Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Schooling Fish Under Attack Are Not All Equal: Some Lead, Others Follow

Abstract

Animal groups such as fish schools, bird flocks and insect swarms appear to move so synchronously that they have long been considered egalitarian, leaderless units. In schooling fish, video observations of their spatial-temporal organization have, however, shown that anti-predator manoeuvres are not perfectly synchronous and that individuals have spatial preferences within the school. Nonetheless, when facing life-or-death situations, it is not known whether schooling fish react to a threat following a random or a hierarchically-based order. Using high-speed video analysis, here we show that schooling fish (Golden grey mullet, Liza aurata) evade a threat in a non-random order, therefore individuals that are first or last to react tend to do so repeatedly over sequential stimulations. Furthermore, startle order is strongly correlated with individual positional preferences. Because school members are known to follow individuals that initiate a manoeuvre, early responders are likely to exert the strongest influence on the escape strategy of the whole school. Our results present new evidence of the intrinsic heterogeneity among school members and provide new rules governing the collective motion of gregarious animals under predator attack.

Citation: Marras S, Domenici P (2013) Schooling Fish Under Attack Are Not All Equal: Some Lead, Others Follow. PLoS ONE 8(6): e65784. https://doi.org/10.1371/journal.pone.0065784

Editor: Gonzalo G. de Polavieja, Cajal Institute, Consejo Superior de Investigaciones Científicas, Spain

Received: February 1, 2013; Accepted: April 29, 2013; Published: June 12, 2013

Copyright: © 2013 Marras, Domenici. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Funding: Funding was provided by the authors' institution, CNR (Consiglio Nazionale delle Ricerche). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Group living is a widespread behaviour, observed in many animal taxa [1]. The collective manoeuvres of gregarious animals reflect remarkable coordination, as is commonly observed in fish schools and bird flocks. In fish, collective behaviour has long been assumed to be generated in a self-organized manner in which individuals within the school appear to move in synchrony like an egalitarian superorganism [2], [3], [4]. This view has changed in more recent years because high-speed video observations on the spatial-temporal organization of schooling fish have shown that their anti-predator manoeuvres are not perfectly synchronous [5], [6] and behavioural observations have demonstrated that individuals have spatial preferences within the school due to a number of individual attributes [1], [7], [8], [9], [10].

Predator avoidance is considered one of the main advantages of schooling in fish [8], however, very little is known about the behaviour of each school member during anti-predator manoeuvres. Previous studies have mainly investigated the escape response of solitary fish, focusing on locomotor kinematics [11], [12], [13]. Escape responses consist of brief and sudden accelerations in a direction away from the threat and are used by most fish species to avoid predation [14]. One of the main variables affecting prey survival is the timing with which individuals react to an approaching predator [15], [16]. A fundamental component of timing is escape latency, defined as the time interval between the stimulus onset and the first detectable movement of the prey [11]. Short escape latencies, of the order of 10–20 ms, are due to the neural control of a pair of giant neurons present in most fish species (Mauthner cells, [17], [18]), although alternative neural circuits with longer reaction times are known [19]. Previous work has shown that schooling fish tend to have longer latencies than solitary fish, possibly as a strategy for avoiding collisions among neighbouring fish [20]. This suggests that social interactions may be the main factor affecting the timing of the reaction of fish in a school, rather than neuromuscular performance as in solitary fish [21]. While some of the rules underlying the rapid decision-making processes at the base of the coordination of collective manoeuvres in fish and other taxa have been revealed [6], [22], [23], [24], [25], [26], the possibility that leadership-followership rules govern the motion of gregarious animals under predator attack remains to be tested.

Here we investigated the hypotheses that the anti-predator response of a school of fish reflects a non-random startle order, whereby school members that are the first or last to respond to a threat, tend to do so throughout a sequence of stimulations, and that startle order of school members is modulated by positional preferences. These hypotheses were tested in schools of 10 juvenile golden grey mullet (Liza aurata). This school size is biologically relevant for many species of nearshore and freshwater fish [9], [27], [28] including mugilids in coastal areas [29], [30].

