Collective movement increases initial accuracy and path efficiency in talitrid amphipod orientation

Abstract

Talitrid amphipods are an extensively studied system for navigation due to their robust ability to navigate back to the optimal burrowing zone after foraging and could be a model system in which to study the impacts of collective behaviour on short-distance navigation and orientation. We investigated whether talitrid amphipods (Megalorchestia pugettensis) differ in their orientation abilities when released individually versus in a group. When released individually, the amphipods took longer to start moving (p < 0.001), travelled longer paths (p = 0.003), moved faster (p = 0.016), had a different initial bearing (p = 0.003) and exhibited more spread in their initial bearing (p = 0.009) than when released in groups. There was no difference between individuals and groups in terms of their trial time nor in the direction or spread of their final orientation. This study introduces a tractable, invertebrate species in which to study the impacts of collective movement and reveals previously unexamined differences in orientation abilities for talitrid amphipods released independently versus in a group that have implications for experimental design in this system.

1. Introduction

Talitrid amphipods are an extensively studied model system for orientation [1]. The many marine species of this family inhabit beaches worldwide and use a variety of cues such as landscape features, the sun and moon, slope, salinity and the geomagnetic field to orient [2–8]. Talitrid amphipod burrows are generally located in the ‘wrack line,’ a strip along the beach where debris is deposited at high tide. However, they must leave this optimal burrowing zone to forage and thus are able to orient quickly and accurately to avoid predation or desiccation [9]. The robustness of this orientation has allowed researchers to extensively study talitrid orientation for over 70 years (reviewed in [1]).

When conducting orientation experiments, researchers have used various numbers of individuals per trial. Studies have released 1–25 animals at a time, with 10 being the most common number of animals per trial [10–15]. The use of multiple animals per trial increases sample size without substantially increasing the total experiment time. This assumes that each individual’s behaviour is not influenced by conspecifics and can be treated as independent. However, recent research on collective behaviour has shown that orientation abilities can be altered and often improved in groups, a feature referred to as ‘collective navigation’ [16,17].

Studies of collective behaviour in Talitridae are limited, with only one study finding no evidence for differential orientation between individuals released alone versus individuals released in groups [18]. However, their analyses were limited by the methods that were typically available at the time; most notably, the authors only examined the location of the animals at each trial’s end. More recent tracking and analytical techniques can record an animal’s entire path and reveal previously undetected differences between individual and group movement. For example, in pigeons, studies that used only the ‘vanishing bearings’—defined as the bearing of the pigeons as they vanish from the researcher’s sight—were unable to determine whether solo individuals travelled differently than when in flocks [19–22]. However, recent experiments using Global Positioning System trackers to record a bird’s entire flight path have shown that, in comparison to solo flights, individuals in small flocks exhibit higher speeds, straighter flight paths and decreased rest time [23,24].

Here we aim to build upon the work done by Scapini et al. [18] by studying the impacts of nearby conspecifics on the orientation behaviour of Megalorchestia, a genus of littoral talitrid amphipods native to the eastern North Pacific. The aims of this study are to (i) investigate the ability of Megalorchestia pugettensis to orient relative to the wrack line and (ii) test whether individuals behave differently when alone versus in the presence of congeners. We predicted that, if there was evidence of collective navigation, the individuals in a group would travel more efficiently (characterized by shorter paths and trial times) and have more consistent orientation (characterized by headings with less spread and more closely oriented in the direction of the shortest distance to the wrack line) when compared to individuals released alone [17]. To test these hypotheses, we conducted experiments where animals were released either alone or in groups and used full-path tracking analyses to examine eight different orientation metrics: latency to move, speed, distance travelled, time to reach the wall, the direction and spread of their initial bearing and the direction and spread of their final orientation.

