Compensatory growth following early nutritional stress in the Wolf Spider Pardosa prativaga


†Author to whom correspondence should be addressed. E-mail:


  • 1Spiders may often be subjected to nutritional stress during their development, either because of lack of prey (starvation) or because the prey available is nutritionally insufficient or chemically defended (i.e. deterrent or toxic). The ability of the Wolf Spider Pardosa prativaga (L. Koch) to catch up on growth and development after treatment with different types and intensities of nutritional stress in the initial period of independent life was tested.
  • 2The stress types (prey) were: (1) starvation; (2) toxicity (the collembolan Folsomia candida Willem); (3) deterrency (the aphid Rhopalosiphum padi L.); (4) nutritional insufficiency (the fruit fly Drosophila melanogaster Meigen). Nutrient-enriched D. melanogaster was used as control and was also the food given to the treated spiders after the stress period ended.
  • 3Though the spiderlings were set back considerably, especially by starvation and toxic and deterrent prey, those that survived achieved the same weight as the control spiders within 3–7 weeks, depending on treatment. The duration of instars after termination of the stress treatment was not affected. Thus, seemingly compensation was accomplished by a burst of supernormal growth initiated shortly after the stress was alleviated.
  • 4Growth compensation is considered a physiological mechanism that allows spiders an optimal seasonal timing of the life cycle even if their phenology is retarded in its early phases.


The life history of most animals is composed of three components: growth, maturation and reproduction (Vollrath 1987). Traditionally, life-history theory assumes that growth rates are maximized and vary only in response to the environment (Roff 1980; Gebhardt & Stearn 1988). However, repeated findings show that growth rates may vary even under similar conditions and that they are not always maximized but rather optimized (Gotthard 2000; Metcalfe & Monaghan 2001). The term growth compensation refers to an animal's capacity to grow faster than normal after a period of environmentally induced growth depression to recover an original body weight or growth trajectory (Jobling, Jørgensen & Sikavupio 1993; Broekhuizen et al. 1994; Nicieza & Metcalfe 1997; Metcalfe & Monaghan 2001). The very existence of compensatory growth responses demonstrates that under normal conditions individuals grow at a rate below that achievable (Broekhuizen et al. 1994; Morgan & Metcalfe 2001).

Growth compensation is seen across a wide range of taxa (Arendt 1997). It is normally achieved through ‘hyperphagia’ (increased feeding activity above the level of control animals) and is usually short-lived, although it can last for months (Jobling & Miglavs. 1993; Jobling & Baardvik 1994; Nicieza & Metcalfe 1997). Compensatory growth is mainly observed among animals that have suffered from deficiencies during early life stages, but it has also been observed among animals born late in the breeding season or animals that have experienced atypically low temperatures (Nicieza & Metcalfe 1997; Nylin & Gotthard 1998). During a stress period, affected animals can preserve irreplaceable tissue and thereby save key functions at the expense of other body functions that are less crucial to long-term fitness (Metcalfe & Monaghan 2001).

Organisms often seem to recover as a result of growth compensation, though the processes during the compensatory phase most likely carry a variety of costs and can have permanent effects on the adult individuals and even on their offspring (Metcalfe & Monaghan 2001). Costs can act at the physiological/cellular level, which are usually paid in the longer term, whereas ecological costs of compensatory responses tend to be more immediate (Gotthard, Nylin & Wiklund 1994; Abrams et al. 1996; Arendt 1997; Metcalfe & Monaghan 2001). The ecological costs may be higher susceptibility to predators and parasites because of the need for increased foraging activity (Abrams 1991; Werner & Anholt 1993; Gotthard et al. 1994; Abrams et al. 1996; Gotthard 2000). The physiological costs are often decreased resistance to starvation and other environmental stresses (Gotthard et al. 1994). Since high growth rates carry fitness costs (Higgins & Rankin 2001), the optimal solution to a setback could be to avoid these costs by delaying a life-history event rather than to pay the costs of keeping to the original time schedule (Abrams et al. 1996; Arendt 1997; Nylin & Gotthard 1998; Grand 1999; Gotthard 2000).

