Purchase this article with an account.
Ben Vermaercke, Hans Op De Beeck; Invariant behavioural templates for object recognition in humans and rats. Journal of Vision 2010;10(7):1013. doi: 10.1167/10.7.1013.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The human visual system is expert in object recognition. We can identify objects under different viewing conditions, that is, object recognition shows a great deal of invariance. How the brain accomplishes this complex task is puzzling neuroscientists. Many studies have applied methods that visualize linear relationships between image properties and performance, such as classification images, ‘bubbles’, and reverse correlation. However, the existence of invariance means that simple relationships between image properties and performance do not exist, except in artificial experimental situations. The validity of results obtained in such situations is not clear. This problem is all the more relevant in studies of other animals that might be prone to rely on simple strategies. Here we extended the bubble technique to explicitly study invariant object recognition in humans as well as rats. We trained humans and five Brown-Norway rats to discriminate two simple shapes (square vs triangle) that were partially occluded with bubbles. At first, the shapes had a fixed position. Behavioural templates from these data were complex, also for rats, and consisted of a few spatially separated spots. Then these shapes were shown at random screen positions to prevent any simple relationship between the content of specific pixels and the stimulus. As expected, we no longer obtained clear behavioural templates with traditional classification image analyses. However, if we adapt the analyses by taking the position shift of the stimulus into account and normalizing the position of bubbles for the position of the stimulus, then again a behavioural template is found - for both species. We conclude that methods to visualize behavioural templates can be adapted to include the full complexity and inherent nonlinearity of object recognition, and allow the investigation of these nonlinearities in humans and even in species that are not typically considered as being ‘visual’.
This PDF is available to Subscribers Only