2. Methods

Animals were released in experimental arenas either alone (hereafter ‘GS1’ or ‘individuals’), or in groups of size five (‘GS5’) or 10 (‘GS10’). The arenas consisted of circular, clear plastic dishes (27 cm diameter), surrounded by white, opaque paper to obscure any landscape cues along the horizon (figure 1). At the beginning of each trial, the arenas were levelled and rotated randomly relative to the beach. Cameras underneath the arenas filmed the animals’ movement through the bottom of the dishes (electronic supplementary material, figure S1). These cameras were always placed in the same position at the centres of the arenas and in the same orientation relative to the arenas to identify whether the sandhoppers were orientating to any features of the experimental set-up itself. Before the animals were introduced to the arenas, the direction of the shortest distance to the wrack line was indicated to each camera, so that the animals’ movement relative to the beach could be calculated (see electronic supplementary material for more details about the experimental conditions).

Figure 1.

(a) Example of the location of arenas (circled in red) relative to the wrack line. (b) Detailed picture of experimental animal. Inset shows a male’s subchelate second gnathopod (circled in red). The average size for males of this species is 13–16 mm [25].

The arenas were placed 2 m landward of the wrack line to motivate the amphipods to move seawards. On macrotidal beaches, such as the study site, talitrid amphipods rarely travel above the wrack line [1]—this zone has the same risks of predation and desiccation as below the wrack line but with far fewer food sources. The amphipods would likely only enter this area through passive displacement, such as a storm, or a navigational mistake. Therefore, talitrid amphipods who find themselves landward of the wrack line are strongly motivated to return to the optimal burrowing zone [1].

Before each trial began, the animals were allowed to acclimate to the arena for 2 min in a transparent plastic cup (6.5 cm diameter) placed at the centre of the arena. After that time had elapsed, the cup was lifted from a randomly chosen direction, and then the arena was covered with a translucent lid, also from a randomly chosen direction. Animals were then allowed to roam freely, moving via a mixture of walking and hopping, until all individuals reached the edge of the arena.

(a). Video digitization

A total of 119 trials (n = 35 for GS1, n = 38 for GS5, n = 46 for GS10) were digitized using the MATLAB application DLTdv8a (described in [26]). The path of each animal was digitized from the lifting of the cup to the moment the animal reached the arena’s edge (electronic supplementary material, figure S2). After an outlier analysis, animals that took longer than 18 s to complete the trial were excluded from all further analyses (electronic supplementary material, figure S3).

(b). Statistical analysis

To evaluate whether the amphipods were oriented in a uniform circular distribution, Rayleigh tests were applied to the animals’ final orientation, and 95% confidence intervals for their mean angle were calculated. This was first done in the original frame of reference, which was relative to the arena itself, to test for orientation behaviour to the experimental set-up. Then this was done in the transformed frame of reference, which was rotated such that 0° was the direction of the shortest distance to the wrack line, to test for orientation to the beach. The (circular) mean angle of their final orientation was calculated across all individuals within a trial for the grouped animals to prevent issues of non-independence (figure 2a).

Figure 2.

(a) The animal’s final orientation relative to (left) the beach (where 0° = the bearing of the shortest distance to the wrack line), and (right) the arena. The black dots represent the mean final orientation for each trial. The centre arrow represents the overall mean angle, where the length of the arrow is scaled by the mean resultant vector length. The black dotted lines represent the 95% confidence interval. The p-value is calculated from a Rayleigh test. (b) The initial location of all animals (blue points) within the acclimation cup (black outer circle). The black dot marks the centre of the arena. The blue arrowhead points in the direction of the wrack line. Individuals (left) include only GS1 animals, grouped (right) include both GS5 and GS10 animals.

Linear mixed-effect regression models were used to determine the effect of cloud cover (cloudy/clear) on the animal’s behaviour, where group size and trial number were set as random effects (see electronic supplementary material). Similar models were used to determine if there was any effect of year, wherein cloud cover was also included as a random effect. Likelihood ratio tests were computed to obtain p-values. The significance threshold for each hypothesis was set as 0.006 after using a Bonferroni correction for multiple testing.

