Fish cognition

Zebrafish (Danio rerio)

Zebrafish display an innate tendency to aggregate with conspecifics. This behaviour is termed `shoaling' and can be elicited by visual cues alone: when presented with a movie of other zebrafish, individual animals demonstrate keen interest in this display. We have capitalized on this behaviour to generate experimental settings akin to psychophysical 2-way testing paradigms: a real zebrafish (test animal) is placed inside a tank flanked by two monitors displaying synthetic zebrafish; the two stimuli compete for attention on the part of the test animal and drive preferential navigation towards the more shoal-like stimulus. This visually-guided behaviour can be tracked and quantified using custom-built hardware/software (yellow dots in Fig. 1).

Figure 1 Yellow dots show animal positions (every 1/4 second) returned by fully automated tracking software. In this example, the zebrafish (highlighted by red tint) displays preference for the high-contrast stimulus on the left.

Using this paradigm, we have demonstrated that zebrafish are able to carry out complex visual computations such as feature binding (Neri 2012). For example, they demonstrate preferential behaviour in relation to stimuli where shape and motion are combined in either congruent or incongruent fashion. Both configurations contain 6 leftward-moving and 6 rightward-moving elements, as well as 6 leftward-pointing and 6 rightward-pointing fish icons. In the congruent stimulus (Fig. 2A), fish icons move in the direction towards which they point; in the incongruent stimulus (Fig. 2B), the relationship between motion direction (the `motion' cue) and pointing direction (the `shape' cue) is reversed.

Figure 2 In the congruent stimulus (A), the motion of direction (indicated by arrows) of fish icons is congruent with their pointing direction. This pairing of motion and shape is reversed in the incongruent stimulus (B). Data points in C show average position of individual test animals (one point per animal) under three variants (red/magenta/yellow) of the congruent-vs-incongruent comparison. Zebrafish consistently prefer the congruent stimulus.

Zebrafish produce measurable preference for the congruent configuration (Fig. 2C and Video 1 below), demonstrating a perceptual ability for correctly binding motion and shape.

SuppMovie4.mpg
Video 1 View from inside the tank. The test animal (indicated by red arrow) spends substantial time inspecting the congruent stimulus on the left, but shows little interest in the incongruent stimulus on the right. Notice that, on the left, the animal often attempts to align itself with the movement of individual synthetic elements in "mirror-like" behaviour.

In subsequent work (Spilioti et al 2016) we have characterized the intrinsic behavioural variability of visually-guided behaviour in the zebrafish, and we have found that its impact on navigation conforms to the expectation of a simple model built around the principles of signal detection theory.


Siamese fighting fish (Betta splendens)

Teleosts represent the most diverse vertebrate group, ranging from highly gregarious species like zebrafish to solitary territorial species like Siamese fighting fish. The latter species (Betta splendens) spontaneously displays aggressive behaviour in response to an approaching opponent by extending its fins and gills ("flaring"). This behaviour is particularly evident in males and can be quantified via automated tracking procedures (Fig. 3).

Figure 3 The test animal (outlined by blue region in A) is placed within a square arena flanked by two displays, one of which (left in the example) shows video of an opponent. Tracking software identifies head position/direction (indicated by black cross) and gill location (indicated by red regions adjacent to blue region in A). B plots head position/direction for stimuli presented to the left (red) and to the right (black). Light-coloured dots show head position when the animal did not produce flaring (gill extension), while full-colour elongated symbols indicate head position and direction (larger part of symbol pointing to front) during flaring events. C plots corresponding distributions for all head positions along tank. Polar plot in D shows corresponding distributions for head directions. Black open histogram shows distribution during the baseline phase (no stimulation). E plots same as D but restricted to flaring events.

We have exploited this idiosyncratic behaviour to understand what kind of visual information is represented by the visual system of fighting fish (Neri 2019). The test animal effectively performed psychophysical 2-alternative-choice experiments by attacking videos depicting manipulated versions of a synthetic opponent (see Video 2).

BettaSuppMovie1.mp4
Video 2 Automated tracking of aggressive display. Top panel shows view from inside the tank: immediately after the stimulus appears, the animal can be seen approaching the display and engaging in aggressive behaviour against the synthetic opponent. Bottom-left panel shows top view of same sequence. Bottom-right panel shows top view with overlayed tracking elements produced by the tracking algorithm: segmented animal is indicated by blue pixels, head template is shown by black parabolic shape, head position (front tip) is marked by black cross, flaring gills are labelled by red pixels when flaring event is detected.

Results from such quantitative behavioural measurements ((Fig. 4) expose a sophisticated degree of visual representation, enabling Siamese fighting fish to discriminate subtle visual manipulations such as warping (Fig. 4D-F) and reverse playback (Fig. 4J-L).

Figure 4 Middle column (B, E, H, K) plots behavioural drive: when tilted away from 0, it indicates measurable preference. Right column (C, F, I, L) shows intact stimulus; left column shows competing stimulus for detection (blank screen in A), discrimination of warping manipulation (D), inversion (upside-down stimulus in G) and reverse-playback (J).

The configural nature of the perceptual operations engaged by fighting fish is reminiscent of how mammals represent socially relevant signals, notwithstanding the lack of cortical structures that are widely recognized to play a critical role in higher cognitive processes. Together with the zebrafish studies (see above), our results indicate that mammalian-centric accounts of social cognition present serious conceptual limitations. Our findings highlight the importance of understanding complex perceptual function from a general ethological perspective.

Relevant publications:

• Neri P Complex visual analysis of ecologically relevant signals in Siamese fighting fish 2019 Animal Cognition in press

• Spilioti M, Vargesson N, Neri P Quantitative assessment of intrinsic noise for visually guided behaviour in zebrafish 2016 Vision Research 127 104-114

• Neri P Feature binding in zebrafish 2012 Animal Behaviour 84 485-493