In temperate latitudes many animals do not have the option to set back development since the mating period is seasonal (Miyashita 1986; Vollrath 1987), unless they change from an annual to a biennial life cycle (Schaefer 1987). Gotthard et al. (1994) found that fast-growing individuals of the Speckled Wood Butterfly Pararge aegeria (Linnaeus) do not become heavier than slow-growing individuals but instead shorten their growth period. This indicates that there may be strong selection for being the ‘right’ size as well as having a short development time. Growth compensation is often incomplete, with affected animals never fully catching up in size, though sometimes animals (notably fish) may exhibit a higher net growth rate and in that way over-compensate to become even larger than control animals (Broekhuizen et al. 1994; Hayward, Noltie & Wang 1997).

We tested the ability of a Wolf Spider Pardosa prativaga (L. Koch) to compensate for early nutritional stress in an experiment with four types of stressful treatments. The stress factors used were: starvation; toxic prey, the collembolan Folsomia candida (Willem); deterrent prey, the aphid Rhopalosiphum padi (Linnaeus); and nutrient-deficient prey, using the fruit fly Drosophila melanogaster (Meigen).

The expectation was that starvation and toxic prey would be strong stress factors, deterrent prey intermediate, and the nutrient-deficient prey the least stressful treatment. The most stressful treatments were expected to cause the highest mortality and the deepest growth depression. Since coping with the stress factors is likely to be energetically costly this should result in a longer latency period until recovery starts. All other things being equal, a slower recovery speed during the subsequent compensatory growth phase might also be expected. However, the investment in compensatory growth may be traded off against the ecological and physiological costs associated with high feeding and growth rates, or may be dependent on the time constraint for reaching a certain size (Gotthard et al. 1994), and recurrent stress treatment may enhance the response, leading to overcompensation (Hayward et al. 1997). Therefore, the speed of recovery may not be simply related to the stressfulness of the treatments and exact predictions cannot be made.

Folsomia candida is considered toxic to the Wolf Spider Schizocosa because survival of hatchlings fed these collembola was lower than that of starved controls (Toft & Wise 1999). It could therefore be expected that F. candida would be a harder stress factor than starvation. Starvation was expected to be very stressful even though polyphagous predators often live with high hunger levels in the field (Juliano 1986; Bilde & Toft 1998), and spiders in particular are known to be adapted to periods of starvation (Wise 1993). Though hunger is a naturally occurring condition, spiders are vulnerable to starvation, especially during early life stages (Anderson 1974). As adults, the spiders are less sensitive (Miyashita 1968b). Spiders may survive starvation by decreasing their basal metabolic rate and by maintaining a relatively motionless sit-and-wait foraging strategy (Miyashita 1969; Anderson 1974; Wise 1993).

Rhopalosiphum padi was expected to be less stressful than starvation or F. candida but more than fruit flies. Rhopalosiphum padi has earlier been ranked as a low-quality prey to a range of generalist arthropod predators (Toft 1995; Bilde & Toft 1997). It is deterrent, as it induces specific feeding aversions (Toft 1997), but it is not toxic and may even have positive effects in mixed diets (Toft 1995). Drosophila melanogaster was expected to be the least stressful prey. It is a highly preferred prey on which most ground beetles and spiders rapidly become satiated, but its nutritional quality is generally insufficient to allow complete development (Mayntz & Toft 2001).

Materials and methods

the spider

The Wolf Spider Pardosa prativaga (Lycosidae) is abundant in Northern Europe in a variety of open biotopes, especially bogs, meadows and agricultural fields. The spider has a body length of 4–6 mm and has a mainly annual life cycle with a fraction of the population being biennial (S. Toft, unpublished observations). During the summer (May–August) the females carry their egg sacs for 2–3 weeks. After hatching, the spiderlings spend some days on their mother's abdomen before dispersing.

For the experiment, females with an egg sac were collected in a field at Stjær, eastern Jutland, Denmark, in June 2001. They were kept in the laboratory at 25 ± 0·2 °C until the egg sacs hatched. Three to five days after hatching the spiderlings were transferred individually to plastic tubes (diam. 2 cm, height 6 cm) with a moistened bottom layer of plaster mixed with charcoal. A foam rubber plug was used as lid. Throughout the experiment the spiders were kept at a constant temperature of 25 ± 0·2 °C and at a photoperiod of 16L : 8D.

stress factor

Starvation and pure diets of Folsomia candida, Rhopalosiphum padi and Drosophila melanogaster were used as stress factors. The cereal aphid R. padi and the collembolan F. candida used in this study were obtained from laboratory cultures. The aphid culture was maintained on wheat seedlings (mixed cultivars) and the collembola were raised on baker's yeast. Mixed stages of collembola and aphids were used to feed the spiders.

