Conspecifics can provide social cues about the presence of key features of the surrounding environment, such as food or predators. Attending to social cues may therefore potentially benefit receivers, or at least be worth following. Yet, bearing social cues could also be costly, particularly if it increases the likelihood of close-range interaction with non-kin. Here, we experimentally-seeded social cues in the wild onto focal individuals of the social hermit crab (Coenobita compressus), testing (1) the ‘potential benefits to receivers’ hypothesis, which predicts that receivers will follow social cues to orient towards valuable resources, and (2) the ‘costs to bearers’ hypothesis, which predicts that bearers of social cues will experience direct (physical) costs or indirect (constrained movement) costs due to interaction with receivers. Consistent with hypothesis (1), in natural encounters, conspecifics that crossed paths frequently made antennal contact, potentially gathering social information at close range. In experiments, naïve conspecifics followed focal individuals bearing ‘positive’ social cues (about a valuable food resource) significantly more often than they followed individuals bearing less attractive (‘neutral’ or ‘ambivalent’) social cues, pointing to a potential benefit. Consistent with hypothesis (2), individuals bearing positive social cues incurred greater direct and indirect costs, being physically flipped more often and achieving shorter displacements compared to individuals bearing other social cues. We conclude that experimentally-seeded social cues in the wild can confer costs to bearers and potentially benefit receivers. Broadly, the costs of bearing social cues, revealed here, underscore the importance of not overlooking that social cues may be costly.

Study site

We conducted extended preliminary observations on highly social terrestrial hermit crabs (Coenobita compressus) in the wild from 2018 to 2022, using these observations as a foundation for the present study. The present study was carried out between January and March 2022 at the beach-forest interface of our long-term study site (Laidre 2010) in Osa Peninsula, Costa Rica (8°23'55.6"N 83°20'52.8"W), where this species occurs naturally in abundance and displays high social activity levels. Our study had three different parts (see below): (1) antennal contact during natural encounters, (2) overall attraction to chemical stimuli in dishes, and (3) experimentally-seeded social cues. All data, both observational and experimental, were collected by the first author between 0500 and 1600.

Antennal contact during natural encounters

To quantify how frequently individuals wandering the beach gathered chemo-tactile social cues via antennal contact we carried out focal and behavioral sampling, distributed evenly across ten separate days. For focal sampling, the observer stood stationary and used binoculars from > 3 m distance to observe specific, randomly-selected focal individuals on the surrounding beach. Each focal individual (N = 20 total individuals) was observed for 5 min. For behavioral sampling, a parallel methodology was used, with each observation period lasting 5 min (N = 20 behavioral samples, for a total of 100 min of cumulative sampling). In behavioral sampling, instead of a specific focal individual being chosen, the observer scanned the entire surrounding beach through binoculars, waiting for any instances of encounters between pairs of individuals.

An ‘encounter’ (Figure 1a) occurred, in both behavioral and focal sampling, when: two individuals, originating more than three body lengths apart, came within a body length of one another, with one or both individuals approaching the other head on. The two individuals that encountered one another therefore had equal opportunity to make ‘antennal contact’ (Figure 1a), which was defined as follows: one or both of an individual’s two antennae touched any part of the other individual (body or shell), at any time during the encounter, regardless of who initiated contact. Every encounter thus had three distinct outcomes: antennal contact could be made by (1) just one individual in the pair, (2) by both individuals, or (3) by neither individual. Note: any subsequent reference to ‘antennal contact’ generally (without specification of whether it was made by just ‘one’ or by ‘both’ individuals in a pair) simply denotes these two separate outcomes being lumped.

Overall attraction to chemical stimuli in dishes

Experimental setup

To determine levels of attraction to different chemical stimuli, we placed each chemical stimulus alone in a dish in the field, recording video from above the dish as free-roaming individuals in the wild moved to and fro. Three different types of chemicals (see below) were tested in randomized order each day across twenty days (with N = 20 trials of each of the three chemical conditions, for a total of N = 60 experimental trials). Each trial involved placing out only one shallow plastic circular dish (5 cm in diameter) and testing only one type of chemical at a time. In between trials the dish was retrieved, and a separate dish was deployed at a new, randomly designated location on the beach. The experimental area consisted of a 65 cm diameter circle drawn in the sand immediately around the dish (Figure 1b). To remove any barrier to access by crabs, the dish was pressed into the sand, with its sides and upper rim flush with the substrate.

To video overhead during the experiments, a video camera (Canon Vixia HF R72) was attached to a long wooden pole driven into the sand. After the experimental area and video camera were setup, the entire experimental area was filmed continuously for twenty minutes (from t = -10 min to t = +10 min). The first ten minute period (from t = -10 min to t = 0 min) provided a baseline with an empty dish; then at t = 0 min the experimenter used a pipette to add 9 mL of chemical to the dish; after which the trial continued for another ten minutes with the filled dish (from t = 0 min to t = +10 min). A visual count of the number of individuals present in the experimental area was made just before the addition of the chemical stimulus at t = 0 min (‘start’ count) and once again at t = +10 min (‘end’ count). From the videos we also quantified the latency for the first individual to make contact with the dish after the chemical stimulus had been added.

