Within-shoal phenotypic homogeneity affects shoaling preference in a killifish

Abstract

Anti-predator benefits associated with living in groups are multiple and taxonomically widespread. In fish shoals, individuals can exploit the confusion effect, whereby predators struggle to target a single individual among several individuals. Theory predicts that the confusion effect could be aided by homogeneity in appearance; thus, individuals should group by phenotypic characteristics, contributing to generating high within-shoal phenotypic homogeneity. While assortments by body size have been extensively documented, almost nothing is known about whether within-shoal homogeneity in body pigmentation affects shoaling preference. To investigate this issue, we used the Mediterranean killifish, Aphanius fasciatus, a shoaling species characterized by conspicuous vertical bars on body sides. Individual females were given a choice between two novel shoals characterized by either a high or low degree of homogeneity in the number of bars. As predicted, individual females preferentially associated with the shoal showing the higher phenotypic homogeneity. Our data demonstrated that fish might associate with the shoal that maximizes phenotypic homogeneity in body pigmentation, irrespective of their own phenotype.

1. Introduction

Living in groups is one of the key components of many animal societies, and the presence of predators has probably been the most important driver for the evolution of this behaviour [1,2]. Anti-predator benefits associated with living in groups are multiple and well documented in many species [2]. Examples include enhanced vigilance and early predator detection, but because predator avoidance is not always possible, the success of anti-predator behaviours is essential for survival. For instance, fish shoals commonly exploit the confusion effect, whereby predators struggle to select and track a single individual when several targets are present [3,4]. Theory predicts that the confusion effect could be aided by homogeneity in appearance [5], suggesting that individuals within shoals should be phenotypically more similar than individuals between shoals (reviewed in [6,7]). While assortments by body size have been extensively documented (reviewed in [8]), assortments by body pigmentation have received much less attention [6]. Almost nothing is known about how within-shoal homogeneity in body pigmentation affects shoaling preference. The presence of vertical bar patterns in several fish species [9] offers a valuable opportunity to investigate this hypothesis. We used the Mediterranean killifish (Aphanius fasciatus) to test whether within-shoal homogeneity in the number of vertical bars affected shoaling preference. In this species, females are particularly social [10,11] and show a series of vertical narrow black bars on a light-grey background (figure 1a) that remain expressed throughout the whole year [12], suggesting a possible functional role in shoaling behaviour. The Mediterranean killifish inhabits shallow brackish waters, organized in a network of intertidal creeks [13]. At each tide cycle, adults are forced to leave refuge for deeper waters in the open lagoon, undergoing predation by large piscivorous fish, such as the European seabass, Dicentrarchus labrax [13,14]. When fish are shoaling in open waters, high within-shoal homogeneity in bar pattern may make it difficult for a piscivorous predator to target a specific individual [3,5]. We thus predicted that, when given a choice between two novel shoals, individual females would prefer to associate with the shoal characterized by the higher degree of phenotypic homogeneity in the number of vertical bars, irrespective of their own phenotype.

Figure 1.

(a) Two female killifish characterized by different number of bars; (b) schematic view of the experimental tank.

2. Material and methods

Aphanius fasciatus females were collected from small creeks in the Venice Lagoon (Italy) and maintained at the Umberto D'Ancona Hydrobiological station (Chioggia, Italy); see electronic supplementary material for details on fish maintenance. We screened 100 individuals for the number of vertical bars: after placing each female in a small glass tank (25 × 10 × 4 cm), we directly counted the number of bars. The number of bars comprised between 7 and 12 (six bar categories overall). Fish characterized by an identical number of bars were allocated for 7 days to three 35 × 35 × 35 glass tanks (A, B, C) (electronic supplementary material, figure S1). We randomly selected 26 females as focal fish and 54 females as stimuli (80 females overall). We performed 26 trials in total, and in each trial, we used two groups of stimuli: a homogeneous and a non-homogeneous group. Each homogeneous group consisted of three females from the same bar category but collected from three different tanks (A, B, C) to ensure that individuals were unfamiliar when the experiment began (electronic supplementary material, figure S1). Each non-homogeneous group consisted of three females from three different bar categories. Females in non-homogeneous groups differed by at least two bars (mean ± s.d. = 2.375 ± 0.489). In order to prevent focal fish from associating with stimuli that most resembled their own phenotype (e.g. [15,16]), in each trial, the focal fish was characterized by a number of bars different from that of the stimuli and could randomly have more or fewer bars than the fish in the homogeneous group. Because also size might affect fish social preference [8], fish were size-matched within each trial by visually comparing the fish before the experiment. The same group of stimuli was used for three subsequent trials and then a new group of stimuli was used. The binary preference test was based on a well-established procedure [17,18] (see electronic supplementary material for further details). The experimental tank was virtually divided into three same-sized areas: a central no-choice area, a choice area for the homogeneous stimulus and a choice area for the non-homogeneous stimulus (figure 1b) and we measured the time spent by the focal fish near the homogeneous versus the non-homogeneous group. Moreover, we recorded also the behaviour of stimulus shoals to control for behavioural differences between the stimuli (see electronic supplementary material). We performed a linear mixed-effects model to compare the absolute time spent near the homogeneous and the non-homogeneous groups. We included stimulus type as fixed factor, number of bars of focal fish and number of bars of fish in the homogeneous group as covariates, trial identity and stimulus group identity as random factors. A one-sample t-test was performed to examine whether the preference for the homogeneous group (calculated as time near the homogeneous group/time near the two groups) was significantly greater than chance (0.5). Finally, we compared the number of visits to each stimulus performing a paired t-test to investigate the accuracy of shoal preference (i.e. the subjects had similar selectivity in the two choice areas). Analyses were performed using R v. 3.3.1 [19].

