Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Ring a bell? Variable recognition of a multicomponent auditory stimulus associated with predation risk by zebrafish responding to full and partial matches

Ring a bell? Variable recognition of a multicomponent auditory stimulus associated with predation... 1.IntroductionAnimals detect and attend to information about the environment to inform behavioural decision-making for navigation, habitat selection, foraging, predator avoidance, and reproduction (Dill, 1987; Dukas, 1998; Shettleworth, 2001; Dall et al., 2005; Webster & Laland, 2017; Bshary & Triki, 2022). Many fitness functions form trade-offs against one another (e.g., risk-sensitive foraging) and weighing relative costs and benefits of any decision is complicated by information that is often incomplete, unreliable, variable or of limited temporal relevance (Lima & Bednekoff, 1999; Sih et al., 2000; Ferrari et al., 2016).Information about predation risk obtained directly through encounters with predators is inherently risky. Therefore, there is a large fitness benefit to detecting risk indirectly through publicly available cues, such as the signature odor of a predator, which can be detected from afar allowing prey to evade detection (Lima & Dill, 1990; Smith, 1992; Wisenden & Chivers, 2006; Wisenden et al., 2010). Similarly, alarm cues released when prey tissues are damaged during a predator attack are reliable indicators of an active predator nearby (Ferrari et al., 2010a). Moreover, alarm cues act as a releaser of associative learning to transfer perception of risk from alarm cues to the novel stimulus (Magurran, 1989; Suboski, 1990; Suboski et al., 1990; Mathis & Smith, 1993; Chivers & Smith, 1994, 1995; see Kelley & Magurran, 2003, Ferrari et al., 2010a for reviews).Predator risk cues, however, are expected to elicit prey responses in direct proportion to the level of risk. A predator risk threshold exists above which a response is appropriate in response to imminent danger. When the predator risk level is relatively low, ignoring the risk cue may be appropriate, allowing prey to invest energy and/or time on other life activities (Ferrari et al., 2008).How fish decide whether or not an indicator of risk surpasses the threshold of mounting a behavioral response is ideally done using quantifiable test stimuli. Auditory stimuli, for example, provide a convenient modality that can be quantified precisely, recorded and presented in standardized ways that allow for experimental tests of learning and risk assessment.Many fishes have well-developed sensory receptors for detecting auditory stimuli (Ladich, 2019). For example, the Otophysi is a speciose group of hearing specialists comprising approximately 67% of all freshwater fish species (Nakatani et al., 2011). The Otophysi possess modified ribs called Weberian ossicles that link the gas bladder to the inner ear allowing these fishes to detect a wide range of sound stimuli (Kasumyan, 2008; Higgs et al., 2011; Ladich, 2019; Fine & Parmentier, 2022; Sprague et al., 2022). Fathead minnows (Pimephales promelas), glowlight tetras (Hemigrammus erythrozonus) and zebrafish (Danio rerio) (all otophysans) can associate novel auditory stimuli with risk of predation via releaser-induced recognition learning (Wisenden et al., 2008; Seigel et al., 2021).Seigel et al. (2021) showed that zebrafish conditioned to fear one tone responded when that tone was replayed but did not respond when a tone of a different frequency was played, suggesting that recognition of an auditory stimulus was frequency-specific and not generalized to all tones. This raised the question of whether or not Otophysan fish can associate risk with multicomponent auditory stimuli, and if risk is associated with each individual component or if risk is associated only with the sum of all components. Here, we report how we conditioned zebrafish to fear a multicomponent auditory stimulus (a three-note chord) followed by tests of their ability to recognize and respond to playback stimuli when all three notes were replayed, when two of the original notes were replayed, or when only one of the original notes was replayed. In essence, we tested the response threshold of zebrafish to recognize and respond behaviourally to full or partial information about predation risk.2.Methods2.1.Animals and housingZebrafish were purchased from a commercial supplier (EkkWill Water Life Resources, Ruskin, FL, USA) and held in 75-litre holding tanks at 24°C on a 12:12 L:D cycle, with dechlorinated tap water filtered by hang-on-back power filters and fed commercial flake food.2.2.Preparation of chemical and auditory test cuesChemical alarm cue was prepared in two batches by euthanizing adult zebrafish with an overdose of methanetricaine sulfonate (MS222). The first batch used 49 zebrafish (mean ± SE total length = 38.3 ± 0.4 mm). The second batch used eight zebrafish (TL = 37.9 ± 0.5 mm). Fish carcasses were placed together in a single beaker with deionized water and homogenized using a hand blender. The homogenate was filtered through a loose wad of polyester fiber to remove connective tissue, and diluted with deionized water to produce a concentration of alarm cue equivalent to one fish per 10 ml. Alarm cue was aliquoted into 10-ml doses and frozen at –20°C until needed. Blank water control cue