Introduction

Information is crucial for organisms making adaptive decisions in various contexts. Better-informed individuals can adjust their behavior to meet demands1, and many species have information about both their current condition and past experience2,3,4. For example, juveniles of the perch Perca fluviatilis decrease their feeding rate in response to a visual cue of a predator and increase the intensity of their antipredator response when they have both visual and olfactory cues5. Males of the wolf spider Schizocosa ocreata can modify their courtship display according to current and past conditions, such as the number of courting male stimuli, the presence of female cues, and field experience6. Females of the Japanese pygmy squid Idiosepius paradoxus tend to abstain from attacking larger prey after learning that large prey are less likely to be captured7.

Contests for limited resources involve information use. Since contest outcomes are strongly affected by asymmetry of fighting ability or resource-holding potential (RHP)8 between contestants, information related to RHP asymmetry is key to the decision to give up or persist in contests9,10. Since weaker contestants often incur a greater cost than stronger contestants11,12, contestants that are potentially weaker but better informed should give up earlier to reduce costs on the basis of their assessment of their own lower RHP and/or relatively lower RHP to opponents’. Body size is a common information source used to assess RHP, and smaller contestants give up sooner13. Past contest outcomes and previously established dominance hierarchy also affect giving-up decisions in weaker contestants14,15,16,17. These experiences might provide information for re-estimation of their own RHP (i.e., self-assessment)18,19, as an additional source of information for assessment of their opponent’s RHP (i.e., social cue)20, and as a reliable source of information for assessment of RHP asymmetry against a particular opponent (i.e., familiar recognition)21.

Males of hermit crabs in the genus Pagurus have direct contests for mates during precopulatory guarding22,23, in which the male grasps the aperture of the gastropod shell occupied by a sexually mature female over several days24,25. The male–male contests are initiated by physical aggression of solitary intruders22,23 and are often settled in favor of the larger male22,26,27. Smaller males are therefore potentially weaker in this context. To decide when to give up the contest, when they encounter larger guarding opponents, smaller intruders might gather information about their chance of success, such as their relative size, their own recent defeats or their opponent’s recent wins, and the dominance hierarchy with their opponent.

Yasuda et al.28 examined whether these four factors affect the decision of P. middendorffii intruders to give up by using pairs of randomly selected smaller intruders and larger guarding males (i.e., random-sized method19). This species shows a large size advantage in male–male contests26, and smaller intruders are less likely to escalate fights23. This avoidance increased when smaller intruders encountered unfamiliar larger opponents with previous experience of wins, and avoidance was greatest when they re-encountered familiar opponents that had established dominance hierarchy28. Their own experience of defeat, however, did not affect the decision to give up in this species28. Thus, smaller intruders of P. middendorffii appear to use three types of information for their decision, but not their own defeats.

Pagurus minutus is another species in which the effects of information on the intruder’s decision to give up has been investigated. Pagurus minutus also shows a large size advantage and a lower probability of escalation by smaller intruders29,30. Yasuda et al.31 reported that P. minutus intruders with previous defeats had decreased eventual fighting success against unfamiliar naïve guarders. Although this suggests that individuals of this species collect information about experience of defeat, the effect of this information on the decision to give up is still unclear, because the study did not assess whether and when the losers gave up. More importantly, it used pairs of similar-sized males (i.e., self-selection method19) to clarify the effect of factors other than male size (i.e., female size in that study) on male–male contests. Hsu et al.19 recommend the random-sized method to examine the effect of experience, because the relationship between size and RHP is not perfect, even between similar-sized contestants. No study has tested whether P. minutus can use information related to opponent status and established hierarchy.

Here, we examined whether the decision of P. minutus intruders to give up is affected by four types of potentially available information related to RHP, namely (1) size relative to their opponent, (2) previous defeats, (3) opponent’s previous wins, and (4) established dominance hierarchy with the same opponent. We used randomly-selected males in male–male contests according to the suggestion of Hsu et al.19 (also see Yasuda et al.31).

