At the root of number competence. Meta-analysis of literature on different animal species and an experimental contribution to the understanding of rudimental numerical abilities in an animal model, the young domestic chick (gallus gallus).

Davis and Pérusse (1988) argued that, although animals can be trained to make numerical discriminations, they do so only as a last-resort strategy, when extensive training is provided and all other cues are eliminated. In spite of this criticism, in the last decade various types of numerical competences have been demonstrated in nonverbal creatures - namely pre-verbal infants and non-human species - that demonstrated the presence of number competence in the absence of language (reviews in Gallistel and Gelman, 1992; Dehaene, 1997; Hauser and Spelke, 2004). In the present study three separate sets of experiments were carried out to investigate respectively: Spontaneous numerical discrimination following imprinting, number discrimination using a conditioning procedure and ordinal numerical competence. In the first set of experiments, by employing a spontaneous choice paradigm, the ability to discriminate small sets of objects was confirmed in 3-day-old-chicks even when the continuous variables were controlled for. These data showed, for the first time, that spontaneous number discrimination can be based on numerical cues only. In the second set of experiment, chicks' ability to discriminate between small sets of object (up to 3) was confirmed by employing operant conditioning procedures. What is interesting is that although training was done with only one specific set of stimuli, in which number co-varied with several continuous physical variables, the chicks seemed to encode number rather than physical variables. These data also provided the first evidence of numerical discrimination of partly occluded objects. Furthermore, discrimination of small numerosities in young chicks seems to be carried out using an Object File System, with a set-size limit of around 4 elements. In the third set of experiments investigating ordinal abilities show that 5-day-old-chicks can successfully learn to identify a target on the exclusive basis of its serial position in a series of 10. A peculiar finding, in Experiment 4.3, was that whenever position had to be identified on a left/right oriented series, in the generalization test, chicks would more often find the correct position by starting from the left end of the series rather than from the right end. In the absence of environmental asymmetries a possible explanation would be that there is a right hemispheric dominance (left visual hemifield) for this sort of task. It would be interesting for future experiments to establish the limit of such ordinal ability, using a different apparatus with more positions as well as trying to understand the hemispheric asymmetry raised in this experiment. All these data demonstrated the presence of rudimentary numerical competence even in a species so distant and so different from humans. These results support the hypothesis that numerical competences are not only a prerogative of adult humans but that such abilities should have an evolutionary precursor in animals.