Inter-specific differences in numerical abilities among teleost fish

Abstract

Adults, infants and non-human primates are thought to possess similar non-verbal numerical systems, but there is considerable debate regarding whether all vertebrates share the same numerical abilities. Despite an abundance of studies, cross-species comparison remains difficult because the methodology employed and the context of species examination vary considerably across studies. To fill this gap, we used the same procedure, stimuli, and numerical contrasts to compare quantity abilities of five teleost fish: redtail splitfin, guppies, zebrafish, Siamese fighting fish, and angelfish. Subjects were trained to discriminate between two sets of geometrical figures using a food reward. Fish initially were trained on an easy numerical ratio (5 vs. 10 and 6 vs. 12). Once they reached the learning criterion, they were subjected to non-reinforced probe trials in which the set size was constant but numerical ratios varied (8 vs. 12 and 9 vs. 12). They also were subjected to probe trials in which the ratio was constant, but the total set size was increased (25 vs. 50) or decreased (2 vs. 4). Overall, fish generalized to numerosities with a 0.67 ratio, but failed with a 0.75 ratio; they generalized to a smaller set size, but not to a larger one. Only minor differences were observed among the five species. However, in one species, zebrafish, the proportion of individuals reaching the learning criterion was much smaller than in the others. In a control experiment, zebrafish showed a similar lower performance in shape discrimination, suggesting that the observed difference resulted from the zebrafish's difficulty in learning this procedure rather than from a cross-species variation in the numerical domain.

Keywords: Betta splendens; Danio rerio; Fish cognition; Poecilia reticulata; Pterophyllum scalare; Xenotoca eiseni.

Figure 1

Experimental apparatus. Subjects were housed in the experimental tank for the entire experiment. Stimuli (two groups of geometric figures differing in numerosity) were presented at the bottoms of the tank.

Figure 2

Schematic representation of the stimuli used. Each panel contained black geometric figures differing in numerosity inserted in a white background. In the figure we depicted a 8 vs. 12 contrast with cumulative surface area controlled to 100% (A), to 85% (B) and to 70% (C). In (A) and (B) stimuli are controlled for overall space, whereas in (C) they are controlled for density.

Figure 3

Numerical contrasts were plotted against the accuracy of the five species. Most of the species significantly discriminated 8 vs. 12 and spontaneously generalized to smaller numerosities (2 vs. 4). All species failed to discriminate 9 vs. 12 or generalize to larger numerosities (25 vs. 50). Asterisks denote a significant departure from chance level (p < 0.05). Bars represent the standard error.

Figure 4

Frequency distribution of the preference score of zebrafish and redtail splitfin in the shape discrimination task of experiment 2. Redtail splitfin significantly preferred the reinforced stimulus, while the performance of zebrafish is indistinguishable from random choice.

Figure 5

Phylogenetic relationship of the five teleost species used in the study.