Materials and Methods

Ethic Statement

The fish were held, and the non-lethal experiments were conducted, in accordance with the laws governing animal experimentation in Italy. The IAMC-CNR facility of Oristano, where the fish were held and the experiments performed, is recognised by the Italian Government as a certified facility for fish rearing and ecophysiological experimentation (D.lgs. 116/92, Decreto n° 136/2011-A).

Animals

A total of 70 juvenile golden grey mullet (Liza aurata) [13.4±1.2 cm total length and weighing 14.7±4.5 g (mean±S.D.)] were used. Fish were maintained in a circular tank (150 cm diameter and 50 cm of depth) supplied with recirculated and filtered natural seawater (salinity 29%, temperature 19–20 C°) under natural photoperiod. Fish were fed ad libitum twice a week on commercial dry pellets. Feeding was discontinued 2 days before the tests.

Experimental Set-up and Protocol

Experiments were carried out in a circular tank (200 cm diameter × 150 cm depth and 36 cm water depth) which was supplied with recirculating seawater at 19–20°C (Fig. 1). Fish were tested in randomly assorted schools of ten individuals. Before the test, fish were marked dorsally, in front of the dorsal fin, using a mix of cyanoacrylate glue and non-toxic white powder insoluble in water (Titanium (IV) oxide, Sigma). Different symbols of approximately 1×0.5 cm were used for each fish so that individuals could be discriminated during the video analysis. Marking took 1 minute for each fish. After marking, fish were placed in a rectangular tank (60×40 cm and 20 cm water depth) and then introduced in the experimental tank all at the same time. They were left undisturbed for 60 minutes of acclimation, after which escape responses were induced by mechanically stimulating the school. The mechanical stimulus was a cylindrical object (10 cm height, 2 cm diameter and weighing 35 g) consisting of a PVC tube with a tapered point, attached to an iron disk in the opposite side. The stimulus was released by an electromagnet from a height of 150 cm above the water surface. To prevent visual stimulation before contact with the water surface, the stimulus was released near the centre of the circular tank, into a vertical opaque PVC tube (10 cm diameter) ending 1 cm above the water surface [31].

thumbnail
Download:
Figure 1. Experimental setup.

Experimental tank in which schooling fish were startled by the stimulus released by an electromagnet and recorded using a high-speed camera positioned above the tank. The contact between stimulus and water surface was reflected by a mirror and recorded by the camera. The field of view of the camera is represented by the area delimited by broken lines.

https://doi.org/10.1371/journal.pone.0065784.g001

A high speed camera (Fastec Imaging, Ranger 1000) was positioned directly above the experimental tank and recorded the escape response at 250 Hz (Fig. 1). A moderate circular flow (∼3 cm s–1) was induced in the experimental tank by the inflow of the re-circulating filter system. This flow elicited positive rheotaxis, which induced the school to maintain a relative stable position in the tank for the duration of the experiment, allowing stimulation to be delivered consistently to the side of the school (88±20°). To video record the time of the impact between stimulus and the water surface, a mirror inclined at 45° was attached to the end of the vertical PVC tube [32] (Fig. 1). The camera was triggered to record from 1 second before the stimulation to 3 seconds after the stimulation. Each school was stimulated 10 times (hereafter defined as trials, with ten trials constituting a “block”, Fig. 2A), at 10-minute intervals, after which fish were removed from the experimental set-up and measured for length and weight. A total of 7 schools were tested. For some schools, technical problems caused the loss of some trials (a total of 6 trials in 3 schools). Escape sequences were analyzed using Redlake MotionScope PCI (Ver. 2.21.1.).

thumbnail
Download:
Figure 2. Fish show a non-random startle order.