Two qualities of wild-type fruit flies, D. melanogaster, were used for the experiment. Both were reared on instant Drosophila medium (Formula 4–2, Carolina Biological Supply; Burlington, NC): ‘Normal’ flies (N-flies) on plain medium and ‘Enriched’ flies (E-flies) on medium mixed with crushed dog food (TechniCal®ADULT, Martin Pet Foods, Ontario, Canada) in a proportion of 100 g of Drosophila medium to 54·5 g of dog food. By enriching the Drosophila medium with dog food, the fruit flies become of a better nutritional quality and both spider growth and survival are increased (Mayntz & Toft 2001). Therefore ‘Enriched’D. melanogaster (E-flies) was used as food for the control treatment and was the food received by all spiders of the stressed treatments after the prescribed stress period ended.

experimental design

This study consisted of three experimental series, each of four treatments, and a control. All were run simultaneously, and the same control group was used for the three experimental series. For the whole set-up, siblings from 59 mothers were equally distributed over the 13 groups to prevent any bias due to maternal effects. The spiderlings were assigned to one of the five diet treatments: (1) starvation, (2) F. candida, (3) R. padi, (4) N-flies and (5) E-flies (control).

Each of the three experimental series consisted of a stress period followed by a recovery period. For Experiments 1 and 2 the time of the stress period was constant while in Experiment 3 the mortality (as an indicator of stress load) was the constant parameter. The stress periods were: Experiment (1) 7-day diet treatment; Experiment (2) 14-day diet treatment; and Experiment (3) diet treatment until approximately 60% of the individuals of each treatment group had died. The stress periods of 7 or 14 days were chosen based on mortality results from previous experiments (Toft & Wise 1999). A stress treatment resulting in 60% mortality was considered a strong stress treatment, also for the individuals that survived. Previous experiments showed that some spiders would die within the first days after the end of stress treatment. With a stress load equivalent to 60% mortality, a reasonable sample size remains for post-stress investigations, provided a large initial sample size is used.

After receiving the prescribed experimental diets for 7 or 14 days, spiders in Experiments 1 and 2 were fed E-flies until termination of the experiment, which was when the animals became adults or after 30 weeks at most. Spiders in Experiment 3 were fed the prescribed experimental diets until a mortality level of approximately 60% was reached. They were then fed E-flies until the end of the experiment as indicated above. The stress periods of Experiment 3 turned out as follows: starvation 10 days, F. candida 13 days, R. padi 16 days and N-flies 70 days. These figures give an indication of the relative stress intensity of the treatments.

During the stress periods all spiders were checked daily to ensure that they were never short of water or prey. Spiders that had only a few prey animals left received a new portion of 7–10 live prey. After the stress periods the spiders were watered, fed E-flies and checked for deaths and moults three to five times a week.

For the first 26 weeks of the experiment the spiders were weighed (Sartorius MC5; 0·001 mg resolution) once a week, after that every second week. The weight before the last moult was used as the ‘maturation weight’. Weight before rather than after the final moult was chosen because spiders stop feeding for some days before a moult, and to avoid any bias from the burst of feeding that spiders have immediately after a moult (Miyashita 1968a; Workman 1978).

Initial sample sizes for the experimental series were different to account for expected mortality. Mortality of spiders fed N- or E-flies was expected to be relatively low (Mayntz & Toft 2001) in Experiments 1 and 2. For spiders in the treatment groups of starvation, F. candida and R. padi, mortality was expected to be high (Toft 1995; Toft & Wise 1999; Bilde & Toft 2001) (Table 1).

Table 1.  Diet treatments, initial sample sizes, mortality and weight change during the stress period (for the control group: mortality and weight change after the first 7 and 14 days, respectively) and duration of the 1st instar (mean ± SD). For each experiment the same letter indicates no significant difference (Tukey-Kramer test)
Exp. no.Days of stress treatmentDietInitial NMortality during stress treatment (%)Mean weight change as percentage of mean initial weightDuration of 1st instar in days
  1. E-flies = Enriched Drosophila melanogaster (same control group for all experiments).