Methods

Study animals

We collected precopulatory guarding pairs of P. minutus, each male with an intact major cheliped, from a sandy mud flat at Nunohiki, in the Waka River estuary, Wakayama, Japan (34°10′23″N, 135°10′49″E), from December 2015 to February 2016; the mating season of this species at this site occurs from November to April32. Each pair was placed in a small vinyl pouch filled with seawater collected in the field. In the laboratory, pairs in which the male was still guarding the female were separated, and each individual was kept in a container (8 cm × 12.5 cm × 8 cm) or a plastic cup (200 mL) with natural seawater (2.5 cm deep), to prevent copulation before the experiment. All pairs were acclimatized to laboratory conditions for at least 1 h before the experiment, and all tests were conducted within 6 h of collection.

After the experiments, all crabs were fixed by freezing (− 18°C) to allow us to measure them. The shield length (SL, calcified anterior portion of the cephalothorax, index of body size) of all males was then measured to the nearest 0.01 mm under a stereomicroscope. Since female size has no effect on random-sized male–male contests in this species29,30, we excluded this value.

Experimental design

We performed two sequential trials of male–male contests (Trials 1 and 2). In Trial 1, two guarding pairs were randomly assigned to an experimental set (N = 92 sets), and in each set the smaller male was designated as the intruder and the larger male as the guarder, owing to the large size advantage in this species29. We then placed a guarding male and his guarded female in a small plastic arena (19.5 cm × 11.0 cm × 8.5 cm) containing seawater about 3 cm deep. After the guarder had returned to guarding the female, the intruder was placed in the arena. We checked the outcome of Trial 1 at 15 min from when the intruder initiated movement; all intruders lost Trial 1 (i.e., did not guard a contested female).

Each intruder was then used again as an intruder in a second trial (Trial 2) after 1 h had elapsed. In Trial 2, we assigned the losers to three experimental groups with different types of guarders. In Group 1, losers encountered larger guarders that had not participated in Trial 1 (N = 31 sets). In Group 2, losers encountered guarders that won Trial 1 against a different intruder (N = 30 sets). In Group 3, losers encountered the same guarders as in Trial 1 (N = 31 sets). The difference in SL between losers and guarders did not differ significantly among groups (ANOVA, F2,89 = 1.030, P = 0.361). Other experimental methods in Trial 2 were the same as in Trial 1.

We recorded all trials using a digital camera (DMC-LF1, Panasonic) from the time the individuals were introduced into the arena until 15 min after the intruder began moving. When the intruder initiated grappling with the guarder (for details of this behavior, see23), we considered that the trial had escalated. After escalation, if intruders did not perform physical aggression for more than 3 min, we defined the fight as settled. We then recorded the duration (seconds) of the series of aggressive interactions as the contest duration until the intruder gave up and the eventual outcome on the basis of which male was guarding the female. Because the duration was defined as ending when the intruder gave up, we excluded contests in which the intruder won (Group 1, N = 1 set; Group 2, N = 1 set; Table 1) from the following analyses. If males continued grappling, with both males grabbing the shell of the contested female, at the end of Trial 2, the trial was defined as a draw (Group 1, N = 1 set; Table 1), and the duration until giving up the trial was censored. The final sample sizes for the analyses were 30 in Group 1, 29 in Group 2, and 31 in Group 3 (Table 1). No crabs were injured or lost any appendages during either trial.

Table 1 Experimental groups for two sequential trials of male–male contests in Pagurus minutus.
Full size table

Since the contest duration of Trial 1 had no effect on the decision to give up without escalation in Trial 2 in Group 1 (see Supplementary Fig. S1), we considered that the loser’s behavior in Trial 2 was independent of energy depletion from Trial 1.

