Symbol-value association and discrimination in the archerfish

Recognition of natural objects in the archerfish

Recognition of individual objects and their categorization is a complex computational task. Nevertheless, visual systems can perform this task in a rapid and accurate manner. Humans and other animals can efficiently recognize objects despite countless variations in their projection on the retina due to different viewing angles, distance, illumination conditions and other parameters. To gain a better understanding of the recognition process in teleosts, we explored it in archerfish, a species that hunts by shooting a jet of water at aerial targets and thus can benefit from ecologically relevant recognition of natural objects. We found that archerfish not only can categorize objects into relevant classes but also can do so for novel objects, and additionally they can recognize an individual object presented under different conditions. To understand the mechanisms underlying this capability, we developed a computational model based on object features and a machine learning classifier. The analysis of the model revealed that a small number of features was sufficient for categorization, and the fish were more sensitive to object contours than textures. We tested these predictions in additional behavioral experiments and validated them. Our findings suggest the existence of a complex visual process in the archerfish visual system that enables object recognition and categorization.

Highlighted Article: Archerfish are capable of natural object recognition and categorization based on a small number of visual features.

Archerfish number discrimination

Debates have arisen as to whether non-human animals actually can learn abstract non-symbolic numerousness or whether they always rely on some continuous physical aspect of the stimuli, covarying with number. Here, we investigated archerfish (Toxotes jaculatrix) non-symbolic numerical discrimination with accurate control for covarying continuous physical stimulus attributes. Archerfish were trained to select one of two groups of black dots (Exp. 1: 3 vs 6 elements; Exp. 2: 2 vs 3 elements); these were controlled for several combinations of physical variables (elements' size, overall area, overall perimeter, density, and sparsity), ensuring that only numerical information was available. Generalization tests with novel numerical comparisons (2 vs 3, 5 vs 8, and 6 vs 9 in Exp. 1; 3 vs 4, 3 vs 6 in Exp. 2) revealed choice for the largest or smallest numerical group according to the relative number that was rewarded at training. None of the continuous physical variables, including spatial frequency, were affecting archerfish performance. Results provide evidence that archerfish spontaneously use abstract relative numerical information for both small and large numbers when only numerical cues are available.

Magnitude integration in the Archerfish

We make magnitude-related decisions every day, for example, to choose the shortest queue at the grocery store. When making such decisions, which magnitudes do we consider? The dominant theory suggests that our focus is on numerical quantity, i.e., the number of items in a set. This theory leads to quantity-focused research suggesting that discriminating quantities is automatic, innate, and is the basis for mathematical abilities in humans. Another theory suggests, instead, that non-numerical magnitudes, such as the total area of the compared items, are usually what humans rely on, and numerical quantity is used only when required. Since wild animals must make quick magnitude-related decisions to eat, seek shelter, survive, and procreate, studying which magnitudes animals spontaneously use in magnitude-related decisions is a good way to study the relative primacy of numerical quantity versus non-numerical magnitudes. We asked whether, in an animal model, the influence of non-numerical magnitudes on performance in a spontaneous magnitude comparison task is modulated by the number of non-numerical magnitudes that positively correlate with numerical quantity. Our animal model was the Archerfish, a fish that, in the wild, hunts insects by shooting a jet of water at them. These fish were trained to shoot water at artificial targets presented on a computer screen above the water tank. We tested the Archerfish's performance in spontaneous, untrained two-choice magnitude decisions. We found that the fish tended to select the group containing larger non-numerical magnitudes and smaller quantities of dots. The fish selected the group containing more dots mostly when the quantity of the dots was positively correlated with all five different non-numerical magnitudes. The current study adds to the body of studies providing direct evidence that in some cases animals' magnitude-related decisions are more affected by non-numerical magnitudes than by numerical quantity, putting doubt on the claims that numerical quantity perception is the most basic building block of mathematical abilities.

Individual behavioural traits not social context affects learning about novel objects in archerfish

Abstract

Learning can enable rapid behavioural responses to changing conditions but can depend on the social context and behavioural phenotype of the individual. Learning rates have been linked to consistent individual differences in behavioural traits, especially in situations which require engaging with novelty, but the social environment can also play an important role. The presence of others can modulate the effects of individual behavioural traits and afford access to social information that can reduce the need for ‘risky’ asocial learning. Most studies of social effects on learning are focused on more social species; however, such factors can be important even for less-social animals, including non-grouping or facultatively social species which may still derive benefit from social conditions. Using archerfish, Toxotes chatareus, which exhibit high levels of intra-specific competition and do not show a strong preference for grouping, we explored the effect of social contexts on learning. Individually housed fish were assayed in an ‘open-field’ test and then trained to criterion in a task where fish learnt to shoot a novel cue for a food reward—with a conspecific neighbour visible either during training, outside of training or never (full, partial or no visible presence). Time to learn to shoot the novel cue differed across individuals but not across social context. This suggests that social context does not have a strong effect on learning in this non-obligatory social species; instead, it further highlights the importance that inter-individual variation in behavioural traits can have on learning.

Significance statement

Some individuals learn faster than others. Many factors can affect an animal’s learning rate—for example, its behavioural phenotype may make it more or less likely to engage with novel objects. The social environment can play a big role too—affecting learning directly and modifying the effects of an individual’s traits. Effects of social context on learning mostly come from highly social species, but recent research has focused on less-social animals. Archerfish display high intra-specific competition, and our study suggests that social context has no strong effect on their learning to shoot novel objects for rewards. Our results may have some relevance for social enrichment and welfare of this increasingly studied species, suggesting there are no negative effects of short- to medium-term isolation of this species—at least with regards to behavioural performance and learning tasks.