was prepared by freezing 10-ml aliquots of deionized water at –20°C until needed.We used the Audacity® application to generate tones C5 (523.25 Hz), E5 (659.25 Hz), and G5 (783.99 Hz) to create mpg files of (1) C, (2) C + G and (3) C + E + G, each of 5 min duration (the subscript ‘5’ refers to the 5th octave on a standard piano keyboard, i.e., one octave above middle C). Zebrafish can detect sounds from 100 Hz to 8000 Hz (Higgs et al., 2011) and in a previous study associated risk with a tone of 285 Hz (Seigel et al., 2021). The three sound files were downloaded to an iPhoneXR. Playback was conducted with volume set to maximum through waterproof earbud headphones (model PWPE10B, PyleUSA.com) suspended inside the test aquaria 12 cm from the tank bottom i.e., equidistant between the surface and bottom, in zone 3 of the grid drawn on the front pane. This method produced clearly discernible auditory stimuli, with each tone of equal amplitude (Figure 1).Figure 1.Power spectra for tones used for conditioning and playback trials using tones C5 (523.25 Hz), E5 (659.25 Hz), and G5 (783.99 Hz). Upper panel, C alone; middle panel, C + G; lower panel C + E + G.2.3.Power spectra analysis of auditory test stimuliThe auditory stimuli C, CG, and CEG were played through waterproof earbuds fixed at one end of a 37-litre aquarium (same size as used for test trials), and recorded with a hydrophone (kithub.cc) fixed at the opposite end. No aerator was in the aquarium during these tests in order to reduce the background noise level. The microphone output was amplified and filtered (gain = 1000, band pass filtering 1 Hz to 5 kHz) using an amplifier built in-house (Land et al., 2001). Open-source software (Audacity®, version 2.4.1) was used to produce a power spectrum of each tone combination (sampling time 10 s, sampling rate 44 kHz, spectrum size 131,072). The power spectra are shown in Figure 1. The sound level was taken relative to the acoustic background noise level. The relative amplitudes of the individual tones C, E and G were 45 ± 5 dB, 39.0 ± 0.7 dB and 39.7 ± 0.7 dB, respectively.2.4.Experimental protocolIndividual test fish were placed in 37-litre aquaria and left for at least 24 h to acclimate to the tank before testing. The front pane of the tank facing the aisle had a 5 × 5 cm grid drawn on it for scoring fish behaviour. Each tank was equipped with an air stone and a second length of airline tubing attached to the air stone to allow for injection and dispersal of test stimuli. Each tank had a small shelter object made from a ceramic tile 11 × 11 cm supported by legs 4 cm tall on each corner.Table 1.Experimental design using tones C5 (523.25 Hz), E5 (659.25 Hz) and G5 (783.99 Hz).The experiment occurred in two steps; a conditioning trial followed by a test trial (Table 1). All fish were conditioned with the multicomponent auditory stimulus C + E + G played simultaneously for 30 s during the stimulus-injection period between pre- and post-stimulus behavioral observations. The negative control treatment group received water while the three tones were being played. This treatment controlled for the potential of a pre-existing aversion to the CEG stimulus. The other three treatment groups received 10 ml of alarm cue solution while the three-note auditory stimulus was played. In all trials, fish activity (total number of grid lines crossed), vertical distribution (horizontal row of the grid, 1 = tank bottom, 5 = surface, occupied by the fish every 10 s), and use of the shelter object (total accumulated time under the shelter) was scored for 5 min before (pre-stimulus observations) and for 5 min after (post-stimulus observations) presentation of test stimuli. Each fish was used only once in a condition-test sequence. Injection hoses were replaced after each trial.After each conditioning trial was completed, tanks were drained and refilled with room-temperature dechlorinated tap water. After at least 24 h (24–168 h, mean ± SE = 67.2 ± 3.9 h), a test trial was conducted. The group conditioned with water + CEG stimuli was tested with the CEG stimulus for 30 s. Of the three treatment groups conditioned with alarm cue + CEG, one was tested with the CEG stimulus for 30 s, one group was tested with two of the three notes play (CG) for 30 s and one group was tested with one of the three notes (C) for 30 s. No chemical stimuli were presented during test trials.2.5.Statistical treatmentVariables for change in behaviour (post-stimulus–pre-stimulus) were not normally distributed (Kolmogorov-Smirnov test, p<0.05). Mann–Whitney U-tests were used to compare the behavioural responses of water-conditioned fish to alarm cue-conditioned fish in conditioning trials. Comparisons of behavioural responses among the four treatment groups in test trials were done by Kruskal-Wallis one-way ANOVAs followed by Bonferroni-corrected pairwise comparisons between treatment pairs. The software used for these analyses was SPSS v26.3.Results3.1.Conditioning trialsRelative to control trials that received blank water, zebrafish that received conspecific alarm cues reduced vertical distribution (U=167.5, N=59, p=0.005), trended strongly to reduce activity (U=223.5, N=59, p=0.064), and increased time under the shelter (U=502.5, N=59, p=0.003; Figure 2).Figure 2.Median ± quartiles change in behaviour by zebrafish in response to either