Summary
Attending to informative stimuli is critical for choosing appropriate actions, but the biological mechanisms
underlying this ability are not well understood. Recent theoretical models of computational rationality propose
that an executive meta-level controller regulates the precision of a cognitive process (e.g., learning or memory)
based on the estimated costs and benefits of the process in a particular context. However, current models have
yet to account for the control of attention to obtain information (reduce uncertainty) and have been limited to
mathematical formalisms rather than biologically plausible architectures. In a Targeted BRAIN Circuits R01, we
will fill these gaps by investigating the cellular and neuromodulatory mechanisms of attention control in visual
and executive structures – i.e., the monkey lateral intraparietal area (LIP) and anterior cingulate cortex (ACC) –
in conjunction with a neurocognitive model developed by the PIs, the attention reinforcement meta-learner (A-
RML). We will test the novel hypotheses that (1) stimuli compete for control of saccades based on their
diagnosticity (predictive accuracy about future events) and the competition is sharpened in favor of more
diagnostic predictors by higher decision uncertainty and (2) dACC and LIP, and DA and NE have distinct roles
in, respectively, the valuation and implementation of an attentional policy. In the current planning proposal, we
set the stage for this goal providing behavioral evidence for our hypothesis (Aim 1), and (2) establishing the
feasibility of using, in monkeys, novel fluorescent molecular sensors – GRAB – that have been demonstrated in
rodents to measure neurotransmitter release with much greater temporal and pharmacological specificity relative
to current techniques (Aim 2). In the full RO1 we will use these tools to obtain high precision measurements of
neuromodulator release in relation to information gathering and neural activity, validate and further extend the
A-RML based on empirical observations, and interrogating the circuit using causal manipulations. The results
will provide a novel account of attention control that spans cellular and neuromodulatory mechanisms and a
computational account linking attention with Bayesian theories of uncertainty minimization.