Analyses

Data from Trial 1 were used to examine whether the intruder’s decision to give up was affected by the opponent’s relative size. For giving-up without escalation, a generalized linear model (GLM) with a binomial error distribution was used. This model was constructed by using whether intruders gave up without escalation (Yes = 1, No = 0; N = 90) as the response variable and the SL difference between intruders and guarders (DSLI–G) as the explanatory variable. Contest duration until giving-up was analyzed by Cox’s proportional hazard model33. The response variable in this model was contest duration (sec, N = 70), and the explanatory variable was DSLI–G.

We then used data from both trials in Group 1 to examine the effect of a defeat in Trial 1 on the decision in Trial 2. Since we observed all intruders twice, we used a generalized linear mixed model (GLMM) and Cox’s model with mixed effects to control for pseudo-replication. In the GLMM, the response variable was whether intruders gave up without escalation (Yes = 1, No = 0; N = 30 × 2 = 60), and the explanatory variables were (1) Trial (Trial 1 or 2) and (2) DSLI–G. In the mixed Cox’s model, the response variable was contest duration (sec; N = 23 + 14 = 37), and the explanatory variable was the same as in the GLMM (i.e., (1) and (2)). Intruder ID was treated as a random factor in both analyses.

The GLMM for giving-up frequency and the mixed Cox’s model for contest duration were also used to assess whether the intruder’s decision was affected by the opponent’s previous wins (Group 1 vs. Group 2; losers faced naïve opponents or opponents that won the previous contest) and established hierarchy with the same opponent (Group 2 vs. Group 3; losers faced a different or the same opponent that won the previous contest). In the GLMMs (Group 1 vs. 2, N = [30 × 2] + [29 × 2] = 118; Group 2 vs. 3, N = [29 × 2] + [31 × 2] = 120), the explanatory variables were (1) Trial (Trial 1 or 2), (2) Group (Group 1 vs. Group 2 or Group 2 vs. Group 3), and (3) DSLI–G. In the mixed Cox’s model (Group 1 vs. Group 2, N = 23 + 14 + 22 + 15 = 74; Group 2 vs. Group 3, N = 22 + 15 + 25 + 7 = 69), all three explanatory variables were included. To examine the effect of information, we also added a (4) Trial × Group interaction in each model if the interaction was significant. Intruder ID was treated as a random effect in all four models.

All analyses in this study were performed in R v. 4.1.134 software, and the R packages “glmmML”35 and “coxme”36 were used to conduct the GLMM and Cox’s model with mixed effects analyses. In Cox’s model, the proportional hazard assumption was satisfied for all explanatory variables (P > 0.062), except for a Trial × Group interaction in the model comparing Groups 2 and 3 (P = 0.034). However, since this model was not used in this study because the Trial × Group interaction was not significant (see “Results”), we concluded the proportional hazard assumption to be acceptable.

Results

Relative size

In Trial 1, the frequency of giving-up without escalation significantly increased as DSLI–G decreased (z = − 3.234, P = 0.001; Fig. 1a). After escalation, smaller intruders also showed significantly earlier giving-up (z = − 3.435, P < 0.001; Fig. 1b). Details of these and following analyses are shown in Supplementary Table S1.

Figure 1
figure 1

Relationship between male size difference and (a) intruders’ frequency of giving-up without escalation and (b) contest duration after escalation. DSLI–G indicates difference in shield length (index of body size) between intruders and guarders. Number in parentheses in (b) indicates sample size.

Full size image

Prior defeat

In Group 1, the frequency of giving-up without escalation was significantly higher in Trial 2 (after defeat) than in Trial 1 (before defeat) (z = 2.264, P = 0.024; Fig. 2a), but was not affected by DSLI–G (z = − 1.687, P = 0.092). Contest duration in this group was independent of both trial number and DSLI–G (Trial: z = − 0.560, P = 0.576; DSLI–G: z = − 1.302, P = 0.193; Fig. 2a).

Figure 2
figure 2

Relationship between trials and (a) the frequency of giving-up without escalation by intruders and (b) duration until giving-up by intruders after escalation. Number in parentheses in (a) indicates sample size in Group 1.

