ANIMAL BEHAVIOUR, 2002, 64, 387–396 doi:10.1006/anbe.2002.3066, available online at http://www.idealibrary.com on Problem solving, inhibition and domain-specific experience: experiments on cottontop tamarins, Saguinus oedipus MARC D. HAUSER*†, LAURIE R. SANTOS*, GEERTRUI M. SPAEPEN* & HEATHER E. PEARSON* Departments of *Psychology & †Program in Neurosciences, Harvard University (Received 1 May 2001; initial acceptance 11 September 2001; final acceptance 11 February 2002; MS. number: A9055) We present the results of experiments on cottontop tamarins designed to explore the relationship between problem solving, inhibitory control and domain-specific experience. The colony was divided into two groups: tool-experienced (TE) and tool-inexperienced (TI). The TE group had previously participated in a series of tool-use experiments and revealed that, when selecting a tool, they used featurally relevant dimensions (e.g. shape, material, orientation) over featurally irrelevant dimensions (e.g. colour). The TI group, although experienced in other laboratory-based experiments, had never been tested on tool or other object manipulation problems. In Phase 1, involving three conditions, all subjects were tested on a series of means–end problems involving the use of a cloth to access a piece of food. Although the correct choice always involved picking the supporting cloth, we also built in an association between the correct cloth and its colour. Once the subjects reached criterion, we reversed the association between the cloth colour and the food reward in Phase 2. If the subjects solved the problems in Phase 1 by attending to cloth colour, then in Phase 2 they should have difficulty, especially given prior findings on tamarins demonstrating that reversal learning is difficult. If the subjects solved Phase 1 by attending to the functionality of the problem (i.e. the physical/causal relationship between the cloth and food), then reversing the colours in Phase 2 should have no effect on the subjects’ performances. Finally, if the subjects attended to both colour and functionality, then reversing the colours should cause some decrement in performance, but less so than in the case where colour alone dominates. In Phase 2, although both groups showed a decrement in performance, indicating problems with reversal learning, TE subjects significantly outperformed TI subjects. Furthermore, the pattern of performance for TE subjects suggested that they had solved the initial problem by attending to a combination of colour and functionality or functionality alone, while TI subjects had attended to colour alone. We conclude that for tamarins with experience as tool users, colour represents a less salient feature, even when it is systematically associated with a food reward. For inexperienced tamarins, however, colour is salient and reversal learning is difficult. Together, these findings highlight the importance of exploring the relationship between inhibitory control and domain-specific problem solving.  2002 The Association for the Study of Animal Behaviour. Published by Elsevier Science Ltd. All rights reserved. To survive, animals must discriminate among a variety of objects and events in the environment: some foods are toxic and some are edible; some heterospecifics are predators and some are prey; some conspecifics are domi- nant and some are subordinate. During the process of acquiring a discriminating palette, individuals must learn which features are relevant and which are irrelevant. In the classic ethological work on recognition systems, including studies of imprinting by Lorenz and aggression by Tinbergen, results showed that discrimination was often based on a simple feature, one tapping an innate releasing mechanism. Thus, young chickens imprint on the first moving object they detect, but show an initial bias to imprint on chicken-like things, especially if they have a head and eyes (Bateson 1966; Bolhuis 1991). Tinbergen’s stickleback work showed that aggression could be released by a red belly, including that of the postman. If the initial discrimination is based on a feature that later turns out to be irrelevant or unreliable, then the individual must inhibit attention to this feature in favour of another with greater reliability. In the case of imprint- ing in chickens and aggression in sticklebacks, the under- lying mechanism taps a statistical bias in the natural environment. In most cases, the first moving object detected by the chicken is its mother, and most red- bellied objects are stickleback competitors rather than postmen. Thus, in the natural world, chickens and Correspondence: M. D. Hauser, Department of Psychology, 33 Kirkland Street, Harvard University, Cambridge, MA 02138, U.S.A. (email: hauser@wjh.harvard.edu). 0003–3472/02/$35.00/0  2002 The Association for the Study of Animal Behaviour. Published by Elsevier Science Ltd. All rights reserved. 387