  2. N-flies = Normal Drosophila melanogaster.

 ControlE-flies78 0 107·3 9·3 ± 3·5 a
17N-flies39 0 109·8 9·7 ± 3·2 a
7R. padi5918·6   2·515·3 ± 3·2 ab
7F. candida5911·9  17·519·8 ± 4·1 b
7Starvation5913·6  −5·318·4 ± 6·0 b
ControlE-flies78 3·9 357·5 9·3 ± 3·5 a
214N-flies39 0 291·910·6 ± 3·5 a
14R. padi5950·9  14·819·6 ± 6·6 b
14F. candida5961·0   9·820·7 ± 10·2 b
14Starvation5994·9 −15·8None moulted
ControlE-flies78 9·3 ± 3·5 a
370N-flies9866·31772·010·8 ± 4·6 a
16R. padi9860·2  24·025·2 ± 4·7 b
13F. candida9857·1  18·924·0 ± 5·9 b
10Starvation9851·0  −7·324·8 ± 7·0 b

statistical analysis

Ideally the growth curves should be compared with repeated measures manova. However, animals that die before the end of the experiment will be deleted completely from this analysis, leaving too few replicates. Therefore, differences in growth between treatments were tested by means of anovas of the weekly weight measurements followed by Dunnett's test to compare the stress treatments with the control. Before the anova the data were tested for variance homogeneity using Levene's test. If variance homogeneity was not fulfilled, a Box Cox transformation was used. Even so, in Experiment 2 week 6 and in Experiment 3 week 5, 6 and 7 variance homogeneity was not fully obtained (P < 0·05). In these cases Welch's anova was used, followed by pairwise comparison of the control and the treatment groups and sequential Bonferroni adjustment of α-values (Rice 1989).

To graphically illustrate the results on growth compensation a weight ratio was calculated as the mean weight of a diet group divided by the mean weight of the control group. To minimize any bias in weight ratio due to spider death during the stress periods, all data from spiders that died during the stress periods were excluded from this analysis. The weight of spiders dying before 70 days was tested against spiders dying later than 70 days. A cut-off of 70 days was chosen because the growth curves and weight ratios were stable after this period of time.

Analysis of survivorship was performed for two periods: (1) during the stress period, to obtain a measure of treatment intensity, and (2) during the compensatory phase. Survivorship analyses were done using the log rank test (Pyke & Thompson 1986). All statistical analyses were performed with JMP, version 4·0.


treatment intensity

Mortality during the stress periods in Experiments 1 and 2 showed significant differences between the treatment groups (overall log rank test, P < 0·05) (Table 1, Fig. 1), in that the E- and N-fly groups survived considerably better than starved spiders and spiders fed F. candida or R. padi. The latter showed no significant differences from each other (P > 0·05). In Experiment 3 mortality during the stress periods was close to the prescribed 60% for all treatment groups.

Figure 1.

Survivorship curves for each experimental group for the whole experimental period (stress period + compensatory phase). Stress periods for the three experiments in days were: Exp. 1 (a) 7 days, Exp. 2 (b) 14 days, Exp. 3 (c) until a mortality level of approximately 60% was reached. The stress periods of Experiment 3 varied as follows: N-flies 70 days, R. padi 16 days, F. candida 13 days and starvation 10 days. Arrows indicate when the stress periods ended. The letters x, y and z indicate significant differences in survival (pairwise log rank tests).

In all three experiments, only starved spiders lost weight during the stress periods, while spiders from the other treatments gained weight (Table 1). The duration of 1st instar reflects the diet treatment during the stress period. In Experiments 1, 2 and 3 the duration of 1st instar showed significant differences between the diet treatments (anova, all P < 0·0001, Exp. 1: F4 = 49·98, Exp. 2: F3 = 34·59, Exp. 3: F4 = 56·85). In Experiment 1 the duration of 1st instar in the starvation and F. candida groups was significantly longer than in the fruit fly groups (Tukey-Kramer, all P < 0·05), with the R. padi group intermediate. In Experiments 2 and 3 the duration of 1st instar in the starvation, F. candida and R. padi groups was significantly longer than in the fruit fly groups (Tukey-Kramer, all P < 0·05). Survival and weight change during the stress periods and duration of 1st instar confirm that F. candida and R. padi are nutritionally insufficient during early life stages when they are the only food.