(A) Example of an experimental block, defined as a series of ten trials (ten successive stimulations of the same school). Numbers in italics indicate SO of a given fish in a given trial (e.g. in trial #1, fish A is the 4th individual to react). (B) Individual fish tended to retain similar startle orders through 10 successive escape responses. Examples of a fast (filled symbols, continuous line) and a slow reacting fish (open symbols, dotted line) within a school. (C) The SOExp-M distribution (black columns, N = 70) showed a statistically wider variance than all of the SORand-M distributions [grey columns represent the mean ± S.E. of the ten simulations of SORand-M (N = 70 for each simulation)] and the SOMC distribution (dark line, N = 20000).

https://doi.org/10.1371/journal.pone.0065784.g002

Experimental and Random Variables

The mass (M) and fork length (BL) of each fish were used to calculate condition factor (Kf, 100*M/FL3), as an index of the relative stoutness of each individual. Latency (Le) was defined as the time interval in ms between the instant when the stimulus broke the water surface and the first detectable escape movement of the fish. If a fish did not react to the stimulus, a latency corresponding to the longest value recorded in our data was assigned to these non-responders. School latency (Le–s) was defined as the mean Le of the individuals of a given school for each trial. Distance from stimulus (Ds) was measured as the distance between the centre of the mechanical stimulus and the centre of mass of the fish measured on stretched-straight specimens [33] (0.4 Length from the tip of the head, [34]). School distance (Ds–s) was calculated as the mean Ds of the individuals of a given school for each trial. The experimental startle order (SOExp) was determined by ranking the individual escape latency (Le) within each school for each of the 10 stimulations (trials). SOExp ranged from 1 (first responder) to 10 (last responder) according to Le. Tie events were exceptional (5 events in total). In these cases the fish that were tied were ranked randomly. In the case of a no-response event for a single fish (14 events in total), that fish was ranked last. In cases when multiple fish did not respond (11 events in total), the non-responsive individuals were randomly placed at the end of the ranking. Overall, the proportion of non-responding fish at the individual level when considering all trials was 7%. A mean experimental startle order (SOExp-M) was calculated for each individual through the 10 trials.

Positional preference was measured, for each individual, as the proportion of times (in % out the ten trials) a fish was in the front (FpExp) and at the edge of the school (EpExp), respectively (Fig. 3). In each trial, within-sector SOExp was determined for each individual by ranking the escape latency within its sector of preference, defined as the sector it occupied most often during the 10 trials. Ties for sector preference were split randomly. Because the number of individuals within each sector varied, within-sector SOExp data were normalized using ((SOExp−1)/(N−1)) [35], where N represents the number of fish within that sector. Thus, within-sector SOExp ranged from 0 to 1. For each individual, within-sector SOExp-M was calculated as the mean of all its within-sector SOExp.

thumbnail
Download:
Figure 3. Fish positions within a school.

Top view of a school of 10 fish. (A) Individuals were numbered according to their position relative to the orientation of the school (Os, indicated by the frontal filled arrow). Fish in the first half of the school (position 1 to 5) were considered to be in the front, while fish in the second half of the school (position 6 to 10) were considered to be in the back. (B) a fish was considered to be at the edge of the school if its tip of the head was at the vertex of the smallest convex polygon enclosing the entire school (1). (C) The four sectors (front-edge, F–E; front-centre, F–C; back-centre, B–C and; back-edge, B–E) were assigned to individuals of the school according to the definitions given for Figures 3A and 3B.

https://doi.org/10.1371/journal.pone.0065784.g003

A set of random variables was determined. Within each school, for each of the ten trials, ten sequences of random numbers (obtained using Research Randomizer, Royal Psychology Network; www.randomizer.org/form.htm) ranging from 1 to 10 were generated (SORand). Seven random schools thus generated were equivalent to the seven experimental schools in terms of number of trials as well as missing data. A mean random startle order (SORand-M) for each individual was calculated as the mean of all its SORand. A random probability distribution of startle order (SOMC) was obtained using a Monte Carlo simulation which generated random permutations. We generated a large number of random data sets using a custom-made Fortran 90 code. We judged SOMC to be stationary when an N of blocks = 2000 was used (i.e. 20000 individuals in 2000 schools composed of 10 fish each), rendering redundant a full analytical computation of all possible permutations (10!10). Two sets of distributions of SO simulations based on sorting rules (see Methods S1) were obtained using the same procedure as for SORand-M (a simulation of seven 10-trial blocks, repeated 10 times) and SOMC (2000 10-trial blocks). Two sets of distributions of SO simulations based on sorting rules were obtained according to the same scheme as for SORand-M (a simulation of seven 10-trial blocks, repeated 10 times) and SOMC (2000 10-trial blocks). Sorting rule A was based on a predetermined circular sequence that followed a fish that was randomly chosen as first responder (see Methods S1). Sorting rule B was based on a first responder chosen at random, followed by responders whose probability of startling was based on the proximity to the previous responders (e.g. fish #9 being proximal to fish #8 and fish #10), (see Methods S1). For each repetition, positions in the front-back (FpRand) and edge-center (EpRand) were randomly reassigned using the proportion observed in any given experimental trial. The within-sector SORand was derived by randomly reassigning the within-sector SOExp. For each individual, this procedure was applied only within the sector for which they had shown a relative preference during the experimental trials. Ties for sector preference were split randomly. An individual mean random startle order (within-sector SORand-M) was calculated as the mean of all its within-sector SORand.

