At 130 ms you “know” where the animal is but you don't yet “know” it's a dog

Abstract

Since the influential studies by Rosch and colleagues in the 70s, it is generally agreed that the visual system can access information at the basic level (e.g., dogs) faster than at the subordinate (e.g., Chihuahua) or superordinate levels (e.g., animals). However, the advantage of the basic category over the superordinate category in object recognition has been challenged recently, and the hierarchical nature of visual categorization is now a matter of debate. In a series of psychophysical studies, we addressed this issue using a forced-choice saccadic task in which two images were displayed simultaneously on each trial and participants had to saccade as fast as possible towards the image containing the designated target category. Kirchner and Thorpe (Vision Research, 2006) previously demonstrated that successful saccades towards animal targets could be initiated with minimum latencies as short as ∼130ms″. This protocol allows us to compare the temporal dynamics of visual recognition at the basic and superordinate level within very short delays after stimulus onset. There were two major findings. First, with such very short processing time, participants' performance was impaired in categorizing an object at the basic level when compared to its superordinate level. Second, when categorizing an object among different basic levels, saccades started at the same mean latency but with a low performance accuracy that was influenced by the degree of morphological similarity between targets and non-targets (dog/bird: 61% correct; dog/cat: 46% correct). Follow-up computational modeling further confirmed that these behavioral results cannot be predicted by pure bottom-up saliency differences between the images used. Therefore, our results support a coarse-to-fine model of visual recognition. The visual system first gains access to relatively coarse visual representations which provide information at the superordinate level of an object, but additional visual analysis is required to allow more detailed categorization at the basic-level.