mortality and development during the compensatory phase

Survivorship curves for the full experimental period are shown in Fig. 1. Excessive mortality rates were observed in the stress treatment groups for several days after the treatment stopped, but eventually levelled off to become negligible. Overall tests for survival during the phase of compensation (i.e. neglecting individuals that died during the treatment phase) revealed a significant treatment effect (log rank test, Exp. 1: χ2 = 49·72, df = 4, P < 0·0001, Exp. 2: χ2 = 69·97, df = 4, P < 0·0001, Exp. 3: χ2 = 79·80, df = 4, P < 0·0001). In all experiments, survival during this phase was highest on fruit flies while the starved spiders and spiders fed F. candida had the lowest survival. Spiders reared on R. padi had intermediate survival. The mortality of the N-fly diet group increased from Experiment 1 through Experiment 2 to Experiment 3 compared with that of E-flies, but the difference between the two fly treatments was significant only in Experiment 3.

Some time after the stress period, but not immediately after, the starved spiders started to gain weight and most spiders from the other treatment groups gained weight at a higher rate (Fig. 2a,c,e). This was not the case in Experiment 2 for starved spiders, because the few remaining individuals of this group died shortly after the stress period ended. During the compensatory phase the duration of instars (2nd to 7th instar) showed no significant differences within each of the three experimental series (anova, all P > 0·05) nor between the three series (anova, all P > 0·05). The duration (mean ± SD) of the instars was as follows: 2nd 8·8 ± 5·5 days (N = 289), 3rd 8·8 ± 5·5 days (N = 247), 4th 16·7 ± 14·3 days (N = 235), 5th 45·1 ± 26·0 days (N = 215), 6th 48·7 ± 21·0 days (N = 140) and 7th 54·4 ± 21·1 days (N = 41).

Figure 2.

Weight curves and weight ratio curves for a 7-day stress period (a, b), a 14-day stress period (c, d) and a stress period until a mortality level of approximately 60% was reached (e, f). Weight ratio was calculated as the mean weight of a diet group divided by the mean weight of the control group. For clarity the error bars of the weight curves are only marked every 5 weeks. At the lower right corner of each weight graph, numbers mark weeks when the indicated treatments are significantly different from the control (anova with Dunnett's tests). Week numbers in italics fall within the stress period.

compensatory growth

For starved spiders and spiders fed F. candida or R. padi the weight ratio curves (Fig. 2b,d,f) in all three experiments declined dramatically during the stress periods. After the stress periods, the decline of the weight ratio curves continued for approximately 1 week, after which the animals started to gain weight at an increased rate. For spiders fed N-flies the weights in Experiments 1 and 2 showed no differences from the control spiders during either the stress periods or the compensatory phases. In Experiment 3 the weight curve for spiders fed N-flies was below the one for spiders fed E-flies (Fig. 2e) from week 2 onwards. The weight ratio curve for N-flies (Fig. 2f) declined almost until the end of the stress period of 70 days.

In Experiments 1 and 2, spider weights of the four nutritionally insufficient treatments and the control showed significant differences after 1 week until after 6 weeks (anova on the weekly weight measurements, all P < 0·0006. Exp. 1, week 1: F4 = 106·43; week 2: F4 = 51·91; week 3: F4 = 39·72; week 4: F4 = 27·45; week 5: F4 = 12·43; week 6: F4 = 5·19. Exp. 2, week 1: F4 = 33·30; week 2: F4 = 57·87; week 3: F3 = 34·74; week 4: F2 = 37·19; week 5: F2 = 18·50; week 6: F2 = 9·62; Fig. 2a,c). From week 7 to week 26 no significant differences were found (all P > 0·05). Thus all the surviving spiders had compensated for the setback in growth. In Experiment 1 significant differences between the R. padi treatment group and the control group were observed from week 1 to week 4 (Dunnett's test, P < 0·05) while the F. candida and starvation treatments were significantly different from the control from week 1 to week 6 (P < 0·05) (Fig. 2a). In Experiment 2 the F. candida and the R. padi treatment groups were significantly different from the control from week 1 to week 6 (P < 0·05). The starvation group was significantly different from the control until week 2 (P < 0·05), after which the group was excluded from analysis owing to mortality (Fig. 2c).

