Introduction

Hermit crabs occupy empty gastropod shells, and these shells act as shelters from biotic factors including predation (Kuhlmann 1992; Angel 2000), and abiotic factors such as desiccation (Taylor 1981; Brodie 1999) and osmotic stress (Shumway 1978; Pechenik et al. 2001). Particular gastropod shells are non-randomly chosen based on shell characteristics that optimally match requirements of individual hermit crabs (Reese 1963; Conover 1978; Brooks and Mariscal 1985; McClintock 1985; Cote et al. 1998). Shell choice occurs following shell investigation, which can vary depending on hermit crab condition and gastropod shell quality (Elwood and Stewart 1985; Elwood 1995; Elwood and Briffa 2001). These portable shell habitats are exchanged frequently for new shells as hermit crabs grow, thus shell choice behavior continues to be important throughout the lifetime of a hermit crab (Reese 1962; Hazlett 1981).

The Atlantic long-wristed hermit crab, Pagurus longicarpus (Say 1817), commonly occupies shells of the periwinkle, Littorina littorea, which are vacated following natural death or predation. Naticid gastropods kill L. littorea by drilling an access hole through the shell, and such drilled shells dominate the population of vacant shells available for P. longicarpus in some populations (Pechenik and Lewis 2000). Shell availability has long been known to influence hermit crab populations (Scully 1979, 1983; Blackstone 1985; Carlon and Ebersole 1995). Yet, despite scarcity of available shells, P. longicarpus avoids drilled or damaged shells to the point of sacrificing ideal shell thickness and size (McClintock 1985; Wilber 1990; Pechenik and Lewis 2000).

Predator-induced changes in prey behavior may have important implications for prey population dynamics and community interactions (Trussell et al. 2002). While behavioral plasticity in response to predator cues has been demonstrated for many species (Van Buskirk 2002; Relyea 2003), the impact of predator risk on hermit crab shell choice has not been addressed. Tropical hermit crabs gain protection from shell-crushing crab predators by their preference for gastropod shells with protective structural traits, such as shell thickness and small aperture size (Bertness 1981; Bertness and Cunningham 1981). In the northeastern United States, the invasive green crab predator, Carcinus maenas, preys upon hermit crabs (Ropes 1968; Elner 1981). These predators use existing structural damage, including drill holes, to break open gastropod shells and gain access to hermit crab inhabitants (Pechenik et al. 2001). As a result, P. longicarpus hermit crabs inhabiting drilled gastropod shells are more vulnerable to green crab predation than hermit crabs occupying intact shells (Pechenik et al. 2001). Taken together, these studies suggest that hermit crabs would have an advantage if they were able to assess predation risks and alter their shell choice behavior to reduce vulnerability to potential predators.

In this study, we investigated whether P. longicarpus hermit crabs respond to increased predation risk by modifying their shell choice behavior. In addition, since previous studies have demonstrated both morphological and behavioral differences among hermit crab populations (Benvenuto and Gherardi 2001), we compared P. longicarpus from two sites differing in drilled shell availability. In laboratory trials, individual crabs were presented with a choice of drilled versus intact L. littorea shells in the presence and absence of chemical cues from C. maenas predators.

Methods

Hermit crab collection and maintenance

P. longicarpus hermit crabs inhabiting L. littorea shells were collected from two sites along the Massachusetts coastline: Long Beach, Nahant (42°25.5′N, 70°55.1′W) and West Beach, Beverly (42°33.5′N, 70°52.8′’W) in May and June 2002. Although L. littorea represents the most common shell type inhabited by P. longicarpus at both sites, the two sites differ in the frequency of drilled L. littorea shells. Drilled shells represent up to 73% of all vacant L. littorea shells at Nahant (with some seasonal variation; Pechenik and Lewis 2000), while drilled shells represent only ~14% of all vacant L. littorea shells at the Beverly site (based on 100 vacant shells haphazardly collected on 8 May 2002). The predatory green crab, C. maenas, occurs at both sites (we were unable to locate sites with P. longicarpus but lacking C. maenas).

Approximately 100 hermit crabs were collected from each site on each of three collection dates, represented by blocks as described below. Ovigerous females (evident by the presence of egg masses) were not used for this study and were returned to their original location. Hermit crabs were maintained in 75.8-l (20-gallon) aquaria in Instant Ocean seawater (27–32 ppt salinity, ~22°C) at a density of ~60 hermit crabs per tank. Hermit crabs were held for 1 week prior to experimentation and were fed imitation crab meat (pollock).