Eight metrics were used to assess the animal’s movement behaviour—for details on how each metric was calculated, see table 1: (i) Latency: the time it took the animal to start moving after the acclimation cup was lifted, (ii) Time: the time between when the animal began moving to when they reached the edge of the arena (not including latency), (iii) Speed: their average speed (calculated as their actual distance, or the sum of all step lengths, divided by time), (iv) Excess Dist: a measurement of their path length (calculated as their actual distance travelled, minus the straight-line distance between the beginning and end of their path), (v) Initial Bearing: the angle between their first and second locations (where 0° indicates the direction of the shortest distance to the wrack line), (vi) Initial MRV: the spread of their initial bearing, calculated as the mean resultant vector (MRV) length (where 0 is high spread and 1 is no spread), (vii) Final Orientation: the angle between their final location and the centre of the arena (where 0° indicates the direction of the shortest distance to the wrack line), and (viii) Final MRV: the spread of their final orientation, calculated as the MRV length.

Table 1.

A summary of all eight movement variables of interest, where: T_start = the time at which the trial began (acclimation cup was lifted); T_move = the time at which the animal began moving; T_end = the time at which the animal reached the edge of the arena; (x_i , y_i) = the x and y coordinates of each animal, for each frame i, where i = 1 is when the animal began moving and i = n is when the animal reached the edge of the arena; the centre of the arena is located at (x = 0, y = 0), and the angle between the wrack line and the centre of the arena is 0°; I_i = initial bearing of individual i; F_i = final orientation of individual i; the actual distance the animal travelled is represented as $D={\sum }_{i=1}^{i=n}\sqrt{{\left(x_i-x_{i+1}\right)}^2+{\left(y_i-y_{i+1}\right)}^2}$; the straight-line distance between the animals start and end position is represented as ${\displaystyle D_{\mathrm{s}\mathrm{t}}=\sqrt{{\left(x_1-x_n\right)}^2+{\left(y_1-y_n\right)}^2}}$.

metric	calculated per	formula	description
latency	individual	${\displaystyle T_{\mathrm{m}\mathrm{o}\mathrm{v}\mathrm{e}}-T_{\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{r}\mathrm{t}}}$	time between the start of the trial and the animals’ first movement
time	individual	${\displaystyle T_{\mathrm{e}\mathrm{n}\mathrm{d}}-T_{\mathrm{m}\mathrm{o}\mathrm{v}\mathrm{e}}}$	time between first movement and the end of the trial
speed	individual	${\displaystyle \frac{D}{T_{\mathrm{e}\mathrm{n}\mathrm{d}}-T_{\mathrm{m}\mathrm{o}\mathrm{v}\mathrm{e}}}}$	actual distance travelled divided by the time between first movement to the trial’s end
excess dist	individual	${\displaystyle D-D_{\mathrm{s}\mathrm{t}}}$	actual distance travelled minus the straight-line distance
initial bearing	individual	${\displaystyle a\mathrm{t}\mathrm{a}\mathrm{n}\left[\left(x_1-x_2\right),\left(y_1-y_2\right)\right]}$	the direction of their first movement
initial MRV	subsample	${\displaystyle {\displaystyle \sqrt{{\left[\sum _{i=1}^{i=n}\cos \left(I_i\right)\right]}^2+{\left[\sum _{i=1}^{i=n}\sin \left(I_i\right)\right]}^2}}}$	the spread of each subsample’s initial bearing
final orientation	individual	$atan\left[\right(0-x_n),(0-y_n\left)\right]$	the angle between their final location and the centre of the arena
final MRV	subsample	${\displaystyle {\displaystyle \sqrt{{\left[\sum _{i=1}^{i=n}\cos \left(F_i\right)\right]}^2+{\left[\sum _{i=1}^{i=n}\sin \left(F_i\right)\right]}^2}}}$	the spread of each subsample’s final orientation

Finally, the eight metrics were compared between animals released individually and animals released in a group using a non-parametric bootstrapping analysis that is described in the electronic supplementary material (electronic supplementary material, figure S4). Briefly, for the animals released in a group, a single animal was randomly selected from each trial, and the average of each movement metric was calculated across this subsample. This sampling process was repeated 10, 000 times to generate a null distribution which was then compared to the average value of each movement metric across all GS1 individuals. The p-value thus represents the percentile of the null distribution that is greater (or less) than this GS1 value.

3. Results

The final orientations of M. pugettensis exhibited significant evidence of unimodal orientation relative to the beach (p < 0.001), with a mean angle of 18^o ± 7^o (figure 2a). Additionally, the animals were not significantly oriented towards any feature of the orientation arena (p > 0.05, figure 2a).