In Experiment 3, weights of the four stressful treatments and the control showed significant differences after 1 week until after 12 weeks (anova, all P < 0·0001. Week 1: F4 = 44·61; week 2: F4 = 49·49; week 3: F4 = 40·04; week 4: F4 = 31·88; week 5: F4 = 26·13; week 6: F4 = 18·96; week 7: F4 = 12·78; week 8: F4 = 9·60; week 9: F4 = 6·07; week 10: F4 = 6·30; week 11: F4 = 4·34; week 12: F4 = 2·82; Fig. 2e). From week 13 to week 26 no significant differences were found (all P > 0·05) and all the surviving spiders had compensated for the setback in growth. The F. candida treatment was significantly different from the control from week 1 to week 5, R. padi treatment from week 1 to week 7, and starvation from week 1 to week 8 (Dunnett's test, all P < 0·05). N-flies were only significantly different from the control in Experiment 3 where they showed differences from week 4 to week 12 (Dunnett's test, P < 0·05).

The weights of animals that died during the first 70 days were tested against the weights of animals that survived for more than 70 days within each experiment. This was done to see if animals dying early in the experiments were smaller than individuals surviving for longer periods of time. In Experiment 1 no significant differences were found between weights of spiders dying early and spiders dying late in the experiment for any of the treatments (Paired t-test, df = 1, all P > 0·05). In Experiment 2, the four nutritionally stressful treatments showed significant differences (all P < 0·05) between spiders dying early and spiders dying late in the experiment. In all treatment groups the small spiders died first. In Experiment 3, weights of spiders from the starvation and the F. candida treatments showed significant differences (all P < 0·05), the small spiders dying first.

number of individuals maturing and total development

The number of individuals of each treatment group that reached maturity and tests for the proportion of individuals reaching the adult stage compared with the proportion of adults in the control group are presented in Table 2. The diet during the early life stages had an impact on the survival and on the developmental processes and thereby on the number of individuals reaching maturity. Only the group of spiders fed N-flies and those fed R. padi for 7 days showed no significant reduction from the control spiders in the proportion of individuals reaching the adult stage.

Table 2.  Total development time to maturity, the number and percentage of individuals of each treatment group that reached maturity, and tests for the proportion of individuals reaching the adult stage compared with the control group (E-flies)
Exp. no.DietInitial NTotal dev. time** (days) ± SDNo. of adults% adults of initial Nχ2*dfP
  • *

    Yates corrected.

  • **

    Total development time to maturity (days).

  • E-flies: Enriched Drosophila melanogaster (same control group for all experiments).

  • N-flies: Normal Drosophila melanogaster.

Control 1E-flies78138·7 ± 37·53646
N-flies39141·3 ± 40·92051 1>0·05
R. padi59139·0 ± 46·81220 1>0·05
F. candida59148·9 ± 34·5 712 8·991   0·0027
Starvation59161·3 ± 38·2 915 6·751   0·0094
Control 2E-flies78138·7 ± 37·53646
N-flies39149·9 ± 37·41128 1>0·05
R. padi59167·3 ± 32·81118 4·941   0·0263
F. candida59200·0 ± 4·2 2 317·091<0·0001
Starvation59 0 021·651<0·0001
Control 3E-flies78138·7 ± 37·53646
N-flies98140·0 ± 40·7141410·851   0·0010
R. padi98149·3 ± 24·5 8 818·351<0·0001
F. candida98126·8 ± 13·1 5 523·511<0·0001
Starvation98174·0 ± 24·4 4 425·491<0·0001

Total development times (Table 2) differed in some experiments depending on stress factors and stress periods. In Experiment 1, spiders from the four nutritionally stressful groups and the control showed no significant differences in total development time to maturity (anova, F4 = 0·55, P = 0·70) or in the number of moults to maturity (anova, F4 = 0·44, P = 0·78), which was 5·9 ± 1·2 moults (average ± SD).

