Welcome to our new website

What can bees really tell us about the face processing system in humans?
Olivier Pascalis, David J. Kelly, Roberto Caldara

Cognitive abilities present in humans, such as face processing, are likely to have evolved under various ecological pressures. A comparable and specialized face processing system observed in sheep and non-human primates suggests a possible common origin in evolution (Pascalis et al., 1999; Kendrick et al., 2001; Parr, 2003). However, it is important to determine when this system emerged during evolution if we are to fully understand it. Studies conducted with honeybees (Apis mellifera) and wasps (Polistes fuscatus) could potentially contribute to our understanding of this ability.

Dyer et al. have demonstrated that bees are able to learn and recognise the picture of a human face when paired with a novel face (Dyer et al., 2005), which is consistent with our existing knowledge of the bees' visual ability. However, we believe that Dyer's extrapolations about how recognition is achieved and whether or not it is facilitated by specialised brain regions are misleading.

Face recognition is carried out by an automated and specific process in humans, which is known as configural processing (perceiving metrical relations between face features). Contrary to Dyer's argument and to earlier research findings (Diamond and Carey, 1977), it is now debatable whether such processing develops late in childhood (Schwarzer et al., 2005). Furthermore, studies that have created `visual experts' who develop configural processing for non-face objects require many more hours of intensive training than reported by Dyer et al. In their study, there is no clear evidence of configural processing and it is likely that the bees' recognition relied on specific features.

In human adults, functional neuroimaging studies have identified a network of areas within the ventral temporal cortex that are highly responsive to faces (Haxby et al., 2000), with maximum selectivity in the right middle fusiform gyrus: the so-called `fusiform face area' (FFA) (Kanwisher et al., 1997). A comparable functional specialization supports face processing in the primate brain (Tsao et al., 2006). Critically, however, normal face identification relies on the integrity of this complex network, as prosopagnosic patients with lesions sparing the FFA show impaired use of optimal information for face identification (Caldara et al., 2005; Schiltz et al., 2006).

With data collected from just five bees, it is too speculative to conclude that specialised brain regions are not necessary for face processing in humans. Humans and bees have not shared a common ancestor for roughly 600 million years and have evolved very differently since this separation. We can therefore expect them to process faces differently. Clearly, more studies are required to determine how the honeybee succeeds in simple face matching tasks before attempting to establish potential similarities between its visual recognition abilities and those of different species. It is first necessary to establish whether bees are able to recognise or categorise conspecifics in a similar way to the wasp (Tibbetts, 2002; Tibbetts and Dale, 2004). Given that humans appear capable of only processing faces confined to human and non-human primate categories, it would be somewhat paradoxical if the bee demonstrated recognition with human faces but not with conspecifics. Finally, artificial computing systems without a neural substrate also demonstrate an optimal ability to recognize individual faces. Does such evidence question the neural specificity of face processing in humans?