Linear and nonlinear effects of Spatial and Temporal attention: human data and drift diffusion model

Attention can be affected by expectations about where and when stimuli will appear. These two influences may interact with each other in complex ways. To explore these possibilities, we used an attention paradigm that manipulated three factors: two explicit cues guiding anticipatory attention before each trial, and implicit temporal statistics learned during task performance. When analyzing the data for linear trends, we found that each of these factors impacted performance but did not interact. When analyzing nonlinear trends in the data, we found these factors interacted: expectations about task duration shaped spatiotemporal attention while performing the task. We evaluated whether attentional effects across spatial and temporal domains could be explained by one unitary mechanism, using an augmented drift diffusion model. In the model, attention is allocated according to an interplay between the costs and benefits of maintaining attention. The model successfully replicated the linear effects observed in the human data, and accounted for some but not all nonlinear features in the data, suggesting that many disparate features of attention can be parsimoniously explained by a cost-benefit framework.