Statistics

To test whether escape latency was affect by fish body length (BL), condition factor (Kf) or stimulus distance (Ds), at the individual level, we ran a multiple regression having the mean Le for each individual as dependent variable, and the individual BL, Kf, and mean Ds as independent variables. To test the effect of Kf, BL and Ds on escape latency at the school level, we ran a multiple regression having the mean Le–s for each school as dependent variable, and Ds–s, mean BL and mean Kf, and as independent variables.

Testing whether SOExp reflects a random order was done in two steps: i) Among-school differences in the variances of SOExp-M were tested using a Bartlett test. ii) If no differences among schools were found, individuals from different schools were pooled and the variances of SOExp-M were compared with those of SORand-M and SOMC using F-tests. The comparison between SOExp-M and SORand-M was performed ten times, with ten different random simulations.

To determine whether each individual had the tendency to keep a preferential position in the school was done in two steps: i) a Bartlett test was used to test whether the variances of FpExp and EpExp were independent of school. ii) If no among-school differences were found, the frequency distribution of the pooled FpExp and EpExp were compared with those of the FpRand and EpRand, respectively, using F-tests.

To test whether SOExp-M was affected by individual positional preference, an ANCOVA with SOExp-M as a dependent variable and school (1 to 7), FpExp and EpExp as independent variables was used. If no effect of school in the ANCOVA was found, data were pooled in order to test the relationship between SOExp and FpExp and EpExp using a multiple linear regression.

In order to partition the effect of individual from that of the sector occupied, defined as one of four possible combinations of front-back and edge-centre (i.e. frontal-edge, F–E; frontal-centre, F–C; back-edge, B–E and; back-centre, B–C; Fig. 3C), we tested if the distribution of within-sector SOExp differed from that of within-sector SORand. Among-school differences in the variances of within-sector SOExp-M and within-sector SORand-M were tested using a Bartlett test, within each sector. If no differences among schools were found, data from different schools were pooled and the distribution of within-sector SOExp-M was compared with that of within-sector SORand-M for each sector using F-tests.

Parametric statistics were used for all tests, except where conditions of normality (D’agostino test) did not apply. In these cases, non-parametric statistics were applied and the Median-based Levene test was used instead of the Bartlett test. A probability less than 5% (p<0.05) was taken as the limit for statistical significance.

Results

Fish body length (BL, 13.4±1.2 cm), condition factor (Kf, value 0.6±0.1) or distance from the stimulus [Ds at the individual level, (62±18 cm); Ds–s at the school level, (62±17 cm)] had no effect on latencies measured at the individual (Le) and at the school level (Le–s) (Multiple linear regressions; p>0.05 in all cases).

The distribution of SOExp-M was independent of school (Bartlett test; p = 0.28) and the variance of the pooled SOExp-M differed significantly from each of ten random simulations (F-test; SOExp-M vs. SORand-M, p<0.001 in all 10 cases, Fig. 2C), with the experimental data showing a wider range of values (1.8–9.0) than those of their random equivalent SORand-M (mean range 3.3–7.6). The pooled SOExp-M differed significantly from SOMC (F-test; SOExp-M vs. SOMC, p<0.001; Fig. 2C). All of the distributions derived from the simulations based on sorting rules are significantly different from the experimental data (F-test; all p<0.05), and not different from SORand-M (F-test; all p>0.05). This indicates that the startle order significantly deviates from random in schooling fish, such that individuals that responded early or late in the first trial tended to respond early or late, respectively, in subsequent trials (Fig. 2B).

