Reconciling age-related changes in behavioural and neural indices of human perceptual decision-making

Ageing impacts on decision making behaviour across a wide range of cognitive tasks and scenarios. Computational modeling has proven highly valuable in providing mechanistic interpretations of these age-related differences; however, the extent to which model parameter differences accurately reflect changes to the underlying neural computations has yet to be tested. Here, we measured neural signatures of decision formation as younger and older participants performed motion discrimination and contrast-change detection tasks, and compared the dynamics of these signals to key parameter estimates from fits of a prominent accumulation-to-bound model (drift diffusion) to behavioural data. Our results indicate marked discrepancies between the age-related effects observed in the model output and the neural data. Most notably, while the model predicted a higher decision boundary in older age for both tasks, the neural data indicated no such differences. To reconcile the model and neural findings, we used our neurophysiological observations as a guide to constrain and adapt the model parameters. In addition to providing better fits to behaviour on both tasks, the resultant neurally-informed models furnished novel predictions regarding other features of the neural data which were empirically validated. These included a slower mean rate of evidence accumulation amongst older adults during motion discrimination and a beneficial reduction in between-trial variability in accumulation rates on the contrast-change detection task, which was linked to more consistent attentional engagement. Our findings serve to highlight how combining human brain signal measurements with computational modelling can yield unique insights into group differences in neural mechanisms for decision making.