Neural and Computational Mechanisms of Categorisation. (360G-Wellcome-106931_Z_15_Z)

Judging a person as a friend or foe, a tumour as cancerous or benign, or a sound as an \l\ or \r\ are examples of categorisation tasks. Category knowledge provides a necessary basis for almost every cognitive act, ranging from assessing the value of an object to problem solving. For example, the perceived value of a shiny object will depend on whether it is categorised as a piece of glass, a synthetic diamond, or the Koh-i-Noor diamond. My research question is whether people's ability to acquire and use categories arises from interacting brain processes that sample aspects of the external (e.g., by eye movements) and internal (e.g., by memory retrieval) world. This recasting of basic constructs in attention and memory is carried by three interrelated aims: Firstly, A) how attention alters category representations in the brain will be investigated, then the mechanisms that give rise to such phenomena will be unpacked in terms of hypothesised B) external and C) internal sampling processe s. In contrast to the standard view, which holds that attention, memory, and decision involve distinct stages of processing, I propose that these functions are rooted in interacting sampling processes. My main hypothesis (see Figure 1) is that categorisation decisions are made by internally sampling from memory until sufficient evidence is accumulated to respond. During this limited-retrieval process, the probability that a memory is retrieved varies as a function of its recency and attentio n-weighted similarity to the external stimulus. These similarity relations can change within a decision as a result of external sampling (e.g., eye movements). External sampling can be strategic (i.e., guided by current beliefs and goals). The interplay of these internal and external sampling processes gives rise to attentional effects and the apparent dynamic character of human knowledge. My main hypothesis concerns sampling processes within a decision episode that shape category representa tions over time. Standard paradigms (e.g., factorial design and comparisons of conditions) are not suited to evaluating how brain state and category representations change over time because these methods fail to specify or evaluate the underlying mechanism. Instead, my team will develop cognitive models that describe how processes unfold and representations change over learning and relate these hypothesised mechanisms to multivariate patterns of neural activity in imaging studies where participa nts learn about novel categories. The integration of behavioural measures (e.g., choice, response time, eye movements), neural measures (e.g., fMRI), and cognitive modelling will provide the necessary theoretical constraints. This formal integration is made possible by the novel cognitive modelling and multivariate analyses developed in this proposal, which will help explain how the brain acquires and uses categories.