The distributions of both FpExp and EpExp were independent of school (Bartlett test; FpExp, p = 0.92; EpExp, p = 0.49). The pooled FpExp and EpExp showed a wider range of values (10–100% and 0–100%, respectively) than those of the FpRand and EpRand (25–86% and 30–89%, respectively) and their variances differed significantly, indicating individual positional preferences (F-test; FpExp vs. FpRand p<0.001; EpExp vs. EpRand p<0.001). Our results show that both individual SOExp and SOExp-M were affected by the position occupied in the school. Individual SOExp was significantly related to the fish position at the time of the stimulation i.e Front or Back and Centre or Edge (Multiple regression; p<0.001). SOExp-M was significantly related to both FpExp (ANCOVA; p<0.001) and EpExp (ANCOVA; p<0.001), independent of school (ANCOVA; p = 0.66), indicating that fish with a tendency to keep frontal and central positions in the school were among the first to escape, while fish that were inclined to keep back and edge positions were among the last (Multiple linear regression, p<0.001; Fig. 4).

thumbnail
Download:
Figure 4. The relationship between mean startle order (SOExp-M) and fish positional tendencies within the school.

The mean startle order depended on the positional tendencies of each individual fish at the time of stimulation. The X- and Z-axes correspond to the % of events in which each individual was in the front and at the edge of the school, respectively. Fish in the frontal and central positions react earlier than fish in the back and at the edge of the school. SOExp-M = 4.78+0.036 EpExp − 0.029 FpExp; p<0.001, R2 = 0.38; N = 70.

https://doi.org/10.1371/journal.pone.0065784.g004

The distribution of within-sector SOExp-M was independent of school (Median-based Levene test; sectors F–E, p = 0.78; F–C p = 0.23; B–E, p = 0.76; B–C, p = 0.40), and their variances of two out of four sectors differed significantly from within-sector SORand-M (F-test; F–E, p = 0.16; F–C p<0.01; B–E, p<0.01; B–C, p = 0.54).

Discussion

The results demonstrate that schooling fish evade a threat in a non-random order, therefore individuals that react early or late tend to do so repeatedly over sequential stimulations. Furthermore, we show that the startle order of each individual fish is modulated by its positional preference.

The finding of non-random startle order is important for elucidating the bases of the collective escape manoeuvres in fish. Predator strikes and prey escapes in fish are typically short lived (in the order of ms) and timing may be a major determinant of survival [14], [15], [16]. In our experiments, while all fish could perceive the stimulus (see Methods S1) the long interval between the first responders (Le, 82±30 ms) and those of the rest of the responders (Le, 255±190 ms) suggests that most school members could potentially have reacted to the motion of the first responders [36], [37]. Although we cannot establish whether late responders reacted to the stimulus or to their neighbours, the motion of startled members of the school is known to affect the response of other fish. Previous studies have shown that gregarious individuals react both to the predator and to their neighbours [8], [38], [39], [40] and that school members tend to follow neighbours that initiate a manoeuvre [41], [42], [43]. Therefore, because certain individuals react consistently before other school members in successive attacks, they are likely to exert a form of leadership affecting the anti-predator manoeuvres of the whole school.

The range of Le observed in the present study is rather wide (36–1468 ms); most values are beyond those expected for minimal response time [44], [45] and higher than those found in solitary fish of the same species (15–19 ms [34]). Therefore, unlike the variation in escape latency observed in solitary sea bass (Dicentrarchus labrax), which was suggested to be caused by differences in neuromuscular performance among individuals [21], the startle order observed here is more likely to be due to behavioural factors. In fish, the riming of escape reactions is influenced by boldness [46], laterality [32] and the perceived risk of predation, e.g. the presence of refuges [47]. In schools, the risk of predation varies in relation to the position occupied [1]; work on schooling chub (Semotilus atromaculatus) showed that the risk of predation tends to be higher in the front of the school [48]. This is consistent with the relationship between positional preferences and reactivity found here, i.e. fish at the rear (with a low perceived risk of predation) tend to react later than front fish (with a high perceived risk). However, we found that fish in the centre of the school tend to react sooner than those at the periphery, in contrast with most previous work which found that fish at the edge of the school have a higher risk of predation [1]. Therefore, the perceived risk of predation may not be the only factor affecting the startle order in schooling fish. In addition to the effect of positional preferences, we found that startle order differed from random in two out of four school sectors, suggesting individual tendencies to keep a given startle order within those sectors.