In Experiment 2, no starved spiders became adults. The remaining stressful treatment groups all took longer to mature than the control group, the difference being significant for the F. candida group (anova, F1 = 5·20, P = 0·029) and the R. padi group (anova, F1 = 5·17, P = 0·028) but not for the N-fly group (anova, F1 = 0·76, P = 0·39). No significant differences were found in the number of moults to maturity (anova, F3 = 2·28, P = 0·09). On average (± SD) the spiders moulted 6·2 ± 0·4 times.

In Experiment 3 the total development time to maturity showed no significant differences between the four stressful treatment groups and the control (anova, F4 = 1·22, P = 0·31). Nor did the number of moults to maturity show significant differences (anova, F4 = 1·78, P = 0·15). On average (± SD) the spiders moulted 6·2 ± 0·7 times. In none of the three experiments were there significant differences in the sex ratio of spiders becoming adults (anova, all P > 0·05). Moreover no significant differences between the sexes were found in total development time to maturity (anova, all P > 0·05).

Diet during the stress period turned out to have an impact on the spiders’ weight at moulting in Experiment 1 in the first and second moult (anova, 1st moult: F4 = 3·72, P = 0·007, 2nd moult: F4 = 3·05, P = 0·02). Spiders previously fed F. candida were significantly smaller at the first and second moults than spiders of the starvation group, which in turn were smaller than spiders from the R. padi, N- and E-fly groups (Tukey-Kramer, P < 0·05). In Experiments 2 and 3, no significant differences in weight at moulting were observed (anova, all P > 0·05).


growth compensation

The results show that Pardosa prativaga is able to respond to short-term nutrient deprivation after the stress treatment is alleviated. After a period of nutritionally stressful treatment P. prativaga responded to the availability of high-quality flies by increasing the growth rate. This compensatory response occurred regardless of the quality of the diet given during the stress period, but the strength of the response varied according to the diet's nutritional quality and the duration of the stress period. Statistically, surviving spiders in Experiments 1 and 2 exhibited weight compensation after 6 weeks. However, the weight curves for starvation, F. candida and R. padi, lie below the curves for N- and E-flies almost throughout the experiment, indicating that growth compensation was actually slightly incomplete. In Experiment 3 all individuals that survived more than 13 weeks exhibited weight compensation. Comparison of weights of animals that died during the first 70 days and those that survived for more than 70 days revealed that small spiders generally had a higher mortality than larger ones.

In many details the responses of the spiders comply well with the compensatory growth model of Broekhuizen et al. (1994) based on fish studies. Thus, the reduction of growth rates was particularly rapid in the initial phase of a stress period. Under nutritional stress animals may experience a considerable decrease in reserves. Spiders respond by reducing their basal metabolic rate during starvation (Anderson 1974), thus reducing the rate of weight loss. Recovery did not start immediately upon introduction of high-quality food but took several days to be initiated. Presumably it takes some time to re-establish a fully functioning metabolism (Broekhuizen et al. 1994). However, after recovery had begun, growth accelerated and remained high until most or all the weight deficit had been regained, after which the previously stressed animals followed the same growth trajectory as the control group.

In Experiment 3 a slower rate of compensation was observed in association with starvation, which corroborates the hypothesis of Nicieza & Metcalfe (1997) that growth reduction during starvation is more severe than that associated with other stress treatments. Since F. candida is toxic to P. prativaga, it was expected to cause greater mortality than starvation, as observed by Toft & Wise (1999), and slower compensation. In Experiment 1, the starved spiders and spiders fed F. candida seemed to compensate equally fast, while in Experiment 3, spiders from the F. candida group compensated faster than the starved spiders. This might be because even food of low quality contains some nutrients (Miyashita 1968b) and because the starved spiders, which survived the stress period in Experiment 3, had a greater size deficit to recover from. A low amount of food or availability of low-quality food (e.g. F. candida or R. padi) may be insufficient for growth and development, but may allow survival for extended periods (Miyashita 1968b; Toft 1995). Thus, spiders fed F. candida did not lose weight during the stress periods, but actually gained a little weight (10–19% of initial weight). In some respects, F. candida is toxic to P. prativaga (E. N. Fisker & S. Toft, unpublished observations) and influences the activity of detoxification enzymes in the spider (Nielsen & Toft 2002). A doubling of the respiration rate was found in P. prativaga fed a single-prey diet of F. candida (Toft & Nielsen 1997).