When clear days were compared with cloudy days (defined as days where the sun was obscured, see electronic supplementary material), we found that M. pugettensis exhibited changes in Time, Excess Dist, Final Orientation and Final MRV (electronic supplementary material, table S1 and figure S5). Owing to the unequal distribution of cloud cover across treatments, the data were filtered to only include clear days for all further analyses. We found no significant relationship between any of the movement metrics and the year in which a trial was conducted (electronic supplementary material, table S1). The total remaining number of trials was 83 (GS1: n_trials = 28, n_animals = 28, GS5: n_trials = 28, n_animals = 139, GS10: n_trials = 27, n_animals = 266).

It was found that the animals released in a group were starting the trial closer to the edge of the acclimation cup, specifically the edge that was closest to the wrack line, compared with those released individually (figure 2b). This resulted in the GS1 animals travelling approximately 1–2 cm longer straight-line distances than those released in a group (electronic supplementary material, figure S6). Therefore, to assess their path length, we opted to use the metric Excess Dist to assess if the grouped individuals were travelling comparatively longer path lengths, regardless of their starting location.

Table 2 shows the mean, standard deviation and p-value from the non-parametric bootstrap analysis for each of the eight movement variables of interest. Detailed results from the non-parametric bootstrap analysis are visualized in electronic supplementary material, figure S7. Five of the eight movement variables of interest were significantly different between GS1 and GS10: animals released individually took approximately 11 s longer to start moving (p < 0.001), moved approximately 1 cm s⁻¹ faster (p = 0.016), travelled approximately 2.7 cm longer paths (p = 0.003), had a different initial bearing by approximately 72^o (p = 0.003) and exhibited more spread in their initial bearing by an MRV of approximately 0.26 (p = 0.009) than GS10 individuals. There was no significant difference between GS1 and GS10 for time or in the direction or spread of their final orientation (figure 3). GS5 showed similar trends to GS10 but with a reduced effect size (electronic supplementary material, figure S8).

Table 2.

The mean ± standard deviation of each of the eight movement variables of interest for each group size (for circular variables 0° is towards the wrack line). For most variables, these numbers are calculated on the raw data. For MRV, these values are calculated on the raw data for GS1 and on the bootstrapped data for GS5 and GS10 to control for sample size. p-Values from the non-parametric bootstrap analysis comparing GS1 to GS5 are reported under GS5, and under GS10 for comparisons between GS1 and GS10, with significant values bolded. Significance codes are as follows: ***p < 0.005, **p < 0.01, *p < 0.05,^•p < 0.09.

	GS1	GS5	GS10
latency (s)	12 ± 6.6	1.3 ± 2.8 (p < 0.001***)	1.2 ± 3.4 (p < 0.001***)
time (s)	3.8 ± 2.2	3.8 ± 1.7 (p = 0.443)	4.1 ± 2.4 (p = 0.255)
speed (cm s−1)	5.9 ± 1.7	5.3 ± 2.3 (p = 0.067 •)	4.9 ± 2.4 (p = 0.016*)
excess dist (cm)	8.5 ± 7.4	7.1 ± 5.4 (p = 0.057 •)	5.8 ± 4.7 (p = 0.003**)
initial bearing (deg)	−55 ± 122	5.7 ± 87 (p = 0.004***)	17 ± 83 (p = 0.003***)
initial MRV	0.11	0.34 (p = 0.012**)	0.37 (p = 0.009***)
final orientation (deg)	19 ± 55	17 ± 40 (p = 0.332)	19 ± 45 (p = 0.45)
final MRV	0.64	0.78 (p = 0.008**)	0.74 (p = 0.086 •)

Figure 3.

Violin plots of the raw data from each animal with the associated p-values from the non-parametric bootstrap analysis comparing GS10 to GS1. For circular data: these are plotted on a circular graph where 0° is towards the wrack line, the arrow points in the mean direction of orientation, the length of the arrow is scaled by the mean resultant vector length and p_d = the p-value comparing the directionality of their orientation while p_MRV = the p-value comparing the mean resultant vector length.

