Research ArticlePREDATION

High-predation habitats affect the social dynamics of collective exploration in a shoaling fish

See allHide authors and affiliations

Science Advances  03 May 2017:
Vol. 3, no. 5, e1602682
DOI: 10.1126/sciadv.1602682
  • Fig. 1 Experimental design and number of movements between the arms of the maze.

    (A) A zoomed-in still from movie S1 of a fish (yellow) initiating a movement into an empty arm and leading the three other fish. (B) The effect of predation risk in the source habitat on the total number of movements between arms per individual fish. The median is shown by the solid line. The interquartile range is enclosed by the box. The whiskers extend to the most extreme data point within 1.5× the interquartile range outside the box, and empty circles show data points beyond the range of the whiskers. Predation in the source habitat had no effect on average levels of activity [linear mixed model (LMM): F = 0.27, P = 0.61], but individuals were less variable when they were from high-predation habitats [negative binomial generalized linear mixed model (GLMM) on model residuals: χ2 = 6.70, P = 0.0098]. This reduced variability in fish from high-predation rivers was confirmed with Levene’s test (F2,307 = 9.68, P = 8.41 × 10−5).

  • Fig. 2 Frequency distribution of the number of initiations versus the number of follows made by each individual fish.

    Individuals are paneled by the predation risk in their source habitat: low (A), medium (B), and high (C). The color scale indicates the proportion of fish from each level of predation with each combination of the number of initiations and the number of follows, so that the total proportion is equal to 1 in each panel. The negative relationship between initiations and follows is greatest in fish from high-predation habitats, in terms of both the slope of the relationship (negative binomial GLMM predation level × number of initiations: χ2 = 5.83, P = 0.016) and the correlation (Spearman’s rank correlation: low: rs = -0.24, n = 84, P = 0.026; medium: rs = −0.11, n = 108, P = 0.24; high: rs = −0.54, n = 118, P = 1.78 × 10−10).

  • Fig. 3 Observed diversity and maximum and minimum number of initiations per group compared to a randomized group membership.

    The test statistic [COV (A), maximum (B), and minimum (C)] is calculated for each group tested (the mean for each predation level is shown as red asterisks) and plotted against the distribution of the mean of that statistic from 10,000 iterations of a randomization that reforms the groups with random membership within each predation level and group size (due to the significant effect group size on the number of initiations). The means (filled squares) and 95% confidence intervals (error bars) of the randomized distributions are shown. The observed group-level statistic falling outside of the 95% confidence intervals serves as evidence that the composition of the groups in the experiment is significantly different from groups being randomly assembled from the fish tested.

  • Fig. 4 Relationship between diversity in the number of initiations (COV) and group cohesion.

    Cohesion is calculated as the number of other fish in the same arm/maximum possible number of other fish, averaged over all time frames and across the individuals in a group. Data are paneled by sex [females (A to C) and males (D to F)] and predation risk [low (A and D), medium (B and E), and high (C and F)]. Fitted lines are calculated from the GLMM fixed-effect coefficients. Circles and dark green lines represent groups of two fish, and crosses and red lines represent groups of four fish.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/3/5/e1602682/DC1

    fig. S1. Relationships between measures of activity and cohesion in this study versus those more typically used, wherein animals are unconstrained by a maze.

    fig. S2. Following behavior in the trials.

    fig. S3. The effect of predation risk in the source habitat on the number of initiations and number of follows per individual fish.

    fig. S4. Relationship between the number of initiations made by each fish in the first versus second half of each trial.

    fig. S5. Relationship between the number of initiations made by two fish in groups from each predation level.

    fig. S6. Frequency of using each arm for initiations and follows by fish from habitats with different levels of predation risk.

    fig. S7. Effect of applying a threshold based on the level of uncertainty of whether a segment of track belongs to a particular fish.

    table S1. Populations, their level of predation, and the location where they were sampled.

    table S2. Full statistical results for linear models including sample sizes.

    table S3. Summary of sample sizes.

    movie S1. Example of group decisions and collective movement of guppies in the three-armed maze.

    data file S1. Data for individual fish behavior, aggregated over each trial.

    data file S2. Data for individual fish behavior, aggregated over the first and second halves of the trials.

  • Supplementary Materials

    This PDF file includes:

    • fig. S1. Relationships between measures of activity and cohesion in this study versus those more typically used, wherein animals are unconstrained by a maze.
    • fig. S2. Following behavior in the trials.
    • fig. S3. The effect of predation risk in the source habitat on the number of initiations and number of follows per individual fish.
    • fig. S4. Relationship between the number of initiations made by each fish in the first versus second half of each trial.
    • fig. S5. Relationship between the number of initiations made by two fish in groups from each predation level.
    • fig. S6. Frequency of using each arm for initiations and follows by fish from habitats with different levels of predation risk.
    • fig. S7. Effect of applying a threshold based on the level of uncertainty of whether a segment of track belongs to a particular fish.
    • table S1. Populations, their level of predation, and the location where they were sampled.
    • table S2. Full statistical results for linear models including sample sizes.
    • table S3. Summary of sample sizes.
    • Legend for movie S1
    • Legends for data files S1 and S2

    Download PDF

    Other Supplementary Material for this manuscript includes the following:

    • movie S1 (.mov format). Example of group decisions and collective movement of guppies in the three-armed maze.
    • data file S1 (Microsoft Excel format). Data for individual fish behavior, aggregated over each trial.
    • data file S2 (Microsoft Excel format). Data for individual fish behavior, aggregated over the first and second halves of the trials.

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article