Paper wasps of the species Polistes fuscatus live in strict hierarchical societies in which the ability to identify superiors and subordinates is crucial. Like humans, these insects have a cognitive tool kit for recognizing familiar faces.
The ability to recognize individuals can convey significant benefits to social animals. In humans, the capacity to recognize different faces is crucial for making individual behaviour predictable for other members of a group — for keeping track of who is aggressive, bold or wise — and so for knowing everyone's place in a family or society. Moreover, there is strong evidence1,2 that primate brains contain specialized modules for face processing and recognition. Writing in Science, Sheehan and Tibbetts3 present evidence that an insect, the paper wasp Polistes fuscatus, not only recognizes the faces of individuals of the same species, but is a veritable expert at face discrimination.
Individual queens of many social insects, including Polistes metricus — a species closely related to P. fuscatus — found colonies in the spring, a perilous and challenging enterprise for a single insect. By contrast, several P. fuscatus queens typically join forces to build, defend and provide for a new nest. This boosts the chances of the project's success, but, for many of the hopeful foundresses, there is a price to pay: a linear hierarchy is established through a series of one-on-one fights. Consequently, the strongest queen dominates egg-laying, whereas subordinates do more menial tasks. After a duel, individuals recognize their opponents by their distinct facial markings, which helps to avoid the repetition of potentially costly battles.
Sheehan and Tibbetts studied face recognition in P. fuscatus and P. metricus. They trained the insects to choose one of two arms of a maze, each marked by an image (Fig. 1). If the wasp turned to the 'wrong' image, it experienced an electric shock, whereas choosing the 'correct' arm provided safety. The image pairs consisted of normal wasp faces, manipulated wasp faces, simple geometric patterns or caterpillars — the typical prey of these wasps.
The authors report that the face-learning ability of P. fuscatus was greater than that of P. metricus. Moreover, P. fuscatus wasps learned to recognize wasp faces more quickly and more accurately than they did other image types. Intriguingly, this insect had difficulties recognizing faces that lacked antennae or that contained scrambled features. This indicates that face learning in P. fuscatus is specific to correctly configured faces, and cannot be explained by a general pattern-recognition system.
Because Sheehan and Tibbetts3 used wasps caught in the wild, it is not clear whether the insects' expertise at face recognition was acquired through evolution, individual experience before capture or an interaction of both. Nonetheless, at the behavioural level, face recognition in P. fuscatus shares some features with that in humans. For humans, too, the configuration of facial features is important for reliable recognition4. Elemental features such as eyes, nose and mouth are not processed individually but as a whole, and their rearrangement significantly impairs recognition performance.
Several types of processing of facial configurations that have been identified in humans could be explored in wasps. These include detection of a face on the basis of its overall arrangement of features; face processing where certain elemental features are bound together; and sensitivity to second-order relationships, whereby the relative distances between various elemental features enhance recognition5.
Neurophysiological investigation of specialized brain areas mediating face processing in humans1,6 and monkeys2 confirmed the existence of bespoke neural mechanisms for analysing the arrangement of facial features. Studies in other mammals7 also indicated that the ability to process configural face information involves specialized neural circuitry and extensive experience. However, work in insects on visual processing of compound patterns (including face images) suggests8 that species without specialized face-processing abilities may also be capable of configural processing. In terms of gross neuroanatomy9, there are no discernible differences between the visual system of P. fuscatus and those of related species that do not use face recognition.
It is therefore likely that neural circuitry similar to that used by insects for prey recognition is co-opted in P. fuscatus for face recognition owing to the insect's social lifestyle. This might require only fairly minor adjustments in neural circuitry to embed visual pattern recognition into the appropriate (social) context. Sheehan and Tibbetts's observation3 that P. metricus lacks specialized face recognition lends further support to the view that specialized cognitive capacities can evolve relatively easily in response to pertinent selection pressures.
The finding that a small-brained insect shares the ability to recognize faces with humans and other primates may come as a surprise to adherents of the 'social brain' hypothesis10 — the idea that group-living and complex social interactions require large brains. However, in asking questions about brain size and cognitive capacity, one must focus on what the computational nature of the task is, and what neural circuitry is actually required to accomplish it. In this view, the specific task that requires a large brain has yet to be discovered.
Analyses of neuronal networks11 show that reliable face recognition can be achieved with a network of only a few hundred neurons — a circuitry that could be easily accommodated in an insect brain — especially because much of the neural hardware for pattern recognition would also be used for other visual tasks such as recognition of landmarks or prey. Small-brained invertebrates, such as insects, are thus exceptionally useful models for studying how social and cognitive abilities can be mediated by modest neural circuitry12, and of the evolutionary adjustments necessary to generate such abilities. As for bigger brains, many of their advantages might relate to higher memory storage capacity (equivalent to bigger hard drives rather than better processors), more parallel processing, or more precision and detail in sensory information, but not necessarily to more complex cognitive processes12.
Kanwisher, N. Nature Neurosci. 3, 759–763 (2000).
Tsao, D. Y., Freiwald, W. A., Tootell, R. B. H. & Livingstone, M. S. Science 311, 670–674 (2006).
Sheehan, M. J. & Tibbetts, E. A. Science 334, 1272–1275 (2011).
Collishaw, S. M. & Hole, G. J. Perception 29, 893–909 (2000).
Maurer, D., Le Grand, R. & Mondloch, C. J. Trends Cogn. Sci. 6, 255–260 (2002).
Lobmaier, J. S., Klaver, P., Loenneker, T., Martin, E. & Mast, F. W. NeuroReport 19, 287–291 (2008).
Peirce, J. W., Leigh, A. E. & Kendrick, K. M. Neuropsychologia 38, 475–483 (2000).
Avarguès-Weber, A., Portelli, G., Benard, J., Dyer, A. & Giurfa, M. J. Exp. Biol. 213, 593–601 (2010).
Gronenberg, W., Ash, L. E. & Tibbetts, E. A. Brain Behav. Evol. 71, 1–14 (2008).
Dunbar, R. I. M. Evol. Anthropol. 6, 178–190 (1998).
Aitkenhead, M. J. & McDonald, A. J. S. Eng. Appl. Artif. Intell. 16, 167–176 (2003).
Chittka, L. & Niven, J. Curr. Biol. 19, R995–R1008 (2009).
About this article
Elevated diversity in loci linked to facial morphology is consistent with the hypothesis that individual facial recognition is important across hominoids
American Journal of Physical Anthropology (2021)
Proceedings of the National Academy of Sciences (2020)
ISIJ International (2020)
Frontiers in Psychology (2019)
Property prediction and properties-to-microstructure inverse analysis of steels by a machine-learning approach
Materials Science and Engineering: A (2019)