Three conditions: ecologically-relevant chemical stimuli

Three ecologically-relevant chemical stimuli, all natural elements of the species’ environment, were chosen for their biological significance and were put into dishes in the field (see above) and also paired with conspecifics (see below): (1) coconut milk signifies a valuable food resource (Laidre 2013; Steele and Laidre 2019), therefore representing a ‘positive’ social cue when paired with a conspecific (for consistency across all our experiments, we used the same brand of coconut milk: ‘AROY-D’ San Jose, Costa Rica, for which the sole ingredient was coconut milk); (2) water from the adjacent lagoon that individuals regularly pass through (Laidre 2010, 2013) served as a control, therefore representing a ‘neutral’ social cue when paired with a conspecific; and (3) urine signifies both a potential predation risk (Laidre et al. 2012) and a potential shell opportunity (Valdes and Laidre 2019), therefore representing an ‘ambivalent’ social cue when paired with a conspecific. Note that nonhuman mammals (e.g., white-nosed coatis Nasua narica and raccoons Procyon spp.) can occasionally act as predators on our study system, but this is rare because generally their bite forces are below that required to break the crabs’ shells (Laidre et al. 2012). As a consequence, most nonhuman mammal predators are less dangerous than human predators (fishermen), which have the dexterity to directly pull crabs out of their shells, leaving behind the empty shells and using the crab body as fish bait. We therefore used human urine as an ecologically-relevant and readily collectable chemical that signifies both a predation risk and shell opportunity in this system. To collect urine, we used voluntary human donors at the field site (including the authors), all of whom consumed omnivorous diets. Urine was collected in the same receptacle at the same time each day. It is worth emphasizing that unambiguously ‘negative’ social cues are rare to non-existent in this species (cf. Doherty and Laidre 2020 on threat display), otherwise we would have included a fourth ‘negative’ chemical condition.

Experimentally-seeded social cues: focal individuals doused in chemicals

Experimental setup

To experimentally seed social cues, the above three chemical stimuli (coconut milk, water, and urine) were paired with conspecifics, thereby generating positive, neutral, and ambivalent social cues, respectively. Each of these three social cue conditions was tested, in randomized order, each day across twenty days (with N = 20 trials per condition, for a total of N = 60 experimental trials). For every trial, a randomly selected (new) focal individual was doused with one of the chemicals. After applying the chemical stimulus to the focal individual, thereby making it a bearer of social cues, we then released the focal individual at a randomly designated location on the beach and recorded all its subsequent interactions with conspecifics from a distance for 5 min using a handheld video camera (Canon Vixia HF R72). To test the impact of chemical-based social cues on bearer-receiver interactions, we quantified from video recordings: the number of encounters each focal individual had with conspecifics and all ensuing interactions (Figure 1c), which could either involve the first step toward a potential benefit to receivers (i.e., receivers followed the bearer) or which could demonstrate a realized cost to the bearer (i.e., the bearer was flipped by receivers; see Table S1 for definitions of each behavior).

Only encounters the focal individual had with moving conspecifics were counted to clearly distinguish conspecifics from stationary shell-like objects (pebbles) in the background of the video. All videos were coded by the first author and by an independent observer, who was blind to both the conditions and the hypotheses. To measure inter-observer reliability for the number of encounters, follows, and flips (see Figure 1 and Table S1), a random sample of videos (N = 30 total, including N = 10 of the experimental trials from each of the three conditions) were coded. There was strong inter-observer reliability (r² > 0.75) for all behaviors (encounters: F_1,28 = 94.2, p < 0.0001, r² = 0.77; follows: F_1,28 = 88.53, p < 0.0001, r² = 0.76; flips: F_1,28 = 705.83, p < 0.0001, r² = 0.96).

Focal individual selection

For each day of experiments, three focal individuals, among usually thousands of social hermit crabs wandering the beach, were randomly selected and then allocated randomly to each of the three chemical conditions. A crab was only included as a focal individual if it had all appendages intact and was occupying the most common (Nerita spp.) shell. Across the three conditions, there was no difference in shell diameter of focal individuals (one-way ANOVA: F_2,57 = 0.04, p = 0.96; positive condition: 11.48 ± 0.33 mm; neutral condition: 11.33 ± 0.45 mm; ambivalent condition: 11.45 ± 0.47 mm).

