The interaction between reward-driven learning and selective attention

Selective attention is the prioritisation of certain pieces of information over others, and is often guided by task goals or by physical salience. Attention can also be guided by stimuli previously learned to be associated with reward which can capture attention in the absence of task relevance or physical salienceâa phenomenon termed value-driven attentional capture (VDAC; reviewed in Chapter 1). The first part of this thesis examines the cognitive and neural mechanisms of VDAC. Chapter 2 shows that stimuli with reward history can rapidly capture spatial attention without being currently task-relevant (akin to exogenous cues) and can also sustain attention over intervals as long as one second. Chapter 3 examines the neural mechanisms of VDAC using multivariate pattern analysis of magnetoencephalography data. Learned value modulates spatial representations of stimuli within posterior visual, parietal, and inferior temporal cortex, and this relates to the magnitude of the behavioural VDAC effect. No evidence was found suggesting that non-spatial features are similarly modulated. A stimulus-independent value signal was also found in insular and posterior visual and parietal cortex. However, the relatively late onset of these modulations during visual processing (~250ms) suggests that value signals may first arise elsewhere. The second part of this thesis describes a behavioural and functional magnetic resonance imaging (fMRI) study in macaques which examines the impact of stimuli with reward history on reward learning and decision-making, and the brain regions that represent reward history and other parameters relevant to these processes. Chapter 4 shows that the reward history of a currently unrewarded option interferes with decision-making, possibly by slowing subjectsâ learning rates. Chapter 5 describes fMRI analyses showing that regions within the ventromedial prefrontal cortex (vmPFC) and ventrolateral prefrontal cortex (vlPFC) track the learned value of stimuli. Analyses also revealed activity consistent with value comparison signals in the vmPFC, reward prediction error (RPE) signals in the amygdala and putamen, and unsigned RPE (âattentionâ) signals in the putamen, anterior cingulate cortex, and the vlPFC. However, fMRI analyses were not able to clearly explain the impact of reward history in the current study; possible reasons for this are discussed. Chapter 6 summarises the findings and their limitations and places them in the broader context of the interaction between learning and attention.