Navigating the risk of predation is a major driver of behavioral decision-making in small fishes. Fish use personal information from olfactory and visual indicators of risk, and also rely upon social cues to inform behavioral trade-offs between risk avoidance and fitness-positive activities such as foraging. Here, fathead minnows (Pimephales promelas), were captured, clipped, and released at 48 field sites chemically labeled with either fathead minnow alarm cue (high risk) or water (low risk). We removed the chemical label after 2 h, then monitored area use by clipped and non-clipped fish. In addition, a shoal was placed in traps in half of the risky and half of the safe locations as a visual social cue of safety. We caught 2919 fish in the first sample, of which 594 were fathead minnows. These were clipped and released. The second sample caught 1500 fish, of which 164 were fathead minnows including 11 bearing marks from the first sample. Fathead minnows and northern redbelly dace in the general community, which lacked personal information about risk status associated with trap sites, avoided areas previously labeled with alarm cues for at least 2 h after the source of alarm cue was removed, unless an experimental shoal was present at the risky site, in which case they joined the shoal in the trap. Clipped fathead minnows with direct personal knowledge of risk showed a significant shift away from areas labeled with conspecific alarm cues and a significant attraction toward sites seeded with a shoal. Moreover, unlike fish in the general community, clipped fathead minnows were not influenced by experimental shoals at sites previously labeled as risky. These data indicate that the influence of social cues on safety depends on whether individual minnows have access to recent personal information about risk assessment.

Preparation of chemical stimuli

Chemical alarm cues were collected from 10 adult fathead minnows (mean ± 1 SE total length = 65.5 ± 1.3 mm) collected by seine net from Deming Lake a few days before the field experiment. Fish were euthanized by an overdose of methane tricaine sulfate (500 mg/L MS222) and cervical dislocation (University of Minnesota IACUC protocol #2103-38900A). We measured the total length and then removed the skin from each side of the fish. The length and height of each skin fillet were measured before placing the fillets into a beaker of reverse-osmosis deionized (RODI) water resting on a bed of crushed ice. Keeping the solution cold during skin collection minimized the biochemical degradation of alarm cues. A total of 61 cm² was collected. We used an immersive blender to homogenize skin fillets in two bouts of 30 s to simulate predation by rupturing epidermal cells. The resulting homogenate was filtered through cheesecloth to remove connective tissue and then diluted to a final volume of 480 mL with RODI water. Skin extract solution was infused into 24 blocks of cellulose sponge (38 x 36 x 50 mm). Each sponge block received 20 mL of skin extract solution, representing the equivalent of about 2.5 cm² of fathead minnow skin per sponge. An additional 24 sponges were prepared and infused with 20 mL of blank RODI water to serve as a control. All sponges were placed in labeled bags and frozen at -20 °C until needed.

Shoal treatment

Shoals comprised five fish, three fathead minnows, and two blacknose shiners. The initial plan to use a shoal of five fathead minnows was modified on the day of the experiment because we could not collect a sufficient number of fathead minnows to apportion five fathead minnows per shoal. Earlier work has demonstrated that fathead minnows respond to shoals of conspecifics (Wisenden et al. 2003) and to shoals of blacknose shiners (Wisenden et al. 2023). Fathead minnows and blacknose shiners used for test shoals were collected by seine net from Deming Lake the day before testing. These fish were held overnight at the University of Minnesota Itasca Biological Station and Laboratories.

Experimental protocol

On May 25, 2023, we placed 48 unbaited Gee’s minnow traps around the perimeter of Deming Lake, placed approximately 10 m apart. Previous work demonstrated that traps placed more than 8 m apart do not influence each other (Wisenden 2008). Traps were set in groups of eight to provide for two replicates of the four treatment combinations (Alarm cue – shoal, Water – shoal, Alarm cue – no shoal, Water – no shoal) to control for spatial and temporal variation in littoral habitat (Fig. 1A, Table 1). All eight traps within each grouping were set simultaneously, approximately 1 – 2 m from shore in a depth ranging from 0.3 – 1.5 m, and secured to riparian vegetation with rope. After setting each group, we waited 15 min before setting the next group of eight traps. In this way, six groups comprising a total of 48 traps were set over 75 min. We pulled the first block of traps at the 120-minute mark giving ourselves 15 min to process the catch before the traps in the next block had been fishing for 120 min. In this way, fishing time for all traps was held constant at 120 min. If excessive processing time required delaying the pulling of the next group of traps (which was never more than a few minutes), the fishing time of all treatments was affected equally.

At the 120-min mark, we pulled each trap, disconnected the clip from the minnow trap, attached the clip to a weighted sponge (Fig. 1B) containing either a frozen alarm cue or blank water control, and placed the sponge in the lake in the same location the trap had been. For each trap, we counted the number of fish caught by species. These data represent the “pre-sponge” use of each area. Fathead minnows received one of four fin clips, each representing one treatment combination (Fig. 1A, Table 1). All fish were returned to the lake directly over the sponge to ensure that they detected the chemical cue being released by the sponge.

At the 240-minute mark, we returned to the first group of traps, removed the sponges reattached, and reset the trap in the same location. Traps now contained a 973-mL glass jar (Fig. 1B) with either a mixed-species shoal of three fathead minnows and two blacknose shiners (shoal treatment) or nothing (control). Five fish fit into a 973-mL jar at natural shoaling density and a shoal of five has successfully attracted fathead minnows into traps in earlier experiments (Wisenden et al. 1995b, 2003, 2023). Fathead minnows and blacknose shiners form mixed-species shoals and earlier work has shown that fathead minnows are drawn into traps by the presence of a five-fish shoal if the shoal is of fathead minnows (Wisenden et al. 2003) or blacknose shiners (Wisenden et al. 2023). The jar lids were replaced with window screen mesh to allow the free exchange of lake water into and out of the jar. We reset the remainder of the traps in groups of eight with jars at 15-minute intervals.

At the 360-minute mark, we began pulling traps in groups of eight at 15-minute intervals, as before. We counted the number of fish caught by species and inspected fathead minnows for the presence of fin clips applied to fish caught in the first sample. These data represent the first post-sponge use of areas previously labeled by sponges as either risky or safe. All captured fish were immediately returned to the lake and the traps and jars were reset for a second post-sponge measure of area use. At the 480-min mark, we pulled the traps, counted fish by species, and noted the presence of clipped fathead minnows.

The mean fishing time for traps was 126 ± 2.3 min (n = six groups of eight traps) for the pre-sponge sample, sponges were in the water for 121 ± 0.8 min before being removed, traps for the first post-sponge sample were in the lake for 122 ± 1.6 min, and traps for the second post-sponge sample were left in the water for a mean of 106 ± 3.0 min.

Statistical analysis

Catch data were not normally distributed (Kolmogorov-Smirnoff tests, p < 0.05). We rank-transformed catch data for each species across the 48 traps within each sample. Separate rankings were performed for each sample. We then tested for independent and interactive effects of the chemical label by sponge (alarm cue or water) and presence of a shoal (shoal or no shoal) using separate factorial ANOVAs for fathead minnows, redbelly dace, blacknosed shiners pumpkinseed sunfish, and brook stickleback. The effect of chemical information (sponge treatment) and the effect of visual information (jar shoal treatment) on recapture rates of clipped fathead minnows were tested using a binomial test. Data were analyzed using SPSS® v 28.0.