3. Results

Female killifish spent 76.22 ± 8.31% (mean ± s.d.) of time in the two choice areas shoaling with the stimuli. In 16 out of 24 trials (approx. 67%), females spent the majority of time near the homogeneous group and only in five out of 24 trials (approx. 21%) did they spend the majority of time near the non-homogeneous group (figure 2). Preference was significantly affected by type of stimulus (χ² = 8.087, p = 0.004), but was not affected by number of bars of focal fish (χ² = 0.025, p = 0.875) or by number of bars of fish in the homogeneous group (χ² = 0.354, p = 0.552). When we tested the preference for the homogeneous group with chance level, we found that females showed a significantly greater preference for the homogeneous group than expected by chance (t_1,23 = 2.272, p = 0.016; figure 2). Finally, we found no significant difference in the number of visits between the two stimuli (t_1,23 = 0.687, p = 0.499; homogeneous group: mean ± s.d. = 5.58 ± 3.58; non-homogeneous group: mean ± s.d. = 5.78 ± 3.26), suggesting that focal fish assessed both stimuli similarly.

Figure 2.

(a) Shoaling preference for the homogeneous group in each trial and the associated 95% binomial confidence intervals, and (b) mean shoaling preference for the homogeneous group and the associated 95% binomial confidence interval.

4. Discussion

Here, we tested whether within-shoal homogeneity in the vertical bar pattern affected shoaling preference. As predicted, when given a choice between two shoals characterized by either high or low degree of homogeneity in the number of bars, individual females preferentially associated with the shoal showing the higher phenotypic homogeneity. The few studies investigating the role of body pigmentation in shoaling preference found that fish preferred to associate with individuals that matched their own pigmentation phenotype (e.g. [15,16,20]), relying on their own phenotype as a reference [6]. Alternatively, fish could prefer a specific pigmentation phenotype on the basis of the experience during development [21]. So far, shoal assortments by body pigmentation have been explained by mechanisms of phenotype matching and learned preference. Our study demonstrated that fish might associate with the shoal that maximizes the homogeneity in appearance, irrespective of their own phenotype. Moreover, we found that number of bars of the homogeneous group did not affect shoaling preference of focal fish, suggesting that focal fish were not attracted to a specific bar phenotype. It is likely that focal fish assessed the homogeneity in the number of bars as trait of the group rather than the number of bars per se. Fish could discriminate between the two types of stimuli by counting the number of bars and/or using continuous variables such as the total area occupied by bars [22]. Although a combination of numerical and continuous information is likely to represent the common condition for fishes [23], the precise mechanism underlying the discrimination between groups of different homogeneity in body pigmentation remains obscure.

Because within-shoal phenotypic homogeneity is thought to reduce predation risk [3,4], phenotypic variability is expected to be lower within shoal than between shoals [5]. In natural populations, shoals are generally composed of familiar individuals [24] and familiarity has been demonstrated to confer multiple benefits, for instance, in terms of foraging efficiency [25,26]. Single fish could associate with shoals composed of familiars to try to gain access to such benefits [27]. Sticklebacks, Gasterosteus aculeatus, are able to discriminate between two novel shoals and to bias the preference towards the shoal composed of individuals familiar to one another [28]. Preferring shoals composed of individuals familiar to one another, outsider fish reached food sources and began feeding more quickly than when preferring shoals of unfamiliar individuals [28]. The exploitation of such benefits could be particularly useful when outsider fish have no prior information about the environment. The Mediterranean killifish is a mobile species within its habitat [13] and shoaling with familiar individuals may facilitate the exploration of novel environments [10]. Outsider fish may obtain benefits from associating with shoals composed of individuals that are familiar to one another [28]. Shoal homogeneity in body pigmentation could be the means by which single fish could distinguish between shoals. Moreover, many species show heritability in body pigmentation [29] and thus fish with a similar number of bars may be more genetically related than fish with a different number of bars. Focal fish could be also attracted to the shoal that reflects the higher genetic familiarity. Clearly, these hypotheses remain open, and further studies are needed to investigate the role of shoal homogeneity in body pigmentation as a social cue to be used by potential outsider individuals.

Ethics

The University of Padova Ethical Committee approved the experimental procedures (protocol no. 10985). Sampling was conducted under permission of Veneto Region (protocol no. 207154).

Data accessibility

The dataset supporting this article has been uploaded as part of the electronic supplementary material.

Authors' contributions

S.C. and M.G. designed the study, conducted the experiment and wrote the manuscript. S.C. performed analysis. Both authors agreed to be held accountable for the content herein and gave final approval for publication.

Competing interests

The authors declare no competing interests.

Funding

This work was supported by a grant from the University of Padova to M.G. (grant no. PRAT-2015-CPDA153859).