blank water (open fill) or chemical alarm cues for conditioning trials (shaded fill).3.2.Test trialsResponse to acoustic stimuli in test trials differed by the completeness of the acoustic stimulus used in the test trial (CEG, CG or C). Treatment group had a significant effect on change in vertical distribution (KW = 10.374, df = 3, p=0.016; Figure 3) but not activity (KW = 0.915, p=0.822; Figure 4), or Shelter Use (KW = 4.669, df = 4, p=0.198; Figure 5).Figure 3.Median ± quartiles of change (post-pre) in vertical distribution (horizontal row of the grid) in response to tones in test trials. All fish were conditioned with a three-note chord (CEG) + either water (W) or conspecific alarm cue (A) and later tested with either all three notes (CEG), two of the three notes (CG) or only one note (C). Shared letters above bars are not statically different in Kruskal-Wallis post-hoc pairwise comparison tests among all four treatment groups (p>0.05).Figure 4.Median ± quartiles of change (post-pre) in activity (number of lines crossed) in response to tones in test trials. All fish were conditioned with a three-note chord CEG + either water (W) or conspecific alarm cue (A) and later tested with either all three notes (CEG), two of the three notes (CG) or only one note (C).Figure 5.Median ± quartiles of change (post-pre) in time under the shelter object in response to tones in test trials. All fish were conditioned with a three-note chord CEG + either water (W) or conspecific alarm cue (A) and later tested with either all three notes (CEG), two of the three notes (CG) or only one note (C).Table 2.Alpha-corrected p-values for post-hoc pairwise comparisons among treatment groups in test trials for change in vertical distribution.Pairwise comparisons of change in vertical distribution revealed that high variability within the water control treatment group caused it to overlap with the other three treatment groups (Table 2). For fish tested with the three-note chord, those conditioned with alarm cues + CEG trended toward a reduction in vertical distribution compared to those conditioned with water + CEG (U=157, N=59, p=0.065). For fish conditioned with alarm cues + CEG, response intensity in test trials to CEG did not differ from CG but response intensity to CEG and CG were each significantly more intense that the behavioural response to C alone (Table 2). These results indicate that zebrafish significantly reduced vertical distribution in response to the full three-note CEG and to the partial two-note CG compared to the one-note C stimulus.4.DiscussionZebrafish conditioned with alarm cues and a three-note auditory stimulus responded with antipredator behaviour when re-exposed to either the same three-note stimulus, or two of the three notes, relative to a playback stimulus containing only one of the three notes. These data corroborate earlier studies that demonstrated that fishes can associate risk with auditory stimuli (Wisenden et al., 2008; Seigel et al., 2021, 2022). Response intensity of learned responses were of lower intensity than behavioral responses during conditioning trials, manifest only as a significant change in vertical distribution but not change in activity or shelter use. In contrast, Seigel et al. (2021) showed significant changes in all three behavioural measures during test trials of zebrafish conditioned to a single tone. Nevertheless, these data are the first to show that recognition of novel auditory indicators of predation risk can occur in response to multicomponent auditory stimuli and that response intensity is sensitive to the completeness of the match between conditioned and test stimuli. As such, these data provide an example of how information uncertainty mediates the trade-off between risk assessment and other life tasks (Lima & Bednekoff, 1999; Sih et al., 2000; Ferrari et al., 2016).Information about the temporal and spatial distribution of risk confers fitness benefits to prey, but monitoring and updating this information incurs a cost in that the best information is available at times and places of greatest risk. Small fishes employ surreptitious olfactory sampling behaviours (“stealthy sniffing”) to monitor olfactory information while minimizing whole-body movements that are easily detected by predators. For example, darters and cichlids engage in opercular pumping (Hawkins et al., 2004; Gibson & Mathis, 2006; Bareto & Volpato, 2006; Bareto et al., 2012; Sanches et al., 2015) and minnows perform fin-flicking (Sutrisno et al., 2014) in response to alarm cues. Both of these behaviours increase the rate of water flow over chemoreceptors in the external nares. In lab and field studies, fish avoid areas labelled with chemical alarm cues released from damaged epidermal tissues of conspecifics but continue to sample risk from afar aided by semiochemicals carried on water currents (Wisenden et al., 1994, 1995, 2010). Alternatively, prey fish may approach the source of danger to acquire information about risk (Brown & Godin, 1999; Wisenden et al., 2010).The “predator recognition continuum hypothesis” predicts that responses to indicators of risk lie on a continuum from innate recognition mechanisms (no learning), to acquired recognition (no generalization) to generalization from known predators to general classes of predictors of risk (Ferrari et al., 2007). Generalization