The spiders fed R. padi all gained weight during the stress periods (1–15% of initial weight), but at a lower rate than the spiders fed fruit flies (107–1772% of initial weight). Shortly after the stress period the spiders fed R. padi accelerated their growth and it remained high until the weight deficit had been recovered after 3–5 weeks. This quick recovery is probably due to the fact that R. padi is only deterrent, but not toxic to P. prativaga in the amounts eaten (Toft 1995). In Experiments 1 and 2 spiders fed R. padi compensated faster than spiders fed F. candida. In Experiment 3 the two groups seemed to compensate equally fast, though statistically F. candida compensated faster than R. padi. This is certainly a sample-size effect since only two spiders from the F. candida group survived throughout the experiment. Spiders fed N-flies for 7 or 14 days had weight curves similar to those of control animals fed E-flies though the spiders fed N-flies for 70 days gained weight at a lower rate during the stress period. In this experiment (Experiment 3) growth accelerated initially after the stress period and remained high until the weight deficit had been recovered.

Early nutritional stress can affect many parameters of development. Total development time is a function of the number of instars and the instar duration (Higgins & Rankin 1996). Significant differences in total development time to maturity were observed only between spiders from the groups of F. candida and R. padi and the control with a stress period of 14 days. In Lepthyphantes tenuis (Blackwall) fed R. padi until the first moult, Beck & Toft (2000) observed that the total development time increased, though only slightly, compared with controls. No increases were observed in the number of instars to maturity with a reduction of nutrients, so the increased time for total development in Experiment 2 must be an increase of the instar duration. Miyashita (1968a) showed that in Pardosa astrigera L. Koch (syn. Lycosa T-insignita (Boes. et Str.)) the number of instars to maturity increased with a decrease in food supply. Duration of the first instar, which was partly within the stress period, was longer for individuals from the starvation, F. candida and R. padi groups than for spiders fed fruit flies. For Leptyphantes tenuis fed R. padi until the first moult the same tendency was found (Beck & Toft 2000).

The hypothesis that spiders should respond differently to various types of nutritional stress was confirmed. It was expected and largely confirmed that the N-flies would be the least stressful of the treatments, that the toxic F. candida would be the most stressful and the deterrent R. padi intermediate. However, the expectation that a toxic prey should be a stronger stress factor than starvation was not supported. The results presented pertain to a specific selection of one of each type of stressful prey. It is possible to imagine that other prey types may be more seriously deficient in certain nutrients than the N-flies used here, and other toxic prey may have toxins with more serious effects than F. candida. Experiments with more prey types need to be done before final generalizations are possible.

The ability for growth compensation probably evolved from strong selection on the spiders to synchronize their reproduction to a specific season. From an analysis of geographical patterns of spider life cycles, Toft (1976) concluded that, within a species the mating and egg-laying periods were always the same in annual and biennial parts of the distribution area and in annual and biennial fractions of a population in one area, even if the spiders completed their maturation moult at different seasons. This is true whether the species is spring, summer or autumn breeding (Toft 1976). Thus, the alternatives to compensatory growth with reproduction at the ‘normal’ time and at the ‘normal’ size are (1) reproduction at a subnormal size, or (2) adding a full year to the developmental period. The first option has large fecundity costs (Schaefer 1987). The second option may allow the spider to mature at a larger than normal body size and thus give a fecundity advantage but has costs in terms of increased mortality by predation and other risks. As long as the costs of compensatory growth are smaller than the costs of either of these alternatives, compensation should be the chosen response. However, it is possible that spiders can be stressed at a time during their development or for such a long period that they are unable to recover sufficiently in due time, and are therefore forced to add a year to their development. In this case they might not utilize their capacity for compensatory growth and they may save these costs. On the contrary, they may take advantage of the time available and use a slower-than-normal growth trajectory, and reach the maturation size at the proper season without having to pay the costs of a high foraging rate and a high growth rate. This may be the reason for the frequent occurrence of mixed annual–biennial life cycles in spiders (Toft 1976).


We thank Søren Aasberg and Shams El-Shamy for help in the laboratory and David Mayntz for help with the statistical analysis. The study was supported by grants from Carlsberg Foundation and the Danish Natural Science Foundation.