Application of chemicals onto focal individual

Before the start of each trial, the experimenter’s hands and a set of forceps were thoroughly washed, and dried, removing any potentially confounding chemical cues. Also, to enable identification of focal individuals in the video, the dorsal side of each crab’s shell was marked with a small square of highly visible duct tape, which was removed at the end of the trial. After its shell was marked, the chemical stimulus was seeded onto the focal individual, by placing the individual (while it continuously remained occupying its original shell) into a temporary holding container and then using a pipette to douse the individual with 9 mL of the chemical (Figure 1c). Dousing ensured that the entire external surface of the crab’s body and shell were covered with the chemical cue, something which occurs commonly and naturally when crabs’ forage on many foods, including coconut (Laidre 2013). The focal individual was then removed from the temporary holding container using forceps and held in the air while any excess chemical ceased dripping, leaving just a thin residue on its exterior surface for close-range detection by receivers.

The focal individual was then carried with clean forceps to a randomized drop point on the beach, where it was placed on top of the sand, with its start position marked by the insertion of a vertical wooden dowel. The number of conspecifics within a 30 cm radius of this start position was recorded, then the focal crab was videoed continuously for 5 min. After 5 min, the focal individual’s end position was marked by the insertion of another vertical wooden dowel, with the number of conspecifics within a 30 cm radius of this end point being recorded, and then the focal individual was recollected. Upon recollection, the mark on the focal individual’s shell was removed, and the individual was rinsed with water before being placed in a separate holding container until the end of the day’s experiments were completed. After all experiments were finished for the day, all focal individuals were released back into the wild. To test if there was a difference in how far focal individuals moved across the different social cue conditions, we measured the distance between their start and end positions (i.e., their displacement; Figure 1c). Displacement is a strong indicator of the extent to which individuals can explore their environment (Trinh and Laidre 2016; Doherty and Laidre 2022), with shorter displacements providing less opportunity for discovering new resources (Laidre 2010; Steele and Laidre 2023a).

Statistical analyses

Antennal contact during natural encounters

We tested whether natural encounters were differentially likely to lead to antennal contact by both individuals, just one individual, or neither individual. For focal sampling, we tested differences in this likelihood of antennal contact by comparing the number of encounters with each of the three different outcomes (note: no formal statistical test was performed, since these three outcomes were non-independent, all deriving from the same N= 20 focal individuals). For behavioral sampling, in which the three outcomes were independent, we tested differences in the likelihood of antennal contact by performing a Chi-square test, which compared observed occurrences of each of the three outcomes to the null assumption of an equal likelihood of these three possible outcomes.

Overall attraction to chemical stimuli in dishes

To test for differences in latency to contact dishes across the three chemical conditions, we performed a one-way ANOVA. To test if the positive (coconut milk) condition attracted individuals faster than the neutral (water) or ambivalent (urine) conditions, we conducted post-hoc All-Pairs Tukey-Kramer tests, which simultaneously allowed us to explore any differences between the neutral and ambivalent conditions. Data for latency were square route transformed (which made these data normally distributed) and analyses were then conducted on the transformed data.

To compare differences in the number of individuals present in the experimental area from before-to-after the addition of each chemical, we performed a Kruskal-Wallace test. To test if the positive condition attracted more individuals than the neutral or ambivalent conditions, we conducted post-hoc All-Pairs Dunn tests, which simultaneously allowed us to explore any differences between the neutral and ambivalent conditions.

Experimentally-seeded social cues: focal individuals doused in chemicals

To test for differences across the three social cue conditions, we used Kruskal-Wallace tests to compare the number of encounters focal individuals had with conspecifics and the number of conspecifics surrounding the focal individual at the end of each experiment. For any variables that showed a significant difference across conditions, we then used post-hoc All-Pairs Dunn tests to determine if the positive condition differed from the neutral or ambivalent conditions and if any difference existed between the neutral and ambivalent conditions. To test differences across the three social cue conditions in the proportion of encounters that led to receivers following the focal individual (the bearer), we arcsine transformed these data (which made these data normally distributed) and performed a one-way ANOVA. A post-hoc All-Pairs Tukey-Kramer test was then conducted to determine if individuals in the positive condition followed bearers more than individuals in the neutral or ambivalent conditions, and if any difference existed between the neutral and ambivalent conditions. To test differences across the three social cue conditions in how many times focal individuals were flipped by receivers, we used a generalized linear model (GLM), with a Poisson distribution and a Log link function. Finally, to test differences across the three social cue conditions in displacement achieved by focal individuals, we performed a one-way ANOVA. A post-hoc All-Pairs Tukey-Kramer test was then conducted to determine if individuals in the positive condition achieved shorter displacements than individuals in the neutral or ambivalent conditions, and if any difference existed between the neutral and ambivalent conditions.

Analyses were performed using JMP® Pro version 16.0.0. For post-hoc analyses, All-Pairs Tukey-Kramer tests controlled for the overall error rate, and All-Pairs Dunn tests used the Bonferroni adjustment. All post-hoc tests are thus reported as above or below p = 0.05.