of antipredator responses occurs when a novel stimulus is either missing some of the components of the conditioned stimulus, and/or has novel components added, reducing the certainty of recognizing the novel stimulus as an indicator of risk, resulting in a response of lower intensity or no response at all (Griffin et al., 2001; Ferrari et al., 2007, 2009; Ferrari & Chivers, 2009; Brown et al., 2011; Chivers et al., 2012, 2014; Mitchell et al., 2013). In the current study, we did not add novel notes, but simply sequentially deleted them until test subjects were too uncertain that the test stimulus matched the conditioned stimulus, and therefore did not show a behavioural response. There are parallels between our current study and studies where visual matching of novel predator identity to a conditioned model is impaired by the presence of turbidity (Ferrari et al., 2010b; Chivers et al., 2012) or when chemical cues are diluted below response threshold (Lawrence & Smith, 1989; Mirza & Chivers, 2003, Mirza et al., 2006). However, in those studies response strength was reduced when stimulus amplitude was reduced, whereas in our study stimulus intensity was held constant but the presence of selected components were deleted. Future experiments with various combinations of deleted and added components would inform the cognitive decision-making thresholds to respond versus ignore.Zebrafish presented with a multicomponent stimulus of three notes in this study failed to respond when only a single note was replayed. This finding contrasts with how fathead minnows can learn to recognize the alarm cues of a heterospecific member of their prey guild (brook stickleback, Culaea inconstans) from a multicomponent olfactory cue that contains the odour of the predator (yellow perch, Perca flavescens) fed a mixed diet of fathead minnows and stickleback (Mirza & Chivers, 2001). In that study, test subjects were able to isolate and remember components of a mixed olfactory stimulus, associate each of them with risk, and later respond with antipredator behaviour when a subset of those chemicals were later presented on their own. There may well be an interaction between the sensory modality and the trade-off between the detect-respond/detect-ignore threshold. This topic is ripe for future work.Releaser-induced recognition learning, and generalization, enable prey to label novel stimuli as dangerous. However, animals must evade risk while simultaneously engaging in other potentially conflicting, fitness-positive activities such as foraging, social and reproductive behaviours. Thus, animals cannot afford to minimize risk; instead, they must manage risk. There is a growing literature dedicated to revealing how animals avoid over-responding to stimuli that may or may not indicate risk (Wisenden & Harter, 2001; Brown et al., 2011; Ferrari & Chivers, 2011; Chivers & Ferrari, 2013; Mitchell et al., 2015). Generalization windows may be broad if the diversity of predators is relatively small (Ferrari et al., 2007) or narrow if predator diversity is large (Mitchell et al., 2013). If the window for predator-recognition generalization is set too broadly, then prey risk responding to non-threatening predators. For example, velvet geckos (Oedura lesueurii) avoid all elapid snakes even though only some are dangerous (Webb et al., 2010). Conversely, a predator-recognition window that is too conservative may lead prey to fail to respond to a predator that poses real and imminent danger. For example, invasive predators can benefit from being outside of the generalized predator window of native prey (Goodchild & Stockwell, 2016; Anton et al., 2016, 2020; Carthey & Blumstein, 2018; Melotto et al., 2020; Stockwell et al., 2022). Similarly, prey can adjust response intensity to the strength of alarm cue or perception of risk represented by test stimuli (Mirza & Chivers, 2003, Mirza et al., 2006).Understanding of the role of sound in mediating fish behaviour is in its infancy (Rountree et al., 2019) but it is well established that there is a rich aquatic soundscape of acoustic signals used in the context of intrasexual male-male competition, and intersexual courtship displays (Putland et al., 2018; Ladich, 2019). In the context of predator–prey interactions, acoustic stimuli are used by prey to detect the presence of predators, by predators to detect prey (in response, prey silence their acoustic signals in the presence of risk) and production of sounds by prey when they are captured or held (Ladich, 2022). The present study reveals that acoustic stimuli may be a relatively unexplored dimension of risk assessment by fishes.In this study, we showed for the first time that fish can associate multicomponent auditory stimuli with predation risk, and respond when a novel stimulus contains all or most of the original components but not when the play back stimulus was missing most of the salient information in the original conditioning stimulus. This work underscores the potential for auditory stimuli as an experimental tool for testing cognitive mechanisms in risk assessment by fishes. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Behaviour Brill

Ring a bell? Variable recognition of a multicomponent auditory stimulus associated with predation risk by zebrafish responding to full and partial matches

Loading next page...