Previous work found that positional preferences can be related to a number of factors such as hunger level [48], behavioural lateralization [49], and metabolic scope [10]. Hunger-related positional preference were minimized here by the standardized feeding protocol. Consequently, the positional preferences observed are likely to be related to individual traits that are repeatable (metabolic scope [50], [51] and inheritable (lateralization [49]). The position occupied by individuals in a group has attracted a lot of attention in the literature, due to the fundamental trade-offs associated with it. Fish in the front were shown to have higher feeding success [1], [7], although they incur higher swimming costs than trailing fish [10], [52]. By demonstrating that positional preferences affect the startle order of fish under attack, the present results provide a novel relationship that can contribute to our understanding of the heterogeneity among the individuals that make up a school.

To conclude, our findings suggest that certain individuals are likely to play a key role on the survival of a whole school under attack. The heterogeneity in the behaviour of schooling fish revealed here can have important ecological implications because of its potential contribution to within and among schools differences in survival during collective anti-predator manoeuvres. The gregarious behaviour of a number of taxa shows similar characteristics to schooling fish in terms of their decision making and anti-predator strategies [1]. It is therefore possible that a non-random pattern of startle order may be a common feature of the coordinated anti-predator manoeuvres displayed by other animal groups.

Supporting Information

Methods S1.

Additional experiment on the escape reaction in solitary fish and fish in a simulated school and simulation based on sorting rules.

https://doi.org/10.1371/journal.pone.0065784.s001

(DOC)

Acknowledgments

We are grateful to J. Krause, D. J. McKenzie, S. S. Killen, R. S. Batty and J. P. Bacon for useful comments on a previous draft of the manuscript, R. S. Ferrari, A. Satta and F. Antognarelli for technical help with the experiments, and A. Cucco and N. Abaid for probability analysis.

Author Contributions

Conceived and designed the experiments: SM PD. Performed the experiments: SM PD. Analyzed the data: SM PD. Contributed reagents/materials/analysis tools: SM PD. Wrote the paper: SM PD.

