Skip to main content
Dryad logo

Predation risk shapes the use of conflicting personal risk and social safety information in guppies


Feyten, Laurence E. A.; Crane, Adam L.; Ramnarine, Indar W.; Brown, Grant E. (2021), Predation risk shapes the use of conflicting personal risk and social safety information in guppies, Dryad, Dataset,


When faced with uncertainty, animals can benefit from using multiple sources of information in order to make an optimal decision. However, information sources (e.g., social and personal cues) may conflict, while also varying in acquisition cost and reliability. Here, we assessed behavioral decisions of Trinidadian guppies (Poecilia reticulata), in situ, when presented with conflicting social and personal information about predation risk. We positioned foraging arenas within high- and low-predation streams, where guppies were exposed to a personal cue in the form of conspecific alarm cues (a known indicator of risk), a novel cue, or a control. At the same time, a conspecific shoal (a social safety cue) was either present or absent. When social safety was absent, guppies in both populations showed typical avoidance responses towards alarm cues, and high-predation guppies showed their typical avoidance of novel cues (i.e., neophobia). However, the presence of social safety cues was persuasive, overriding the neophobia of high-predation guppies and emboldening low-predation guppies to ignore alarm cues. Our experiment is one of the first to empirically assess the use of safety and risk cues in prey, and suggests a threshold level of ambient risk which dictates the use of conflicting social and personal information.


Data was collected in March 2019 in the Lopinot and Upper Aripo Rivers of the Northern Range Mountains of Trinidad. 

For Experiment 1, we conducted field observations by gently placing a novel foraging arena (corrugated plastic; ~30x25x23 cm) into stream pools at locations that visibly contained at least 5 guppies in locations without visible currents or higher velocities. For half of the trials, the social safety cue was present, consisting of 3 female guppies inside a clear 1-L chamber (~7x7x24 cm) filled with stream water and gravel (~3 cm depth). For the other half of trials, the chamber was filled with water and gravel but did not contain guppies. We began our trials by gently injecting 10 mL of one of the three personal cue treatments (alarm cue, novel cue, or stream water) through airline tubing on the back wall of the arena before flushing the tubing with 60 mL of stream water and beginning a 5-min observation period. We recorded the latency for guppies to enter the foraging zone (crossing from the approach zone into the foraging zone), as well as the latency to forage inside the foraging zone. These two latencies could correspond to two different guppies. After each trial, we moved at least 1 m upstream before initiating the next trial, which ensured we avoided re-testing guppies or testing downstream guppies potentially exposed to the experimental cues from previous trials. The cue treatments were randomly assigned, and we conducted a total of 90 trials per population (180 total), with a sample size of 15 per treatment group. 

For Experiment 2, we conducted a behavioral assay to determine whether the movement activity of our experimental shoals differed between the high- and low-predation streams (30 trials from each population). From each stream, we collected 90 female guppies in total, moving 3 individuals at a time into 1-L chambers that were positioned on a flat surface along the stream bank. The chamber again contained stream water (0.75 L) and gravel, was surrounded by opaque walls on 3 sides to minimize external visual stimuli, and was demarcated into 4 quadrants by exterior lines. After a 1-min acclimation period, we tallied the total number of lines crossed by all 3 guppies over a 1-min observation period. We also recorded whether the line crosses were dashes (rapid bursts of movement with the entire body crossing the line in under 0.5 s) or were calmer movements that are typical of normal swimming behavior (> 0.5 s). After each trial, we recorded the standard length (head to caudal peduncle) of each fish and released them at their site of capture. 

All analyses were conducted in RStudio 3.4.1 using α = 0.05.

Full Methodology can be found in Feyten et al. (2021), Behavioral Ecology.

Usage Notes

We have attached a ReadMe file for usage notes and variable lists.


Natural Sciences and Engineering Research Council of Canada, Award: Discovery Grant