 
/lp/brill/ring-a-bell-variable-recognition-of-a-multicomponent-auditory-stimulus-UCIqBXi9tT

References (79)

Publisher
Brill
Copyright
Copyright © Koninklijke Brill NV, Leiden, The Netherlands
ISSN
0005-7959
eISSN
1568-539X
DOI
10.1163/1568539x-bja10210
Publisher site
See Article on Publisher Site

Abstract

1.IntroductionAnimals detect and attend to information about the environment to inform behavioural decision-making for navigation, habitat selection, foraging, predator avoidance, and reproduction (Dill, 1987; Dukas, 1998; Shettleworth, 2001; Dall et al., 2005; Webster & Laland, 2017; Bshary & Triki, 2022). Many fitness functions form trade-offs against one another (e.g., risk-sensitive foraging) and weighing relative costs and benefits of any decision is complicated by information that is often incomplete, unreliable, variable or of limited temporal relevance (Lima & Bednekoff, 1999; Sih et al., 2000; Ferrari et al., 2016).Information about predation risk obtained directly through encounters with predators is inherently risky. Therefore, there is a large fitness benefit to detecting risk indirectly through publicly available cues, such as the signature odor of a predator, which can be detected from afar allowing prey to evade detection (Lima & Dill, 1990; Smith, 1992; Wisenden & Chivers, 2006; Wisenden et al., 2010). Similarly, alarm cues released when prey tissues are damaged during a predator attack are reliable indicators of an active predator nearby (Ferrari et al., 2010a). Moreover, alarm cues act as a releaser of associative learning to transfer perception of risk from alarm cues to the novel stimulus (Magurran, 1989; Suboski, 1990; Suboski et al., 1990; Mathis & Smith, 1993; Chivers & Smith, 1994, 1995; see Kelley & Magurran, 2003, Ferrari et al., 2010a for reviews).Predator risk cues, however, are expected to elicit prey responses in direct proportion to the level of risk. A predator risk threshold exists above which a response is appropriate in response to imminent danger. When the predator risk level is relatively low, ignoring the risk cue may be appropriate, allowing prey to invest energy and/or time on other life activities (Ferrari et al., 2008).How fish decide whether or not an indicator of risk surpasses the threshold of mounting a behavioral response is ideally done using quantifiable test stimuli. Auditory stimuli, for example, provide a convenient modality that can be quantified precisely, recorded and presented in standardized ways that allow for experimental tests of learning and risk assessment.Many fishes have well-developed sensory receptors for detecting auditory stimuli (Ladich, 2019). For example, the Otophysi is a speciose group of hearing specialists comprising approximately 67% of all freshwater fish species (Nakatani et al., 2011). The Otophysi possess modified ribs called Weberian ossicles that link the gas bladder to the inner ear allowing these fishes to detect a wide range of sound stimuli (Kasumyan, 2008; Higgs et al., 2011; Ladich, 2019; Fine & Parmentier, 2022; Sprague et al., 2022). Fathead minnows (Pimephales promelas), glowlight tetras (Hemigrammus erythrozonus) and zebrafish (Danio rerio) (all otophysans) can associate novel auditory stimuli with risk of predation via releaser-induced recognition learning (Wisenden et al., 2008; Seigel et al., 2021).Seigel et al. (2021) showed that zebrafish conditioned to fear one tone responded when that tone was replayed but did not respond when a tone of a different frequency was played, suggesting that recognition of an auditory stimulus was frequency-specific and not generalized to all tones. This raised the question of whether or not Otophysan fish can associate risk with multicomponent auditory stimuli, and if risk is associated with each individual component or if risk is associated only with the sum of all components. Here, we report how we conditioned zebrafish to fear a multicomponent auditory stimulus (a three-note chord) followed by tests of their ability to recognize and respond to playback stimuli when all three notes were replayed, when two of the original notes were replayed, or when only one of the original notes was replayed. In essence, we tested the response threshold of zebrafish to recognize and respond behaviourally to full or partial information about predation risk.2.Methods2.1.Animals and housingZebrafish were purchased from a commercial supplier (EkkWill Water Life Resources, Ruskin, FL, USA) and held in 75-litre holding tanks at 24°C on a 12:12 L:D cycle, with dechlorinated tap water filtered by hang-on-back power filters and fed commercial flake food.2.2.Preparation of chemical and auditory test cuesChemical alarm cue was prepared in two batches by euthanizing adult zebrafish with an overdose of methanetricaine sulfonate (MS222). The first batch used 49 zebrafish (mean ± SE total length = 38.3 ± 0.4 mm). The second batch used eight zebrafish (TL = 37.9 ± 0.5 mm). Fish carcasses were placed together in a single beaker with deionized water and homogenized using a hand blender. The homogenate was filtered through a loose wad of polyester fiber to remove connective tissue, and diluted with deionized water to produce a concentration of alarm cue equivalent to one fish per 10 ml. Alarm cue was aliquoted into 10-ml doses and frozen at –20°C until needed. Blank water control cue was prepared by freezing 10-ml aliquots of deionized water at –20°C until needed.We used the Audacity® application to generate tones C5 (523.25 Hz), E5 (659.25 Hz), and G5 (783.99 Hz) to create mpg files of (1) C, (2) C + G and (3) C + E + G, each of 5 min duration (the subscript ‘5’ refers to the 5th octave on a standard piano keyboard, i.e., one octave above middle C). Zebrafish can detect sounds from 100 Hz to 8000 Hz (Higgs et al., 2011) and in a previous study associated risk with a tone of 285 Hz (Seigel et al., 2021). The three sound files were downloaded to an iPhoneXR. Playback was conducted with volume set to maximum through waterproof earbud headphones (model PWPE10B, PyleUSA.com) suspended inside the test aquaria 12 cm from the tank bottom i.e., equidistant between the surface and bottom, in zone 3 of the grid drawn on the front pane. This method produced clearly discernible auditory stimuli, with each tone of equal amplitude (Figure 1).Figure 1.Power spectra for tones used for conditioning and playback trials using tones C5 (523.25 Hz), E5 (659.25 Hz), and G5 (783.99 Hz). Upper panel, C alone; middle panel, C + G; lower panel C + E + G.2.3.Power spectra analysis of auditory test stimuliThe auditory stimuli C, CG, and CEG were played through waterproof earbuds fixed at one end of a 37-litre aquarium (same size as used for test trials), and recorded with a hydrophone (kithub.cc) fixed at the opposite end. No aerator was in the aquarium during these tests in order to reduce the background noise level. The microphone output was amplified and filtered (gain = 1000, band pass filtering 1 Hz to 5 kHz) using an amplifier built in-house (Land et al., 2001). Open-source software (Audacity®, version 2.4.1) was used to produce a power spectrum of each tone combination (sampling time 10 s, sampling rate 44 kHz, spectrum size 131,072). The power spectra are shown in Figure 1. The sound level was taken relative to the acoustic background noise level. The relative amplitudes of the individual tones C, E and G were 45 ± 5 dB, 39.0 ± 0.7 dB and 39.7 ± 0.7 dB, respectively.2.4.Experimental protocolIndividual test fish were placed in 37-litre aquaria and left for at least 24 h to acclimate to the tank before testing. The front pane of the tank facing the aisle had a 5 × 5 cm grid drawn on it for scoring fish behaviour. Each tank was equipped with an air stone and a second length of airline tubing attached to the air stone to allow for injection and dispersal of test stimuli. Each tank had a small shelter object made from a ceramic tile 11 × 11 cm supported by legs 4 cm tall on each corner.Table 1.Experimental design using tones C5 (523.25 Hz), E5 (659.25 Hz) and G5 (783.99 Hz).The experiment occurred in two steps; a conditioning trial followed by a test trial (Table 1). All fish were conditioned with the multicomponent auditory stimulus C + E + G played simultaneously for 30 s during the stimulus-injection period between pre- and post-stimulus behavioral observations. The negative control treatment group received water while the three tones were being played. This treatment controlled for the potential of a pre-existing aversion to the CEG stimulus. The other three treatment groups received 10 ml of alarm cue solution while the three-note auditory stimulus was played. In all trials, fish activity (total number of grid lines crossed), vertical distribution (horizontal row of the grid, 1 = tank bottom, 5 = surface, occupied by the fish every 10 s), and use of the shelter object (total accumulated time under the shelter) was scored for 5 min before (pre-stimulus observations) and for 5 min after (post-stimulus observations) presentation of test stimuli. Each fish was used only once in a condition-test sequence. Injection hoses were replaced after each trial.After each conditioning trial was completed, tanks were drained and refilled with room-temperature dechlorinated tap water. After at least 24 h (24–168 h, mean ± SE = 67.2 ± 3.9 h), a test trial was conducted. The group conditioned with water + CEG stimuli was tested with the CEG stimulus for 30 s. Of the three treatment groups conditioned with alarm cue + CEG, one was tested with the CEG stimulus for 30 s, one group was tested with two of the three notes play (CG) for 30 s and one group was tested with one of the three notes (C) for 30 s. No chemical stimuli were presented during test trials.2.5.Statistical treatmentVariables for change in behaviour (post-stimulus–pre-stimulus) were not normally distributed (Kolmogorov-Smirnov test, p<0.05). Mann–Whitney U-tests were used to compare the behavioural responses of water-conditioned fish to alarm cue-conditioned fish in conditioning trials. Comparisons of behavioural responses among the four treatment groups in test trials were done by Kruskal-Wallis one-way ANOVAs followed by Bonferroni-corrected pairwise comparisons between treatment pairs. The software used for these analyses was SPSS v26.3.Results3.1.Conditioning trialsRelative to control trials that received blank water, zebrafish that received conspecific alarm cues reduced vertical distribution (U=167.5, N=59, p=0.005), trended strongly to reduce activity (U=223.5, N=59, p=0.064), and increased time under the shelter (U=502.5, N=59, p=0.003; Figure 2).Figure 