Representational Geometry of categorical perception of animate and inanimate objects

Abstract

Visual perception begins with simple sensory information and is transformed in the brain into increasingly complex representations of recognizable objects. In order to recognize existing items, concepts - mental representations of things that exists - are necessary. These concepts are organized into classes called categories that allow humans to organize their perception, generalizing between similar items. Visual perception is only one way of accessing conceptual representations, another way is language. While language can be perceived visually as written language, its fundamental form is auditory and it also uses a different (lexical) format. Despite these differences, conceptual organization for vision and language comprehension appears to be similar. It has even been claimed that conceptual representations are independent of the modality and format they are accessed through.I investigated modality-independent conceptual representations for two modalities (visual and auditory) and two formats (pictorial and lexical) using behavioral and fMRI data and using the category of animacy as a window into conceptual organization. I carried out a behavioral experiment and two fMRI experiments expanding upon previous research by directly comparing the two modalities and formats in within-subject designs, by including a larger part of the animacy spectrum and a perceptual tasks. Categorical perception emerged in the behavioral experiment in the absence of a task suggesting categorical organization of the stimulus set, but, while animacy did seem to play a role in participants´ judgments, other factors such as the distinction between man-made and natural items and possibly real-life co-occurrence of objects also seemed to play a role. In the brain, candidate regions for modality-independent convergence zones were observed, for example the bilateral superior and middle temporal lobes and supramarginal gyri, but also some regions that have not been strongly associated with modalityindependent processing like the pre-SMA.Directly testing for categorical representations for both specifc modalities and in a modality-independent way revealed that categorical processing was not as present as previous research suggested. In most previous studies investigating this topic, a task was used that involved categorical processing which might have led to an overestimation of categorical processing effects. By varying the conceptual demands of the task, future research can investigate how top-down vs. bottom-up aspects influence both behavioral similarity judgments and brain activity.