Prior to the experiment, hermit crabs were removed from their shells by pulling them as gently as possible, and were then placed in individual 100 ml containers with seawater for 24 h. For this experiment, only intact individuals with both chelipeds and all legs (~90% of crabs extracted) were used since tactile cues are likely to be important for shell investigation and selection (Reese 1963; Scully 1986). Wet mass was measured for each crab, and optimal L. littorea shell size was determined using the equation developed for P. longicarpus (Angel 2000): log(shell aperture length in mm)=0.2623(log crab mass inmg)+0.4342

For each crab, two similarly-sized L. littorea shells, one with a natural drill hole and one intact, were identified: these shells were within 0.3 mm of the optimal aperture size. Before being used in experiments, shells were soaked in hydrogen peroxide for 15 min to remove organic residues, and then soaked in freshwater for 24 h. Only shells with no visible defects (other than drill holes) were used.

Experimental conditions

To investigate whether P. longicarpus shell choice behavior changed in the presence of predator cues, we conducted tests with each crab in both predator effluent (green crab-conditioned seawater) and normal seawater. Predator effluent was obtained by keeping six C. maenas collected from Beverly (weight range 1–16 g, mean=6.2 g) in 8.5 l Instant Ocean for 48–72 h. Green crabs were fed imitation crab meat (identical to hermit crab diet) until approximately 48 h before the experiment. Mechanisms for shell selection using sensory information have previously been explored in hermit crabs (Reese 1963; Hazlett 1981). Chemical cues may be used by hermit crabs for predation risk assessment (Hazlett 1996, 1997; Rittschof and Hazlett 1997), shell discrimination (Hazlett 1982), attraction to gastropod predation sites (McLean 1974; Rittschof 1980; Hazlett 1996, 1997), and attraction to dead conspecifics (Rittschof et al. 1992; Hazlett 1996). Based on these previous studies, we used seawater with green crab effluent as a predation threat in this experiment.

After 24 h of isolation, naked hermit crabs were each transferred to rectangular 8×12-cm opaque plastic arenas containing 120 ml predator effluent or control seawater. Each hermit crab was offered a drilled and an intact L. littorea shell, matched to individual crab wet mass as described above. Shells were approximately 8 cm apart, and the crab was placed equidistant from the two shells. Crabs were observed continuously for 15 min, during which behaviors were recorded using The Observer 3.0 software (Noldus Technology). For each crab, the following behaviors were recorded: the number of contacts with each shell type (drilled and intact), the duration of shell investigation (in seconds) for each shell type, and the number of times (and duration) a crab occupied each shell type. The experimental design was paired, such that each hermit crab (64 from Nahant, 59 from Beverly) experienced predator effluent and control treatments in random order. Crabs were removed from their shells following each 24-h trial and again isolated for 24 h before starting the next treatment. Following behavioral observations, crabs remained in their arenas and final shell choice was recorded after 24 h.

Data analysis

We separately measured (in seconds) the duration of initial shell investigation by naked crabs and the duration of continued investigations by crabs already in shells. Latency to shell occupancy was calculated as the time it took for each naked crab to move into a shell. Total duration of shell occupancy in either shell was calculated as the sum of time each crab spent occupying either shell, and this was expressed as a percentage of the 15-min observation period. Intact shell occupancy was calculated as the percentage of the total shell occupancy duration that a crab spent in the intact shell. We also analyzed shell-switching, the number of times hermit crabs switched between the two shells. Finally, shell choice was examined by comparing the number of hermit crabs occupying intact versus drilled shells at three time points: initial (first shell occupied), at 15 min, and after 24 h. For the analysis of latency to shell occupancy, intact shell occupancy, and shell-switching behavior, hermit crabs remaining naked for the entire trial were excluded.

We compared shell investigation and occupancy behavior of hermit crabs between predator effluent and control treatments using paired t-tests to control for between-crab variability. To check for any block effects (three hermit crab collection dates) or site by predation treatment interactions, we also used repeated measures ANOVAs that included block, site, and predation treatment as a repeated measure. As there were no significant block effects, we removed blocks from the statistical model. Data were examined to check assumptions of each analysis. We report paired t-tests within sites when comparing the behavior of the same crabs with and without predator cues, and we report repeated measures ANOVAs when examining possible predator × site interactions. To account for the fact that naked crabs generally investigated only one shell before entering it (thus spending no time investigating the alternate shell from a naked state), investigation times of zero were eliminated: a repeated measures ANOVA was thus conducted on the subset of crabs that investigated the same shell in both predator effluent and control treatments.

