Reinterpreting decision boundary height in sequential sampling models as task attention

Sequential sampling models (SSMs) are widely used in studies of cognitive control and decision-making to decompose the cognitive and neural mechanisms underlying behavior. Of particular interest in the application of SSMs are parameters that are presumed to be sensitive to deliberate control, particularly drift rate and decision boundary/response threshold. Condition- and group-dependent differences in these parameters are routinely observed across a wide range of studies and contexts, leading to their use as sufficient explanations for behavioral and neural effects in cognitive psychology and cognitive neuroscience. These explanations generally correspond to a widespread convention that changes in drift rate relate to attentional processes, while changes in the decision boundary are related to response execution or response inhibition. Although considerable evidence relates neural activity with drift rate parameters, there is substantially less empirical support for direct modulation of the decision boundaries predicted by SSM approaches, and frequently this evidence only indirectly supports decision boundary shifts through mathematically equivalent mechanisms such as baseline shifts or changes in noise. In this manuscript, I provide an alternative interpretation of decision boundary height as the overall level of attention allocated to task-relevant information. By reformulating standard SSM models to allow attention to be diverted toward a non-informative noise component in addition to task cues, I show that withdrawing attention from task cues produces behavioral effects mathematically identical to reducing the effective decision boundary. In this framework, total task attention determines effective boundary height, while the relative allocation of attention across task cues determines drift rate. Simulations of a reliability-based model of learned attention show how controlled withdrawal of attention from unreliable task cues recapitulates classic boundary effects, providing a unified attentional account of cognitive control in SSMs.