2.Median ± quartiles change in behaviour by zebrafish in response to either blank water (open fill) or chemical alarm cues for conditioning trials (shaded fill).3.2.Test trialsResponse to acoustic stimuli in test trials differed by the completeness of the acoustic stimulus used in the test trial (CEG, CG or C). Treatment group had a significant effect on change in vertical distribution (KW = 10.374, df = 3, p=0.016; Figure 3) but not activity (KW = 0.915, p=0.822; Figure 4), or Shelter Use (KW = 4.669, df = 4, p=0.198; Figure 5).Figure 3.Median ± quartiles of change (post-pre) in vertical distribution (horizontal row of the grid) in response to tones in test trials. All fish were conditioned with a three-note chord (CEG) + either water (W) or conspecific alarm cue (A) and later tested with either all three notes (CEG), two of the three notes (CG) or only one note (C). Shared letters above bars are not statically different in Kruskal-Wallis post-hoc pairwise comparison tests among all four treatment groups (p>0.05).Figure 4.Median ± quartiles of change (post-pre) in activity (number of lines crossed) in response to tones in test trials. All fish were conditioned with a three-note chord CEG + either water (W) or conspecific alarm cue (A) and later tested with either all three notes (CEG), two of the three notes (CG) or only one note (C).Figure 5.Median ± quartiles of change (post-pre) in time under the shelter object in response to tones in test trials. All fish were conditioned with a three-note chord CEG + either water (W) or conspecific alarm cue (A) and later tested with either all three notes (CEG), two of the three notes (CG) or only one note (C).Table 2.Alpha-corrected p-values for post-hoc pairwise comparisons among treatment groups in test trials for change in vertical distribution.Pairwise comparisons of change in vertical distribution revealed that high variability within the water control treatment group caused it to overlap with the other three treatment groups (Table 2). For fish tested with the three-note chord, those conditioned with alarm cues + CEG trended toward a reduction in vertical distribution compared to those conditioned with water + CEG (U=157, N=59, p=0.065). For fish conditioned with alarm cues + CEG, response intensity in test trials to CEG did not differ from CG but response intensity to CEG and CG were each significantly more intense that the behavioural response to C alone (Table 2). These results indicate that zebrafish significantly reduced vertical distribution in response to the full three-note CEG and to the partial two-note CG compared to the one-note C stimulus.4.DiscussionZebrafish conditioned with alarm cues and a three-note auditory stimulus responded with antipredator behaviour when re-exposed to either the same three-note stimulus, or two of the three notes, relative to a playback stimulus containing only one of the three notes. These data corroborate earlier studies that demonstrated that fishes can associate risk with auditory stimuli (Wisenden et al., 2008; Seigel et al., 2021, 2022). Response intensity of learned responses were of lower intensity than behavioral responses during conditioning trials, manifest only as a significant change in vertical distribution but not change in activity or shelter use. In contrast, Seigel et al. (2021) showed significant changes in all three behavioural measures during test trials of zebrafish conditioned to a single tone. Nevertheless, these data are the first to show that recognition of novel auditory indicators of predation risk can occur in response to multicomponent auditory stimuli and that response intensity is sensitive to the completeness of the match between conditioned and test stimuli. As such, these data provide an example of how information uncertainty mediates the trade-off between risk assessment and other life tasks (Lima & Bednekoff, 1999; Sih et al., 2000; Ferrari et al., 2016).Information about the temporal and spatial distribution of risk confers fitness benefits to prey, but monitoring and updating this information incurs a cost in that the best information is available at times and places of greatest risk. Small fishes employ surreptitious olfactory sampling behaviours (“stealthy sniffing”) to monitor olfactory information while minimizing whole-body movements that are easily detected by predators. For example, darters and cichlids engage in opercular pumping (Hawkins et al., 2004; Gibson & Mathis, 2006; Bareto & Volpato, 2006; Bareto et al., 2012; Sanches et al., 2015) and minnows perform fin-flicking (Sutrisno et al., 2014) in response to alarm cues. Both of these behaviours increase the rate of water flow over chemoreceptors in the external nares. In lab and field studies, fish avoid areas labelled with chemical alarm cues released from damaged epidermal tissues of conspecifics but continue to sample risk from afar aided by semiochemicals carried on water currents (Wisenden et al., 1994, 1995, 2010). Alternatively, prey fish may approach the source of danger to acquire information about risk (Brown & Godin, 1999; Wisenden et al., 2010).The “predator recognition continuum hypothesis” predicts that responses to indicators of risk lie on a continuum from innate recognition mechanisms (no learning), to acquired recognition (no generalization) to