction norm as well. While our data are for a single genotype, if this variation in size and age across predator treatments is common, it could represent a mechanism for the maintenance of genetic variability in plasticity.

Our data suggest how plasticity in size and age, and in growth, might support a frequency dependent expression of maturation responses in multiple predator environments. Kingsolver et al. (2007) point out that ‘theoretical models predict that selection on reaction norms should depend on the relative frequency of environmental states experienced by a population.’ Daphnids experience seasonally changing environments, often including transitions between multiple predators or other forms of stress. The population from which this clone was extracted experiences fish predation early in the season and midge predation later in the season (Hammill, Rogers & Beckerman 2008). We can thus hypothesize that the plasticity shown in (Fig. 1) may facilitate environment specific frequency dependent expression of adaptive life history under multiple predator environments. And the reaction norms representing this frequency dependence are likely to have evolved in response to varying, unpredictable predator environments – the same selective pressures that contribute to the maintenance of variation in maturation reaction norms (Tollrian & Harvell 1999; Kingsolver et al. 2007).


We thank Stewart Plaistow, Wai Meng Au-Yeong, Dylan Childs, Mark Rees and Matt Robinson for in-valuable discussion and insightful comments. Alison Blake maintained the clones and laboratory. APB and SRD were funded by NERC (NE/D012244/1).