References

  1. 1. Krause J, Ruxton GD (2002) Living in groups. Oxford: Oxford University Press.
    • 2. Pyke GH (1978) Are animals efficient harvesters? Animal Behaviour 26: 241–250.
    • 3. Wilson DS, Sober E (1989) Reviving the superorganism. Journal of Theoretical Biology 136: 337–356.
    • 4. Huth A, Wissel C (1992) The simulation of the movement of fish schools. Journal of Theoretical Biology 156: 365–385.
    • 5. Domenici P, Batty RS (1994) Escape maneuvers of schooling Clupea harengus. Journal of Fish Biology 45: 97–110.
    • 6. Marras S, Batty RS, Domenici P (2012) Information transfer and antipredator maneuvers in schooling herring. Adaptive Behavior 20: 44–56.
    • 7. Krause J, Bumann D, Todt D (1992) Relationship between the position preference and nutritional state of individuals in schools of juvenile roach (Rutilus rutilus). Behavioral Ecology and Sociobiology 30: 177–180.
    • 8. Pitcher TJ, Parrish JK (1993) Function of shoaling behaviour in teleosts. In: Pitcher TJ, editor. Behaviour of teleost fishes. London: Chapman & Hall. 363–439.
      • 9. Krause J, Hoare DJ, Croft D, Lawrence J, Ward AJW, et al. (2000) Fish shoal composition: mechanisms and constraints. Proceedings of the Royal Society Biological Sciences Series B 267: 2011–2017.
      • 10. Killen SS, Marras S, Steffensen JF, McKenzie DJ (2012) Aerobic capacity influences the spatial position of individuals within fish schools. Proceedings of the Royal Society Biological Sciences Series B 279: 357–364.
      • 11. Domenici P, Blake R (1997) The kinematics and performance of fish fast-start swimming. Journal of Experimental Biology 200: 1165–1178.
      • 12. Webb PW (1978) Fast-Start Performance and Body Form in 7 Species of Teleost Fish. Journal of experimental biology 74: 211–226.
      • 13. Langerhans RB (2010) Predicting evolution with generalized models of divergent selection: A case study with poeciliid fish. Integrative and Comparative Biology 50: 1167–1184.
      • 14. Domenici P (2010) Context-dependent variability in the components of fish escape response: integrating locomotor performance and behavior. J Exp Zool 313A: 59–79.
      • 15. Walker JA, Ghalambor CK, Griset OL, McKenney D, Reznick DN (2005) Do faster starts increase the probability of evading predators? Functional Ecology 19: 808–815.
      • 16. Fuiman LA, Rose KA, Cowan JH, Smith EP (2006) Survival skills required for predator evasion by fish larvae and their relation to laboratory measures of performance. Animal Behaviour 71: 1389–1399.
      • 17. Eaton RC, Lee RKK, Foreman MB (2001) The Mauthner cell and other identified neurons of the brainstem escape network of fish. Progress in Neurobiology 63: 467–485.
      • 18. Stefanelli A (1980) I neuroni di Mauthner degli Ittiopsidi. Valutazioni comparative morfologiche e funzionali. Lincei Mem Sci Fis Natur XVI: 1–45.
      • 19. Kohashi T, Oda Y (2008) Initiation of Mauthner- or Non-Mauthner-Mediated Fast Escape Evoked by Different Modes of Sensory Input. Journal of Neuroscience 28: 10641–10653.
      • 20. Domenici P, Batty RS (1997) Escape behaviour of solitary herring (Clupea harengus) and comparisons with schooling individuals. Marine Biology (Berlin) 128: 29–38.
      • 21. Marras S, Killen SS, Claireaux G, Domenici P, McKenzie DJ (2011) Behavioural and kinematic components of the fast-start escape response in fish: Individual variation and temporal repeatability. Journal of Experimental Biology 214: 3102–3110.
      • 22. Nagy M, Akos Z, Biro D, Vicsek T (2010) Hierarchical group dynamics in pigeon flocks. Nature 464: 890–893.
      • 23. Sumpter DJT (2010) Collective animal behavior. Princeton, NJ: Princeton University Press.
        • 24. Ward AJW, Sumpter DJT, Couzin ID, Hart PJB, Krause J (2008) Quorum decision-making facilitates information transfer in fish shoals. Proceedings of the National Academy of Sciences of the United States of America 105: 6948–6953.
        • 25. Herbert-Read JE, Perna A, Mann RP, Schaerf TM, Sumpter DJT, et al. (2011) Inferring the rules of interaction of shoaling fish. Proceedings of the National Academy of Sciences of the United States of America 108: 18726–18731.
        • 26. Robert M, Dagorn L, Lopez J, Moreno G, Deneubourg J-L (2013) Does social behavior influence the dynamics of aggregations formed by tropical tunas around floating objects? An experimental approach. Journal of Experimental Marine Biology and Ecology 440: 238–243.
        • 27. Seghers BH (1981) Facultative Schooling Behavior in the Spottail Shiner (Notropis hudsonius): Possible Costs and Benefits. Environmental Biology of Fishes 6: 21–24.