generalization from known predators to general classes of predictors of risk (Ferrari et al., 2007). Generalization of antipredator responses occurs when a novel stimulus is either missing some of the components of the conditioned stimulus, and/or has novel components added, reducing the certainty of recognizing the novel stimulus as an indicator of risk, resulting in a response of lower intensity or no response at all (Griffin et al., 2001; Ferrari et al., 2007, 2009; Ferrari & Chivers, 2009; Brown et al., 2011; Chivers et al., 2012, 2014; Mitchell et al., 2013). In the current study, we did not add novel notes, but simply sequentially deleted them until test subjects were too uncertain that the test stimulus matched the conditioned stimulus, and therefore did not show a behavioural response. There are parallels between our current study and studies where visual matching of novel predator identity to a conditioned model is impaired by the presence of turbidity (Ferrari et al., 2010b; Chivers et al., 2012) or when chemical cues are diluted below response threshold (Lawrence & Smith, 1989; Mirza & Chivers, 2003, Mirza et al., 2006). However, in those studies response strength was reduced when stimulus amplitude was reduced, whereas in our study stimulus intensity was held constant but the presence of selected components were deleted. Future experiments with various combinations of deleted and added components would inform the cognitive decision-making thresholds to respond versus ignore.Zebrafish presented with a multicomponent stimulus of three notes in this study failed to respond when only a single note was replayed. This finding contrasts with how fathead minnows can learn to recognize the alarm cues of a heterospecific member of their prey guild (brook stickleback, Culaea inconstans) from a multicomponent olfactory cue that contains the odour of the predator (yellow perch, Perca flavescens) fed a mixed diet of fathead minnows and stickleback (Mirza & Chivers, 2001). In that study, test subjects were able to isolate and remember components of a mixed olfactory stimulus, associate each of them with risk, and later respond with antipredator behaviour when a subset of those chemicals were later presented on their own. There may well be an interaction between the sensory modality and the trade-off between the detect-respond/detect-ignore threshold. This topic is ripe for future work.Releaser-induced recognition learning, and generalization, enable prey to label novel stimuli as dangerous. However, animals must evade risk while simultaneously engaging in other potentially conflicting, fitness-positive activities such as foraging, social and reproductive behaviours. Thus, animals cannot afford to minimize risk; instead, they must manage risk. There is a growing literature dedicated to revealing how animals avoid over-responding to stimuli that may or may not indicate risk (Wisenden & Harter, 2001; Brown et al., 2011; Ferrari & Chivers, 2011; Chivers & Ferrari, 2013; Mitchell et al., 2015). Generalization windows may be broad if the diversity of predators is relatively small (Ferrari et al., 2007) or narrow if predator diversity is large (Mitchell et al., 2013). If the window for predator-recognition generalization is set too broadly, then prey risk responding to non-threatening predators. For example, velvet geckos (Oedura lesueurii) avoid all elapid snakes even though only some are dangerous (Webb et al., 2010). Conversely, a predator-recognition window that is too conservative may lead prey to fail to respond to a predator that poses real and imminent danger. For example, invasive predators can benefit from being outside of the generalized predator window of native prey (Goodchild & Stockwell, 2016; Anton et al., 2016, 2020; Carthey & Blumstein, 2018; Melotto et al., 2020; Stockwell et al., 2022). Similarly, prey can adjust response intensity to the strength of alarm cue or perception of risk represented by test stimuli (Mirza & Chivers, 2003, Mirza et al., 2006).Understanding of the role of sound in mediating fish behaviour is in its infancy (Rountree et al., 2019) but it is well established that there is a rich aquatic soundscape of acoustic signals used in the context of intrasexual male-male competition, and intersexual courtship displays (Putland et al., 2018; Ladich, 2019). In the context of predator–prey interactions, acoustic stimuli are used by prey to detect the presence of predators, by predators to detect prey (in response, prey silence their acoustic signals in the presence of risk) and production of sounds by prey when they are captured or held (Ladich, 2022). The present study reveals that acoustic stimuli may be a relatively unexplored dimension of risk assessment by fishes.In this study, we showed for the first time that fish can associate multicomponent auditory stimuli with predation risk, and respond when a novel stimulus contains all or most of the original components but not when the play back stimulus was missing most of the salient information in the original conditioning stimulus. This work underscores the potential for auditory stimuli as an experimental tool for testing cognitive mechanisms in risk assessment by fishes.

Journal

BehaviourBrill

Published: Feb 15, 2023

There are no references for this article.