Integrating modality-specific expectancies for the deployment of spatial attention

The deployment of spatial attention is highly sensitive to stimulus predictability. Despite evidence for strong crossmodal links in spatial attentional systems, it remains to be elucidated how concurrent but divergent predictions for targets in different sensory modalities are integrated. In a series of behavioral studies, we investigated the processing of modality-specific expectancies using a multimodal cueing paradigm in which auditory cues predicted the location of visual or tactile targets with modality-specific cue predictability. The cue predictability for visual and tactile targets was manipulated independently. A Bayesian ideal observer model with a weighting factor was applied to trial-wise individual response speed to investigate how the two probabilistic contexts are integrated. Results showed that the degree of integration depended on the level of predictability and on the divergence of the modality-specific probabilistic contexts (Experiments 1–2). However, when the two probabilistic contexts were matched in their level of predictability and were highly divergent (Experiment 3), higher separate processing was favored, especially when visual targets were processed. These findings suggest that modality-specific predictions are flexibly integrated according to their reliability, supporting the hypothesis of separate modality-specific attentional systems that are however linked to guarantee an efficient deployment of spatial attention across the senses.