Individualism versus collective movement during travel

Study site

Social hermit crabs (Coenobita compressus) were studied in Osa Peninsula, Costa Rica, at a long-term field site (Osa Conservation’s Piro Biological Station), where the population has been under study since 2008¹⁷. Experiments were carried out from January to March 2019 at the beach-forest interface (Fig. 1A), an area where ‘fission–fusion’ social groupings³⁰ continuously form and dissolve³¹ and where free-roaming individuals regularly travel¹⁷. All studies were undertaken during daylight hours (06:30–11:30 h) during periods of peak social activity.

We conducted two separate sets of experiments, both involving a similar stimulus design (below). First, to determine whether free-roaming individuals were biased in their movement decisions by a collective, we performed a set of free-roam experiments (see “Experiment 1: Free-roam”). The free-roam experiments were conducted directly on the beach (Fig. 1B; 8° 23′ 39.5″ N, 83° 20′ 10.2″ W). Second, to determine whether an increase in danger influenced the relative independence versus social bias in individual movement, we performed a set of handled experiments (see “Experiment 2: Handled”). The handled experiments were conducted on a platform (Fig. 1C; 8° 23′ 33.2″ N, 83° 19′ 50.6″ W), which was immediately adjacent to the beach and situated within the range of the crabs’ normal daily movements. All reported compass bearings are relative to magnetic North (0°) unless otherwise specified.

Stimulus design

As conspecific ‘stand-ins’, we used N = 60 Nerita scabricosta shells (C. compressus’ preferred shell species²³), spanning a natural range of sizes (9–32 mm) within this population (Table S1; Fig. S1). To create a group of these stand-ins that we could manoeuvre as a collective, each shell was affixed using epoxy to one of four strands of clear fishing line, which were each 4 m long. These lines were spaced approximately 30 cm apart on a long wooden dowel (Figs. 2A,B, 3A,B). An equal number of shells (N = 15 shells per line) were distributed randomly along the 2 m of each fishing line furthest from the dowel. To allow the experimenter to manoeuvre the stimuli, without disturbing live crabs’ behaviour, another fishing line (4 m in length) was attached to the top of the dowel. With this line, the entire apparatus could be pulled by the experimenter from a distance, thereby simulating synchronised movement of the entire collective. To control for any influence the apparatus might have on focal individuals (other than that produced by the movement of the shell ‘stand-ins’), the entire apparatus—dowels and fishing lines—was replicated, just without any attached shells, for use as a control (Figs. 2C, 3C).

Experiment 1: Free-roam

To test whether the movement of the collective influenced free-roaming individuals’ travel direction, the stimuli were pulled across the beach at a uniform speed (1 m per min), within the natural range of the walking speed of social hermit crabs^17,22,23. Each trial lasted 1 min. A total of N = 80 free-roam trials were conducted, N = 40 experimental (with the full collective, represented by all the shells) and N = 40 controls (with only the raw materials, but no shell collective). For each of the N = 80 trials, the movement of a single free-roaming focal individual was recorded.

It is not uncommon to see multiple crabs moving parallel to (or perpendicular to) the shore, since many individuals will often be collectively attracted to eviction sites, injured conspecifics, or food items, with all the attracted individuals travelling in a roughly parallel formation^16,17. For each trial in the free-roam experiments, the stimuli were pulled parallel to the shore (Fig. 1B), either to the right (116.1°) or to the left (296.1°). We did not pull the stimuli perpendicular to the shore, given the substantial slope from the forest down to the ocean, which would have confounded any such comparisons. Condition (experimental or control) and stimulus direction (right or left) were selected randomly, with balanced sample sizes (N = 20 for each). To ensure there was a free-roaming focal individual, whose movement we could measure in response to the stimulus, a trial was only carried out when at least one live crab was walking within approximately 30 cm of the stationary stimulus. Then pulling was initiated.

To avoid disturbing live individuals by moving through or near the vicinity, we gathered overhead video footage of all experiments using a drone (Phantom advanced model GL300C). Drone video recorded all interactions between the focal individual and the simulated collective while the drone hovered at a height of approximately 2 m above the beach. At this height, there was no disturbance to natural behaviour or movement of the crabs, and the drone remained positioned overhead for at least 1 min prior to the start of a trial. Minor adjustments to position were then made between trials due to drone drift (i.e., slight movement of the drone due to wind).

To randomly select focal individuals for video coding, we first split an image of the starting frame of each video file into a 4 × 4 matrix, with N = 16 equally-sized sections, and then used a random number generator to choose one section (repeating this step if no crabs were present in the selected section). Second, we numbered all individuals in the selected section and again used a random number generator to select the individual.

To calculate bearings relative to magnetic North for the direction each focal crab moved, we first measured the angle of divergence (°) between the stimulus trajectory and the focal crabs’ trajectory. Focal crab trajectory—a proxy for the overall direction of the crab’s movement—was measured by drawing a straight line from the start-to-end position of that individual (see Fig. S2 and Vid. S1 for further explanation). Stimulus trajectory was measured in the same manner, using the shell closest to the focal at the beginning of the trial. Using Google Maps and the IGIS Map bearing angle calculator, we calculated the bearing of our stimuli (right and left) relative to true North (right: 114°, left: 294°). To determine bearings for our stimuli relative to magnetic North, we then used the Enhanced Magnetic Model (EMM) magnetic field calculator, provided by NOAA, to calculate the relevant declination (− 2.1°) for our coordinates on the dates the experiments were carried out, subtracting this value from true North. Thus, for the free-roam experiments, the bearing of a stimulus moving to the right, relative to magnetic North, was 116.1°, and the bearing of a stimulus moving to the left, relative to magnetic North, was 296.1°. Lastly, bearings for focal crabs’ directions, relative to magnetic North, could then be calculated using the new bearings of the stimuli and the angle of divergence between stimulus and crab trajectories.

