The ability to detect and respond to the presence of predation risk is under intense selection, especially for small-bodied fishes that coexist with predators. Fish use visual, olfactory, and auditory cues to assess predation risk. Damselfishes (Pomacentridae) use auditory vocalizations during inter- and intrasexual interactions, but it is not known if they can use vocalizations in the context of predator-prey interactions. Here, we test if yellowtail damselfish, Chrysiptera parasema, can learn to associate the territorial vocalization of heterospecific humbug damselfish Dascyllus aruanus with predation risk. In conditioning trials of yellowtail damselfish we played the territorial call of humbug damselfish while introducing either blank water (control treatment) or chemical alarm cue derived from damaged skin of conspecific yellowtail damselfish. In conditioning trials, fish exposed to alarm cue increased activity and spent more time in the water column relative to fish that received the control treatment. After a single conditioning trial, conditioned fish were exposed again to the territorial call. Fish conditioned with the call + alarm cue increased activity and time in the water column relative to fish that had been conditioned with the control treatment. These data indicate associative learning of an auditory stimulus with predation risk in a species that regularly uses auditory signaling in other contexts. Recordings of conditioning and test trials failed to detect any acoustic calls produced by test fish in response to the perception of predation risk. Thus, although yellowtail damselfish can associate risk with auditory stimuli, we found no evidence that they produce an alarm call.

Study animals and experimental apparatus        

Yellowtail damselfish, Chrysiptera parasema, were purchased from a commercial supplier for the ornamental fish trade. Fish were acquired in two batches, shipped from the wholesaler 74 days apart. Individual damselfish were set up in 37-L glass-bottom aquaria filled with salt water made from reverse osmosis deionized water and Instant Ocean® marine salt adjusted to a salinity of 35 ppt. Fish were kept at 26° C. and maintained on a diet of commercial flake food and thawed adult brine shrimp. Each tank was equipped with an air-powered sponge filter, heater, and a shelter constructed from a 108 x 108 mm ceramic tile supported by legs 37 mm long on each corner. A length of airline tubing used for delivering test stimuli was affixed to the edge of each tank and terminated inside the rigid plastic tube through which the rising column of air and water exit from the air-powered sponge filter. These currents quickly dispersed test stimuli throughout the test aquarium. Injection tubing was replaced after every trial. The front pane of each aquarium had a 5 x 5 cm grid drawn on the outside of the tank used for scoring behaviors, described below. Test fish were left to acclimate to test aquaria for 24 h before testing.

Alarm cue preparation

We anesthetized 18 yellowtail damselfish (mean ± 1SE = 38.6 ± 1.0 mm) by immersion in ice water then euthanized them by cervical dislocation with a razor blade. We used ice instead of a chemical anesthetizing agent to avoid contamination of chemical contents of alarm cue derived from skin extract (Achtymichuk et al. 2022). The 18 carcasses were homogenized in 100 mL of reverse osmosis deionized (RODI) water for 60 s using an immersion blender. The resulting solution was diluted with additional RODI water to a final volume of 180 mL. The solution was spun in a centrifuge at 3000 rpm for 4 min. We decanted off the supernatant into a separate beaker and added an additional 25 mL of RODI water to compensate for liquid retained in the pellet to bring the volume of alarm cue solution to 180 mL. The alarm cue solution was aliquoted into 18, 10-mL doses and frozen at -20° C until needed. We prepared 18 10-mL aliquots of blank RODI water to serve as cue for the control treatment. Water cue was frozen at -20° C until needed.

Acoustic stimulus playback and audio recording of trials

We acquired a recording of a territorial call of humbug damselfish from https://www.youtube.com/watch?v=qnxl9ka_qqA, isolated a single call, removed background noise, and saved it in mp3 format using Audacity 3.2.5 open-source software (Muse Group, Boston, MA, 2023) on a laptop (Hewlett Packard Elitebook). A USB-to-3.5 mm audio jack adapter was used to connect the laptop audio output to a speaker, and audio input from a hydrophone. Fish calls were passed through an audio amplifier (Kinter K3118) and played through a speaker (8 ohm, 5 cm diameter) taped to the outside surface of the tank in the center of the end panel of the aquarium. Vibrations from the speaker passed through the aquarium glass into the water. Sound stimuli detected by a hydrophone (Aquarian Audio and Scientific, model H1a) were amplified using an instrumentation amplifier (Land, 2017) with gain = 100 and a low pass filter cutoff frequency set at 5 kHz. The signal was digitized (sampling frequency 44 kHz) and recorded by Audacity. The hydrophone was centered on the long side of the aquarium. The distance between the hydrophone and the speaker was (28 ± 1) cm.

Variables that influence the detected sound amplitude (amplifier gains, filter settings, Audacity microphone gain, speaker position, hydrophone position) were calibrated before trials began and kept fixed for all trials. We verified that the fish call was above the damselfish threshold of hearing (100 dB re: 1μPa, Parmentier et al., 2016) when played through the system. With no fish in the aquarium, the damselfish call played through the speaker was recorded by the hydrophone. A spectrogram (Sonic Visualizer 4.5.1, window = 32,768) was generated from the recording, and a python program was used to create a calibrated spectrogram. The first 60 s (when the damselfish call plays) shows frequency components well above the detection threshold for damselfish and distinguishable from background noise.

Experimental protocol

Each fish was tested twice; first in a conditioning trial and then again in a test trial. We alternated between alarm cue and water cue treatments for each consecutive trial to ensure that external factors such as time in captivity, shelf position in the lab, random disturbances in the research facility, or fish batch, were not spuriously correlated with cue treatment. To measure behavioral responses in conditioning trials, we recorded 5 min of pre-stimulus behavior, followed immediately by introduction of either alarm cue or water (control), and played the recording of the humbug damselfish. The auditory stimulus was played for 60 s (eight repetitions of the recorded call) while chemical stimuli were being introduced. Stimulus injection required less than 30 s, typically about 10 s. A 5-min post-stimulus observation period began immediately after the cessation of stimulus delivery.

We recorded vertical distribution at 10-s intervals as the row in the grid occupied by the fish’s eye. We recorded activity as the total number of grid lines crossed over the 5-min observation period. Shelter use was the cumulative number of seconds that the fish spent under the shelter. Antipredator behavioral responses typically manifest as a reduction in activity, movement out of the water column, and increased use of shelter. However, for some species, exposure to alarm cues in confined lab aquaria results in an increase in activity and movement into the water column because test fish attempt to flee the area where risk has been detected (Wisenden, 2011). In addition to recording behavioral responses, we made an audio recording of the entire trial (i.e., pre-stimulus, stimulus injection, and post-stimulus periods) to detect any evidence of vocalizations of focal fish during the trial in response to conspecific chemical alarm cue or to the recording of the heterospecific territorial call. We conducted 15 trials in which test fish received chemical alarm cues + territorial call playback, and 15 trials in which fish received blank water (control) + territorial call playback. When conditioning trials for any given day were completed, tanks were drained by siphon to a depth of 3 cm then refilled with fresh, pre-warmed salt water.

After a minimum of 24 h, fish were retested with the auditory stimulus alone. Using the same protocol described above, we recorded vertical distribution, activity and shelter use for 5 min before and after the acoustic stimulus was presented. We again made audio recordings inside each tank to detect the presence of vocalizations by the focal fish in response to test stimuli.