< Back to previous page

Publication

Visual object representations in the human cortex: Part-whole relationships and the effects of learning

Book - Dissertation

Viewing the world, we are confronted with multiple visual objects, each of which might be composed of multiple parts. How exactly our brain is able to recognize and identify these objects very quickly and easily, and how it constructs a representation of these objects is not yet clear. In this doctoral thesis, we focused on two aspects of visual object representations: (1) how do we learn to recognize objects and (2) what kind of relationship exists between the neural representations of complex visual stimuli and their constituent parts? For the first question, we studied characteristics of perceptual learning of complex objects, using a combination of behavioral and transcranial direct current stimulation (tDCS) experiments. One studied aspect was how general the learning effect is when participants are trained to recognize objects. We found that both general and specific learning processes are involved: while the learning effect is specific for the trained stimulus set, partial transfer of learning was found over different paradigms and changes in appearance of the stimuli (different exemplars and orientation). Additionally, we found that object learning benefits from sleep between training sessions and that mere exposure to the objects while performing another task is not sufficient to train object recognition. Finally, the involvement of the lateral occipital complex in object learning was directly tested using tDCS. Stimulation of lateral occipital cortex with tDCS during training resulted in a larger training effect. In the second part, we looked at what kind of relationship exists between the representations of ‘wholes’ and their constituent parts. In an fMRI study, we studied the relationship between neural responses to single objects and object pairs and the modulation of this relationship based on co-location for action and a task effect. Response patterns of object pairs were best predicted by a weighted average of the response patterns of the constituent single objects, with the maximum response weighted more than the minimum response. The max and min response were weighted differently for different types of object pairs when participants attended to the configuration. In a high-field fMRI study we investigated the effect of changing one or more parts (letters) of a word by means of a parametric manipulation of orthographic similarity in real and pseudowords. We found that the visual word form area is responsive to lexical status, but both real words and pseudowords were further processed in terms of orthographic similarity. Similarity of the underlying neural response patterns is thus related to overlap of parts (letters) between different words and word-like stimuli.
Publication year:2015
Accessibility:Closed