To determine if predator effluent altered hermit crab shell choice, we used contingency table tests of heterogeneity to compare the proportion of crabs occupying intact versus drilled shells between treatments. All statistical analyses were conducted using Systat 10 (SPSS).

Results

Shell investigation

Overall, hermit crabs spent approximately twice as much time investigating intact shells as drilled shells (Fig. 1). Exposure to predator effluent did not significantly alter the total amount of time hermit crabs spent investigating drilled shells (Fig. 1; Beverly paired t=1.70, df=58, P=0.09; Nahant paired t=0.67, df=62, P=0.51). When investigating intact shells, hermit crabs from the two sites responded differently to predator effluent: Nahant hermit crabs spent more time investigating intact shells in the presence of predator effluent, whereas Beverly hermit crabs showed the opposite behavior (Fig. 1; repeated measures ANOVA predator × site interaction F 1,117=5.29, P=0.023).

Fig. 1
figure 1

Investigation times (s) by P. longicarpus hermit crabs of intact and drilled L. littorea shells in predator effluent versus control treatments during 15-minute observation periods. Means+1 SE are represented for hermit crabs from two different populations

Full size image

We also partitioned shell investigation behavior into time spent by naked crabs investigating either shell type, and time spent in continued investigation by crabs already in shells. There was no effect of predator effluent on the time naked crabs spent investigating drilled or intact shells (Tables 1, 2). However, there was a significant predator × site interaction for naked crabs investigating drilled shells (Table 2); crabs from the Beverly site that were exposed to predator effluent spent more time investigating drilled shells (Table 1).

Table 1 Comparison of P. longicarpus behavior between predator effluent and control treatments. Latency to occupy either of two L. littorea shells and the number of times hermit crabs switched between them was measured during 15-min observation periods for crabs from two sites. Shell investigation time by naked crabs and the duration of continued investigations by crabs already in shells were measured separately during the observation period
Full size table
Table 2 Statistical analysis of P. longicarpus behavior for naked hermit crabs initially investigating drilled or intact L. littorea shells, and for further investigation once crabs have entered either shell type. Reported F and P-values based on repeated measures ANOVAs for behavioral differences with and without predator cues (predator effect), differences between two sites, and interaction effects. For the naked crab investigations: df=1,35 drilled; df=1,33 intact; for all investigations from either shell type: df =1,122
Full size table

Shell occupancy

The percentage of the 15-min observation period that hermit crabs spent in either shell did not differ between predator effluent and control treatments (Fig. 2i, Beverly paired t=0.04, df=58, P=0.965; Nahant paired t=1.03, df=63, P=0.306). Interestingly, across both sites hermit crabs spent significantly more time in intact shells in the presence of predator effluent compared to control treatments (Fig. 2ii; repeated measures ANOVA, predation treatment F 1,82=10.18, P=0.002). This difference was most pronounced at the Nahant site (Beverly paired t=1.81, df=43, P=0.077; Nahant paired t=2.37, df=43, P=0.022).

Fig. 2
figure 2

Percentage time that P. longicarpus hermit crabs in predator effluent versus control treatments spent occupying L. littorea shells (does not include naked crabs). Means+1 SE are represented for hermit crabs from two different sites. i Percentage of 15-min observation time spent occupying either shell (Beverly n=59, Nahant n=64). ii Percentage time spent occupying intact shells out of total time spent in either shell (Beverly n=44, Nahant n=43). Asterisks indicate probability <0.05

Full size image

There was no difference in the amount of time taken for hermit crabs to occupy either shell (latency) between the predator effluent and control treatments (Table 1; Beverly paired t=0.77, df=52, P=0.440, Nahant paired t=0.33, df=56, P=0.743). Once hermit crabs initially occupied a shell, we examined the number of times they switched to occupy the other shell, which ranged from zero to three, and was not affected by predator effluent at either site (Table 1; Beverly paired t=1.12, df=51, p=0.266 Nahant paired t=0.65, df=56, p=0.517).