Full size image

Opponent’s prior wins

In the comparison between Groups 1 and 2, Trial × Group interaction was excluded from the GLMM because it was not significant (z = − 0.562, P = 0.574). In the GLMM without the interaction, giving-up without escalation was significantly affected by Trial and DSLI–G but not by Group (Trial: z = 2.925, P = 0.003; DSLI–G: z = − 2.512, P = 0.012; Group: z = − 0.412, P = 0.680). The frequency of intruders that did not escalate was higher in Trial 2 than in Trial 1 of both groups (Fig. 3a) and increased as DSLI–G decreased.

In the mixed Cox’s model, Trial × Group interaction was significant (z = 2.362, P = 0.018). While contest duration until giving-up was similar between trials in Group 1, it was lower in Trial 2 than in Trial 1 in Group 2 (Fig. 3b). However, although Trial 1 was the same for all groups, duration in Trial 1 seemed shorter in Group 1 than in Group 2 (Fig. 3b; see also Fig. 4b for the value of Group 3). The significant Trial × Group interaction might have been caused by this unexpected difference. Smaller intruders gave up significantly sooner after escalation (z = − 2.592, P = 0.010).

Figure 3
figure 3

Differences between Groups 1 and 2 in (a) the frequency of giving-up without escalation by intruders and (b) duration until giving-up by intruders after escalation. Numbers in parentheses in (a) indicate sample size. Group 1: naïve opponent; Group 2: opponent that won against a different intruder.

Full size image

Established dominance hierarchy

In the comparison between Groups 2 and 3, the Trial × Group interaction was significant in the GLMM (z = 2.182, P = 0.029). Although the frequency of giving-up without escalation increased in Trial 2 in both groups, the intensity of the trend was greater in Group 3 than in Group 2 (Fig. 4a). DSLI–G also had a significant effect on the decision (z = − 2.897, P = 0.004).

Figure 4
figure 4

Differences between Groups 2 and 3 in (a) the frequency of giving-up without escalation by intruders and (b) duration until giving-up by intruders after escalation. Numbers in parentheses in (a) indicate sample size. Group 2: opponent that won against a different intruder; Group 3: same opponent as in Trial 1.

Full size image

In the mixed Cox’s model, we excluded the Trial × Group interaction from the model because it was not significant (z = − 0.374, P = 0.710). In the mixed Cox’s model without the interaction, intruders gave up significantly earlier in Trial 2 than in Trial 1 (z = 3.411, P < 0.001; Fig. 4b) and with decreasing DSLI–G (z = − 2.826, P = 0.005). No difference was found between groups (z = 0.037, P = 0.970).

Discussion

We conducted random-sized male–male contests of the hermit crab P. minutus to examine whether smaller solitary intruders give up contests earlier on the basis of four types of information: size relative to a larger guarding opponent, previous defeats, opponent’s previous wins, and the established dominance hierarchy. We found that three factors increased the frequency of giving-up without escalation: smaller size (Trial 1; Fig. 1a), a previous defeat (Trial 1 vs. Trial 2 in Group 1; Fig. 2a), and re-encountering the same dominant opponent (Group 2 vs. Group 3 in Trial 2; Fig. 4a). The opponent’s previous wins, on the other hand, appeared to contribute to a shorter duration after escalation (Group 1 vs. Group 2 in Trial 2; Fig. 3b). Our previous studies have also shown that the motivation of P. minutus intruders to fight is decreased when they encounter a larger opponent in the randomly-chosen contests29,30 and had experienced a recent defeat in the similar-sized contests31. This is the first study to show that Pagurus hermit crab intruders can use all four types of information to decide whether to give up male–male contests and to confirm that the intruders use information about their recent defeats regardless of the experimental method.

