Monkeys categorize objects in the same way as humans

Being so closely related to our own species, monkeys serve as important model organisms, and have provided many insights into the workings of the human brain. Research performed on monkeys in the past 30 years or so has, for example, been invaluable in the development of brain-machine interfaces.

Monkeys have also contributed a great deal to our understanding of the visual system -  they were the subjects in many of the classic experiments of Hubel and Wiesel, which showed that the primary visual cortex contains neurons that are responsive to edges and bars moving in specific orientations.

Yet, little is known about higher order visual processing in the monkey brain. For example, do monkeys categorize objects in the same way as we do? A new study by an international team of researchers addresses this question, and shows that the brains of monkeys and humans do indeed use a common code for categorizing objects in a hierarchical manner. 

Numerous previous research has shown that the inferior temporal (IT) cortex is involved in higher order processing of visual information, and is associated with the representation of the large scale features of compelx objects. Discrete regions of this region of the brain are known to respond selectively to different categories of objects - one particularly well characterized region of the IT is the fusiform face area, so named because it responds to faces but not to other categories of visual stimuli.

The new study, which was led by Nikolaus Kriegeskorte of the Laboratory for Brain and Cognition at the National Institutes for Mental Health, involved presenting macaque monkeys and humans with the same set of real-world images, while measuring the response pattern in the IT. The same 92 isolated images were presented in quick succession in a randomized order. The images were subdivided into a number of categories, including the faces and bodies of humans and animals and natural and artificial inanimate objects. In the humans, neuronal activity was measured using high resolution functional magnetic resonance imaging (fMRI), while in the monkeys, the activity of some 700 single IT neurons was recorded using intracellular electrodes.


"Fiber-flow" visualization of the responses of human and monkey inferior temporal lobe in response to the same set of visual stimuli. The stimuli are arranged such that the distance between each reflects the similarity in the response pattern, whereby images placed close to each other elicited a similar response. Each pair of dots represents the same stimulus and is linked by a fiber, whose thickness corresponds to the inconsistency with which that stimulus is represented between the two species
(From Kriegeskorte et al 2008).

Because the researchers used different measuring techniques to record the neural activity of the monkeys and humans, they could not directly compare the responses of each. (In the human subjects, a single voxel in fMRI corresponds to groups of thousands of neurons.) Instead, they compared the response patterns to the different stimuli within the same individual subject; by analysing these similarities and differences in object representation, they were therefore to compare how each of the objects is represented between the two species, even though the features of object representation in the two species may not directly correspond to one another. 

This analysis revealed several important similarities in the responses of the two species. First, two objects represented differently in the human IT tended also to be represented differently in the monkey IT, and vice versa. This dissimilarity in the way two objects are represented was usually largest when one of the presented objects was an animate object and the other an inanimate one, and smaller when the objects were either both animate or inanimate. Third, the dissimilarity was particularly small when the response to both human and animal faces was measured.

These data show that IT activity in both species encodes the same category boundaries and reflects very similar distinctions between objects. In both monkeys and humans, the top-level distinction is between between animate and inanimate objects, with faces and body parts forming subclusters within the category of animate objects. These distinctions are behaviourally relevant, so it seems plausible that the brain mechanisms underlying their categorization has been conserved during the evolution of the primate lineage.

Interesting differences were also found in how human and ape faces were represented. The researchers observed greater dissimilarities in the human than in the monkey representations of human faces. A selective analysis of the representation of monkey, ape and human faces in the monkey IT showed larger dissimilarities in the representations of monkey and ape faces. Thus, despite the similarities in the responses, the brain of each species seems to be especially adept at recognizing faces of the same or more closely related species.


Kriegeskorte, N. et al (2008). Matching Categorical Object Representations in Inferior Temporal Cortex of Man and Monkey. Neuron 60: 1126-1141. DOI: 10.1016/j.neuron.2008.10.043


More like this

Not sure about that last conclusion.

We would expect a human to have seen relatively few monkey faces in their lifetime, and vice versa. Greater granularity could just be more exposure. (Eskimos don't really have 40 words for snow, but skiers do have lots more than skateboarders).

I wonder how this interacts with Alan Hein's findings on training in the visual cortex, and with more recent research on arousal in different areas of the brain when presented with the faces of same/other races.