Shell choice

More hermit crabs initially chose intact shells in the predator effluent compared to the control treatment; this difference was not statistically significant (data combined for both sites, 2×2 test of homogeneity, G=3.65, df=1, P=0.056). At the end of the 15-min observation period, a large percentage of hermit crabs in both treatments had moved into intact shells (Table 3). This percentage did not differ between the two treatments (2×2 test of homogeneity, G=0.076, df=1, P=0.783). After 24 h, nearly all hermit crabs occupied intact shells and again there was no difference between the two treatments (2×2 test of homogeneity, G=0.867, df=1, P=0.352).

Table 3 Effect of predator effluent on the number of P. longicarpus hermit crabs occupying drilled or intact L. littorea shells, or remaining naked, at three different time points: initially (first shell occupied), at 15 min, and at 24 h
Full size table

Discussion

This study shows that effluent of the invasive predatory green crab, C. maenas, influences P. longicarpus assessment and choice of drilled versus intact L. littorea shells. We found that hermit crabs spent significantly more time occupying intact shells in the presence of predator effluent than they did in control treatments (Fig. 2ii), indicating that hermit crabs modify their shell selection behavior in the face of predation risk. Overall, hermit crabs spent less time investigating drilled shells than intact shells (Fig. 1), suggesting that drilled shells are relatively quickly identified and then avoided by hermit crabs. Previous studies have shown that P. longicarpus hermit crabs avoid shells with damage, including holes drilled by naticid gastropods, and prefer to occupy shells of suboptimal fit (McClintock 1985; Wilber 1990; Pechenik and Lewis 2000). P. longicarpus hermit crabs forced to occupy drilled shells are more vulnerable to predation, as C. maenas exploit drill holes to gain access to hermit crab inhabitants (Pechenik et al. 2001). Green crabs attack hermit crabs in drilled shells by enlarging the drill hole until they can remove the hermit crab from its shell. However, hermit crabs can survive attacks by predatory green crabs; in laboratory predation trials, 25% of hermit crabs in drilled shells remained alive after 24 h (Pechenik et al. 2001). Thus, hermit crab preference for intact shells may result from intense selection against hermit crabs occupying drilled shells.

Our results indicate that predator cues did not affect hermit crab shell switching behavior or the percentage of hermit crabs that occupied intact shells at 15 min or 24 h. This is consistent with previous research which has shown that P. longicarpus switch into preferred shells, even in the presence of other predators (Brooks and Mariscal 1985). Our results are also consistent with previous studies showing strong avoidance by P. longicarpus of drilled L. littorea shells (Pechenik and Lewis 2000).

Shell investigation behavior by hermit crabs varied between two sites. Predator cues increased the amount of time that Nahant hermit crabs spent investigating intact L. littorea shells, whereas the opposite occurred in Beverly hermit crabs (Fig. 1). One possible explanation for this is that these sites differ in the availability of intact shells. Among vacant L. littorea shells, intact shells are rare at Nahant (20–30%; Pechenik and Lewis 2000), while much more common (86%) at Beverly. As a result, P. longicarpus hermit crabs from Nahant are likely to have encountered few intact shells and therefore increased investigation of intact shells might be necessary to confirm shell condition. P. longicarpus from Beverly showed reduced investigation of intact shells when exposed to predator effluent, perhaps because predation threat reduces exploratory behavior in general. Interestingly, naked crabs from the Beverly site that were exposed to predator effluent spent more time investigating drilled shells (Table 1). This might be due to crabs with more limited experience with drilled shells taking longer to assess a shell before eventually accepting or rejecting it. An alternate explanation for the observed predator by site interaction might be a difference in the risk of predation by green crabs between sites.

Previous comparisons of hermit crabs from different populations have demonstrated both morphological and behavioral plasticity (Benvenuto and Gherardi 2001). Shell choice behavior of P. longicarpus from different populations differed in response to shell availability (Scully 1979). The different responses to predator threat observed in this study suggest that crabs from different P. longicarpus sites may differ in their motivation to find optimal shells, perhaps influenced by their prior experience with the disadvantages of occupying drilled shells.

These results suggest that predation threats may be a key determinant of hermit crab shell investigation and choice behavior. The green crab C. maenas has likely been an important predator of hermit crabs throughout the Atlantic rocky intertidal in Europe, and in the United States as well, following its multiple introductions over the past 150 years (Grosholz and Ruiz 1996). Despite the brevity of their local co-occurrence, our data show that P. longicarpus hermit crabs recognize chemical cues from C. maenas effluent and modify their shell investigation behavior in response. Future studies might further investigate P. longicarpus response to C. maenas or other crab predators by exploring how differences in hermit crab early experience with predators and shell availability affect shell choice behaviors.