The decreased aggression in losers is referred to as the loser effect19,20. Many animals show a loser effect when competing against naïve opponents14,37,38,39, and one explanation of this is that losers decrease their self-assessment of RHP relative to others18,20. Although Yasuda and Koga30 suggest that P. minutus assess their RHP relative to their opponent before and after escalation in male–male contests, the greater giving-up frequency in Trial 2 of Group 1 indicates that they also assess their own RHP before escalation, at least after losing. Since intruders’ persistence after escalation did not change even after losing, the loser effect might be based on a perceived but not actual RHP19. Although fighting behavior, such as rapping in shell fights of hermit crabs40, can carry a cost of depletion of energy reserves41, contest duration in Trial 1 did not affect giving-up decision in Trial 2 of Group 1 (Supplementary Fig. S1), and few Pagurus males are injured during male–male contests22,23. Thus, one contest might not affect actual RHP, via depletion of energy reserves, in our context.

Whereas intruders of P. middendorffii ignore a past defeat in their next contest28, intruders of P. minutus used this information to decide whether to avoid a contest. Yasuda and Koga30 have pointed out the possibility that P. minutus is more sensitive to fighting costs than P. middendorffii; they suggest that P. minutus may have more potential mating opportunities than P. middendorffii30, because P. minutus has a longer reproductive period than P. middendorffii and only P. minutus has multiple oviposition32,42,43. Animals showing the loser effect avoid even smaller opponents38,44, suggesting that it might be so important for losers to minimize fighting costs that they would not engage even in a fight with a high probability of winning. The loser effect in P. minutus might also contribute to cost avoidance and support cost sensitivity in this species.

If P. minutus losers maintain their actual RHP after a contest, our analysis (Group 1 vs. Group 2) suggests that intruders of this species can gain information about an opponent’s previous wins after escalation. Because we found no difference in giving-up frequency before escalation in these groups (Fig. 3a), a loser’s decision to give up might be independent of the opponent’s prior wins in the pre-escalation phase. After escalation, however, a significant Trial × Group interaction was detected, and intruders retreated sooner from prior winners than from naïve opponents (Fig. 3b). One possible explanation is that losers could detect an opponent’s cues of a prior win, as can other crustaceans45 even after escalation, but why P. minutus use this information only after escalation, rather than before escalation like P. middendorffii28, remains to be explored. Another possible explanation is that the opponent’s actual RHP was increased by a prior win via improved motivation19 or fighting skill46. On the other hand, since contest duration in Trial 1 was shorter for Group 1 than for the other two groups, this interaction might also be caused by the unexpected value. To determine the relative importance of an opponent’s cues indicating a prior win and an opponent’s enhanced RHP for losers’ decisions, further direct investigation (i.e., random-sized contests between naïve intruders and guarding males with a prior win) is needed.

Greater giving-up frequency in Trial 2 of Group 3 than of Group 2 (Fig. 4a) suggests that P. minutus intruders used the previously established dominance hierarchy to avoid familiar, dominant opponents. Pagurus minutus could therefore be the second species of Pagurus hermit crabs that shows familiar recognition in male–male contests, after P. middendorffi28. As above, interspecific differences in the loser effect suggest that males of the two species might have different sensitivities to the cost of fighting for a potential mating opportunity. Nevertheless, since intruders of both species avoid escalation against familiar dominant individuals, the established hierarchy should provide reliable information about subsequent decreased chance of success. Since opponent recognition has also been reported in other contexts in Pagurus species (P. bernhardus47 and P. longicarpus48), this cognitive ability might be widely shared in this group.

Recent studies indicate that information use is flexible within and among species. For example, contestants rely on different information about RHP based on contest phases23,49,50, their own positioning51, and prior contest experience17. In P. minutus, intruders use mainly mutual-assessment of RHP to make a giving-up decision but also re-assess their own RHP after losing; information about an opponent’s prior wins might affect different contest phases based on the familiarity of the opponent. Pagurus minutus therefore provides an additional example of flexibility in information use. Information use can also vary among closely related species according to interspecific differences in sociality52, habitat choice and response to predation53, and reproductive characteristics30,54. Social cognition is an active field of study but investigations of invertebrates are limited55; therefore, investigations in taxonomically diverse organisms are required.