To gauge the level of interaction that focal individuals had with the collective, we recorded whether or not individuals initiated contact with shells in the experimental condition. An individual was classed as having initiated contact if it climbed onto a shell or touched a shell with its claws (Vid. S2). Additionally, we noted whether individuals were bumped by passing shells. An individual was classed as having been bumped if a moving shell hit it while the individual was withdrawn, stationary, or facing away from the moving shell (Vid. S3).

To assess whether drone drift during experiments was a problem, we examined a random sample (N = 20) of the videos, both control (N = 10) and experimental (N = 10). We took N = 40 images from these 20 videos (i.e., two images from each video: one at the start of the 1-min trial and one at the end of the 1-min trial) and used a system wherein we marked the same two distinguishable fixed points on the landscape in each pair of images. We then overlaid the images in each pair, allowing us to see any longitudinal or latitudinal movement as well as any potential rotation of the drone. Nineteen of the N = 20 pairs of images showed virtually identical overlap of the markers, with just one image showing a minor gap between 1 of the 2 landmarks, suggesting slight rotation of the drone. We were therefore confident that drone drift was not an issue in our analyses.

All videos were coded by CD. To measure inter-observer reliability for the angle of divergence (°) between stimulus trajectory and focal crabs’ trajectory (see Fig. S2), a random sample of videos (N = 41 total, N = 22 of experimental and N = 19 of control) were also coded by a second observer (MP) who was naïve to the competing hypotheses. There was strong inter-observer reliability in the measurements (F_1,39 = 142.8, p < 0.0001; r² = 0.79). Indeed, excluding a single outlier, the r² value was 0.995 (F_1,38 = 7233.6, p < 0.0001). And the vast majority (N = 35) of the angles measured by both observers fell within 10° of each other.

Experiment 2: Handled

To investigate whether danger levels may mediate the impact a collective has on individual movement, we ran another set of experiments, in which focal crabs were handled prior to testing. Unlike the free-roam experiments, where individuals only interacted with conspecifics in the wild, in these handled experiments, individuals were picked up by the experimenter—a strong negative stimulus—immediately before being tested. Furthermore, we carried out the handled experiments on an artificial beach (Fig. 1A,C), involving a flat platform, which eliminated the slope of the natural beach, enabled us to precisely measure each focal individual’s displacement (below), and ensured no other free-roaming individuals were present besides the single focal individual. The artificial beach consisted of a 4 × 4 m tarpaulin, topped with a layer of natural sand collected from the adjacent beach. The artificial beach thus afforded a high level of control, while still involving semi-naturalistic field conditions. The same experimental and control stimuli (see section on “Stimulus design”) were used to test focal individuals’ responses in both the free-roam and the handled experiments.

Individuals in the handled experiments were collected from the wild, on the beach adjacent to the platform, shortly before the start of the experiment. A focal individual was then placed under an opaque cup in the centre of the stimulus (Fig. 3), where it remained for 1 min before being released. This 1-min buffer allowed the experimenters to leave the vicinity and get in position to manoeuvre the stimulus. The cup containing the focal individual had fishing line attached and was removed via a pulley system. At the same time, the stimulus, either experimental (N = 80 trials) or control (N = 80 trials), was pulled at a speed of 1 m per min for 1 min. Both the stimulus type (control or experimental) and direction (forest = 27°, ocean = 207°, left = 297° or right = 117°) were randomly selected prior to the start of the trial. The handled experiments were not videoed, since measurements could be directly taken in situ. At the end of each 1-min trial, the compass bearing was taken of the focal individual, based on a straight line from its start-to-end position. Also, to test whether the simulated collective affected the focal individual’s travel distance, we measured the focal individual’s displacement (cm) as the same straight line from its start-to-end position. Note that degrees for left and right are slightly different between the handled versus the free-roam experiments. Left and right were defined as parallel to the shoreline, which differed marginally between the two experimental sites (Fig. 1A).

Statistical analyses

To assess variability in direction of focal individuals, we calculated circular variance for each condition (control versus experimental) and analysed data separately for each stimulus direction. Circular variance ranges from 0 to 1 (with 0 meaning no variance, i.e., all individuals go in exactly the same direction, and with 1 meaning maximum dispersion in all directions, such that a mean angle cannot be described). We considered the level of variation in individual direction to be indicative of bias, with less variation signifying stronger bias. Hence, if little or no bias occurred due to the collective, then variation in individual direction should remain high across all conditions and stimulus directions.

To test for directed orientation (i.e., whether a true mean or median direction existed) within each condition, we used the Rayleigh test for any conditions that had a von Mises distribution (the equivalent of a normal distribution for circular data). For conditions with a distribution other than von Mises, we used the Hodges-Ajne test (hereafter referred to as an omnibus test). Significant p-values for the Raleigh or omnibus tests, respectively, indicate that a true mean or median exists³². Data for these tests were analysed separately for each different stimulus direction.

To test for differences in displacement (i.e., the absolute distance individuals moved during the trials), we ran an ANOVA model, which included the following factors: condition (with two categories: control and experimental); stimulus direction (with four categories: right, left, forest, and ocean); and the interaction between condition and stimulus direction. We used an orthogonal contrast test to specifically examine the impact of condition (i.e., control versus experimental) on displacement.

All circular statistics were calculated in R version 1.3.1056, with the exception of the omnibus tests, which were carried out in MATLAB R2020a. All analyses of displacement and inter-observer reliability were performed in JMP® Pro 15.0.0.

Ethics approval and consent to participate

All experiments were approved by the Costa Rican Ministerio de Ambiente y Energía (MINAE).

Source: Ecology - nature.com