Cookies on this website
We use cookies to ensure that we give you the best experience on our website. If you click 'Continue' we'll assume that you are happy to receive all cookies and you won't see this message again. Click 'Find out more' for information on how to change your cookie settings.

The significance of prefrontal cortex for reward-guided choice is well known from both human imaging and animal neurophysiology studies. However, dialogue between human and animal research remains limited by difficulties in relating observations made across different techniques. A unified modelling framework may help reconcile these data. We have previously used attractor network models to demonstrate that varying decision values can elicit changes in local field potentials (LFP) dynamics, causing value correlates observable with human magnetoencephalography (Hunt et al., Nature Neuroscience 2012). Extended, hierarchical forms of such models can also predict human functional MRI signal in different frames of reference during a multi-attribute decision process (Hunt et al., Nature Neuroscience 2014). In light of this framework, we have recently sought to relate simultaneously recorded LFP and single-unit data from prefrontal cortex of macaque monkeys performing a cost-benefit decision task. By performing principal component analysis of unfiltered LFPs timelocked to choice, components emerged that resembled the main choice-related LFP signature (PC1) and its temporal derivative (PC2). PC1 thus indexes LFP event-related amplitude, but crucially PC2 indexes its latency, reflecting the speed at which choice dynamics occurred on each individual trial. We found PC1 scores were correlated with overall value sum: the inputs of a decision process. PC2 scores, however, were correlated with chosen (but not unchosen) value: the outputs of a decision process. To relate LFP dynamics to single unit activity, we regressed single-trial PC1 and PC2 scores onto simultaneously recorded single-unit firing rates. PC2, indexing the internal latency of each individual choice, predicted single-unit activity in nearly half of all recorded units – in particular those cells that showed a value-to-choice transformation. Similar results could be found by principal component decomposition of the attractor network model. These results provide a novel bridge between LFP single-trial dynamics and simultaneously recorded single-unit activity. Perhaps more significantly, it allows us to relate value correlates in human neuroimaging studies to their cellular origins.

Host: Maryann Noonan