        • 28. Godin JGJ, Morgan MJ (1985) Predator Avoidance and School Size in a Cyprinodontid Fish, the Banded Killifish (Fundulus diaphanus Lesueur). Behavioral Ecology and Sociobiology 16: 105–110.
        • 29. deCarvalho CD, Corneta CM, Uieda VS (2007) Schooling behavior of Mugil curema (Perciformes: Mugilidae) in an estuary in southeastern Brazil. Neotropical Ichthyology 5: 81–83.
        • 30. Cosolo M, Ferrero E, Sponza S (2010) Prey ecology and behaviour affect foraging strategies in the Great Cormorant. Marine Biology (Berlin) 157: 2533–2544.
        • 31. Lefrancois C, Domenici P (2006) Locomotor kinematics and behaviour in the escape response of European sea bass, Dicentrarchus labrax L., exposed to hypoxia. Marine Biology (Berlin) 149: 969–977.
        • 32. Dadda M, Koolhaas WH, Domenici P (2010) Behavioural asymmetry affects escape performance in a teleost fish. Biology Letters 6: 414–417.
        • 33. Webb PW (1976) Effect of Size on Fast-Start Performance of Rainbow Trout Salmo gairdneri, and a Consideration of Piscivorous Predator-Prey Interactions. Journal of Experimental Biology 65: 157–177.
        • 34. Lefrançois C, Shingles A, Domenici P (2005) The effect of hypoxia on locomotor performance and behaviour during escape in Liza aurata. Journal of Fish Biology 67: 1712–1730.
        • 35. Barelli C, Boesch C, Heistermann M, Reichard UH (2008) Female white-handed gibbons (Hylobates lar) lead group movements and have priority of access to food resources. Behaviour 145: 965–981.
        • 36. Hunter JR (1969) Communication of velocity changes in Jack Mackerel (Trachurus symmetricus) schools. Animal Behaviour 17: 507–514.
        • 37. Couzin ID, Krause J, Franks NR, Levin SA (2005) Effective leadership and decision-making in animal groups on the move. Nature 433: 513–516.
        • 38. Treherne JE, Foster WA (1981) Group transmission of predator avoidance in a marine insect: the Trafalgar Effect. Animal Behaviour 29: 911–917.
        • 39. Cavagna A, Cimarelli A, Giardina I, Parisi G, Santagati R, et al. (2010) Scale-free correlations in starling flocks. Proceedings of the National Academy of Sciences of the United States of America 107: 11865–11870.
        • 40. Harcourt JL, Sweetman G, Manica A, Johnstone RA (2010) Pairs of fish resolve conflicts over coordinated movement by taking turns. Current Biology 20: 156–160.
        • 41. Reebs SG (2000) Can a minority of informed leaders determine the foraging movements of a fish shoal? Animal Behaviour 59: 403–409.
        • 42. Makris NC, Ratilal P, Jagannathan S, Gong Z, Andrews M, et al. (2009) Critical population density triggers rapid formation of vast oceanic fish shoals. Science 323: 1734–1737.
        • 43. Parrish JK, Viscido SV, Grünbaum D (2002) Self-Organized Fish Schools: An Examination of Emergent Properties. Biological Bulletin (Woods Hole) 202: 296–305.
        • 44. Domenici P (2010) Escape responses in fish: Kinematics, performance and behavior. In: Domenici P, Kapoor BG, editors. Fish Locomotion: An eco-ethological Perspective: Scientific Publishers.
          • 45. Eaton RC, Hackett JT (1984) The role of Mauthner cells in faststarts involving escape in teleost fish. In: Eaton RC, editor. Neural mechanisms of startle behavior. New York: Plenum Press.
            • 46. Jones KA, Godin JGJ (2010) Are fast explorers slow reactors? Linking personality type and anti-predator behaviour. Proceedings of the Royal Society Biological Sciences Series B 277: 625–632.
            • 47. Dill LM (1990) Distance-to-cover and the escape decisions of an african cichlid fish, Melanochromis chipokae. Environmental Biology of Fishes 27: 147–152.
            • 48. Bumann D, Krause J, Rubenstein D (1997) Mortality risk of spatial positions in animal groups: The danger of being in the front. Behaviour 134: 1063–1076.
            • 49. Bisazza A, Dadda M (2005) Enhanced schooling performance in lateralized fishes. Proceedings of the Royal Society Biological Sciences Series B 272: 1677–1681.
            • 50. Nespolo RF, Franco M (2007) Whole-animal metabolic rate is a repeatable trait: a meta-analysis. Journal of Experimental Biology 210: 2000–20005.
            • 51. Marras S, Claireaux G, McKenzie DJ, Nelson JA (2010) Individual variation and repeatability in aerobic and anaerobic swimming performance of European sea bass, Dicentrarchus labrax. Journal of Experimental Biology 213: 26–32.
            • 52. Herskin J, Steffensen JF (1998) Energy savings in sea bass swimming in a school: measurements of tail beat frequency and oxygen consumption at different swimming speeds. Journal of Fish Biology 53: 366–376.