ANIMAL BEHAVIOUR, 2003, 66, 903–910 doi:10.1006/anbe.2003.2269 Discrimination of closed coloured shapes by honeybees requires only contrast to the long wavelength receptor type NATALIE HEMPEL DE IBARRA* & MARTIN GIURFA†‡ *Institut fu ¨r Biologie—Neurobiologie, Freie Universita ¨t Berlin †Centre de Recherches sur la Cognition Animale, CNRS—Universite ´ Paul-Sabatier, Toulouse ‡Departmento de Biologı ´a, Universidad de Buenos Aires, Argentina (Received 11 July 2002; initial acceptance 22 August 2002; final acceptance 2 March 2003; MS. number: 7416) Floral shape is a visual cue used by pollinators to discriminate between competing flower species. We investigated whether discrimination is possible between closed shapes presenting the same colour and lacking a centrally presented fixation point. Free-flying honeybees, Apis mellifera L., had to discriminate between a solid square and a solid triangle of the same colour presented on the back walls of a Y-maze. Different colours were used to vary chromatic contrast and receptor-specific contrasts. Discrimination was possible whenever shapes presented contrast to the long wavelength receptor but was independent of chromatic contrast, overall intensity contrast or short and middle wavelength receptor contrast. We suggest that the bees used the edges of the closed shapes to solve the task. Bees failed when shapes were rotated, showing that a single shape edge was not sufficient for recognition.  2003 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. Flowers are not only colourful, but also variable in their morphology. The variation in shape and pattern provided by flower corollas represents an additional source of information that pollinators can use to discriminate between competing flower species. Although the flower image for an insect becomes coarser with increasing distance (Land 1993; Vorobyev et al. 1997), the fact that flowers have evolved a large variety of shapes and pat- terns suggests that such visual information is important for close-up flower recognition (Wehner 1981; Dafni et al. 1997; Giurfa & Lehrer 2001). The honeybee, Apis mellifera, has been a powerful model for investigating the mechanisms by which insects learn and recognize patterns (reviewed by Wehner 1981; Srinivasan 1994; Dafni et al. 1997; Ronacher 1998; Giurfa & Lehrer 2001). Studies on honeybee pattern recognition have identified two essential strategies used to process and recognize visual patterns: (1) retinotopic template matching and (2) feature extraction. The former assumes that the template of a learned pattern is stored in a way that preserves the retinotopic coordinates (Wehner 1967, 1972; Cruse 1972; Gould 1985, 1986; Dill et al. 1993; reviewed by Heisenberg 1995). Recognition depends on the amount of overlap between an actual retinal image and the stored template. Consistent with this hypothesis, it has been shown in the context of scene recognition during navigation, that bees can store landmark shapes as retinotopic snapshots in which the retinal positions of the edges and not the area overlap are important (Cartwright & Collett 1983; reviewed by Collett 1996). The feature extraction hypothesis assumes that in pattern discrimination, selective attention and processing are focused on a single feature (orientation, symmetry, con- tour density, size, etc.) such that recognition depends on the presence or absence of such a feature and its specific values (van Hateren et al. 1990; Srinivasan et al. 1994; Horridge & Zhang 1995; Giurfa et al. 1996a; Ronacher 1998). Early studies suggested that bees do not discriminate between closed shapes, such as a square, a disc or a triangle, either in the horizontal (von Frisch 1914, 1965) or in the vertical (Baumga ¨rtner 1928) plane of presen- tation. In addition, dissected shapes, that is, shapes with many contours, could not be discriminated from each other but could be easily distinguished from closed shapes. Such experiments led to the conclusion that the spatial scale of a figure (i.e. its degree of dissectedness or ‘figural intensity’) was the primary cue used for pattern discrimination by bees (Hertz 1929, 1933; Zerrahn 1933; Wolf & Zerrahn-Wolf 1935). However, researchers later doubted that contour density alone was sufficient to describe the bees’ discrimination abilities for patterns. Correspondence: N. Hempel de Ibarra, Institut fu ¨r Biologie– Neurobiologie, Freie Universita ¨t Berlin, Ko ¨nigin-Luise-Strasse 28/30, D-14195 Berlin, Germany (email: nhempel@neurobiologie.fu- berlin.de). M. Giurfa is at the Centre de Recherches sur la Cognition Animale, CNRS—Universite ´ Paul-Sabatier—Toulouse III—UMR 5169, 118 Route de Narbonne, 31062 Toulouse cedex 4, France. 0003–3472/03/$30.00/0  2003 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. 903