4. Discussion

The goals of this study were to investigate whether M. pugettensis demonstrated beachward orientation and to explore whether there were differences in their movement behaviour when released individually versus in a group. We predicted that animals in a group would travel more efficiently (characterized by shorter paths and trial times) and have more consistent orientation (characterized by headings with less spread and more closely oriented towards the shortest distance to the wrack line).

Megalorchestia pugettensis exhibited significant evidence of unimodal orientation. Their mean angle was closely aligned in the direction of the wrack line, providing the first evidence for beachward orientation in this genus and a useful system for future comparative studies. The finding that their orientation was highly affected by cloud cover indicates the potential use of a celestial compass, consistent with the capabilities observed in other talitrid amphipods [1]; however, this result could also be driven by differential motivation, perhaps due to a decreased risk of desiccation as a result of lower air temperatures.

The differences in latency between individuals and animals in groups may be a result of a threat response or driven by social interactions. Similar cases have been observed in several other animals. For example, in pigeons, birds released individually spend longer circling at the release site before homing in comparison to birds released in small flocks [23]. In fruit flies, the introduction of a potential threat caused individual animals to ‘freeze’ more often and for longer than those in groups [27]. Finally, in phrynosomatid lizards, movement latency was lower when male lizards were in the presence of females [28]. It is largely unknown how sex or other social demographics may specifically affect talitrid amphipod behaviour. However, our animals’ confinement together during the acclimation period was likely more socially and physically stimulating than for an animal confined alone.

In terms of path efficiency, we found that animals released individually travelled longer paths. This was due in part to the fact that they were travelling more circuitous routes, but also because, unlike the grouped animals, they were not already orienting during the acclimation period. Surprisingly, this did not result in the individual animals taking more time to reach the edge of the arena once they started moving, due to their higher speed. In terms of orientation, at the beginning of the trials, we found that animals released individually were less oriented towards the wrack line and had greater spread in their mean orientation than grouped individuals. However, this difference disappeared by the end of the trial, a result that agrees with the findings of Scapini et al. [18]. This indicates that, at some point, the individual animals successfully reoriented themselves.

Similarly to the latency results, it is possible that these differences in path efficiency and initial orientation were the result of a threat response. This could be an example of the ‘selfish herd’ phenomenon, whereby animals can reduce their individual level of danger by increasing their number of neighbours [29]. Speculatively, individual animals may have opted to move more quickly and erratically, thereby sacrificing their path efficiency, to elude potential predators. Alternatively, it is possible these differences occurred because of the ‘many wrongs’ phenomenon, which predicts that individuals who each possess a noisy estimate of the ‘correct’ direction can improve their navigational accuracy by travelling in a group and averaging across each individual’s estimate, thereby resulting in more consistent orientation towards the goal, and straighter, more efficient paths [16,17]. If this is the case, the amphipods released in a group may have been better able to immediately determine the direction of the wrack line. Further experimentation is needed to distinguish between these two options.

Notably, although Talitridae are gregarious and frequently found in dense populations in the wild, it is currently unknown whether they can be classified as social [30,31]. While previous work on the many-wrongs phenomenon has primarily focused on social animals, this research suggests that the many-wrongs phenomenon could extend to animals outside of the classical definition of sociality, or that these animals exhibit a previously unrecognized level of social interaction. However, such interpretations are speculative at this stage and require further investigation. Further experiments should also explore whether results differ based on life stage or sex.

Although there is uncertainty in the underlying mechanism driving the differences seen in the movement behaviour of amphipods released individually versus in a group, we conclude that solo individuals travelled less efficient paths than those in a group. The reduced effect size shown in the GS5 animals fits with current theoretical work on collective navigation, and indicates the potential for Allee effects, whereby sudden collapse of populations can occur if the densities drop below a critical threshold [32,33]. Finally, our results suggest that future researchers investigating talitrid orientation should put careful thought into how group dynamics may affect their experimental design.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

Data are available from the Dryad Digital Repository [34].

Supplementary material is available online [35].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

M.L.P.: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, writing—original draft, writing—review and editing; J.N.G.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, writing—original draft, writing—review and editing; S.J.: conceptualization, funding acquisition, project administration, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

Air Force Office of Scientific Research, grant no. FA9550-20-1-0399.