Categorical face perception in fish: How a fish brain warps reality to dissociate "same" from "different"

Categorical perception (CP) is the phenomenon by which a smoothly varying stimulus property undergoes a nonlinear transformation during processing in the brain. Consequently, the stimuli are perceived as belonging to distinct categories separated by a sharp boundary. Originally thought to be largely innate, the discovery of CP in tasks such as novel image discrimination has piqued the interest of cognitive scientists because it provides compelling evidence that learning can shape a category's perceptual boundaries. CP has been particularly closely studied in human face perception. In nonprimates, there is evidence for CP for sound and color discrimination, but not for image or face discrimination. Here, we investigate the potential for learned CP in a lower vertebrate, the damselfish Pomacentrus amboinensis. Specifically, we tested whether the ability of these fish to discriminate complex facial patterns tracked categorical rather than metric differences in the stimuli. We first trained the fish to discriminate sets of two facial patterns. Next, we morphed between these patterns and determined the just noticeable difference (JND) between a morph and original image. Finally, we tested for CP by analyzing the discrimination ability of the fish for pairs of JND stimuli along the spectrum of morphs between two original images. Discrimination performance was significant for the image pair straddling the boundary between categories, and chance for equivalent stimulus pairs on either side, thus producing the classic "category boundary" effect. Our results reveal how perception can be influenced in a top-down manner even in the absence of a visual cortex.

Archerfish vision: Visual challenges faced by a predator with a unique hunting technique

Archerfish are well-known for their ballistic hunting behaviour, in which they shoot down aerial prey with a well-aimed jet of water. This unique hunting strategy poses several challenges for visual systems. Archerfish face significant distortion to the appearance of targets due to refraction at the air/water interface, they search for prey against a complex background of foliage, they change prey targeting behaviour as conditions change, and they must make high speed decisions to avoid competition. By studying how archerfish have overcome these challenges, we have been able to understand more about fundamental problems faced by visual systems and the mechanisms used to solve them. In some cases, such as when searching for targets, the visual capabilities of archerfish are functionally similar to those of humans, despite significant differences in neuroanatomy. In other cases, the particular challenge faced by archerfish magnifies fundamental problems generally faced by visual systems, such as recognizing objects given strong viewpoint dependent changes to appearance. The efficiency of archerfish retrieving fallen prey to avoid kleptoparasitism, demonstrates that their visual processing excels in both speed and accuracy. In this review, we attempt to provide an overview of the many facets of visually driven behaviour of archerfish, and how they have been studied. In addition to their hunting technique, archerfish are ideal for visual processing experiments as they can be quickly trained to perform a range of non-ecologically relevant tasks. Their behavioural flexibility moreover, introduces the opportunity to study how experience-dependence and choice affects visual processing.

What pops out for you pops out for fish: Four common visual features

Visual search is the ability to detect a target of interest against a background of distracting objects. For many animals, performing this task fast and accurately is crucial for survival. Typically, visual-search performance is measured by the time it takes the observer to detect a target against a backdrop of distractors. The efficiency of a visual search depends fundamentally on the features of the target, the distractors, and the interaction between them. Substantial efforts have been devoted to investigating the influence of different visual features on visual-search performance in humans. In particular, it has been demonstrated that color, size, orientation, and motion are efficient visual features to guide attention in humans. However, little is known about which features are efficient and which are not in other vertebrates. Given earlier observations that moving targets elicit pop-out and parallel search in the archerfish during visual-search tasks, here we investigate and confirm that all four of these visual features also facilitate efficient search in the archerfish in a manner comparable to humans. In conjunction with results reported for other species, these finding suggest universality in the way visual search is carried out by animals despite very different brain anatomies and living environments.

Quantity discrimination in fish species: fish use non-numerical continuous quantity traits to select shoals

Fish typically prefer to live in big shoals due to the associated ecological benefits. Shoaling is a behavior that depends on the ability to quantitatively discriminate. The fundamental mechanism involved in quantity discrimination determines whether fish can discriminate a shoal using numerical discrete cues (e.g., number of shoal members), non-numerical continuous traits (e.g., total body surface area) or both; however, the mechanism is currently a controversial topic. In the present study, we used a spontaneous choice experiment to test whether guppy (Poecilia reticulata), zebrafish (Danio rerio), Chinese crucian carp (Carassius auratus) and qingbo (Spinibarbus sinensis) rely on continuous (i.e., body surface area) or discrete (i.e., number of shoal members) information for shoal selection by altering the body surface area (cumulative body surface area ratio of 3:2 or 1:1) between two stimulus shoals with a different number of members (2 individuals vs 3 individuals). All four fish species preferred to shoal with the stimulus shoal with the larger cumulative surface area even if the shoal had fewer members; however, fish showed no shoal preference when the cumulative surface body areas of both stimulus shoals were equal. Furthermore, qingbo did not numerically discriminate between a shoal with 1 individual and a shoal with 3 individuals when the cumulative surface areas of both stimulus shoals were equal; however, qingbo showed a preference for the shoal with the larger cumulative surface area when the two stimulus shoals each had 3 individuals. In conclusion, the present study demonstrated that all four fish species relied only on non-numerical continuous quantity information for shoal selection, at least under a difficult task (i.e., 2 vs 3).