Abstract

We recently demonstrated that neural activity in the same frontal and parietal cortical areas persists when humans 1) maintain a location in working memory, 2) covertly maintain attention peripherally, and 3) maintain a spatially directed motor intention. We concluded that spatial working memory, attention, and intentions share a common neural mechanism that is implemented in these areas. To further test these conclusions, here, we use multivoxel pattern classification of fMRI data to test two hypotheses. First, can we predict the position of a working memory representation, the direction of covert attention, and the target of a motor intention based on the multivariate pattern of delay period activity. Indeed, we find that frontal and parietal cortex activity can correctly classify whether subjects are remembering, attending, and planning a movement to the right or left hemifields. Second, we predict that these classifiers can even predict the correct hemifield across tasks. We trained classifiers on one task (e.g., working memory) and tested its performance on the other tasks (e.g., spatial attention and motor intention). Remarkably, despite that subjects were performing a different task, we observe robust cross-task classification. A classifier trained to discriminate the position of a working memory representation can predict the direction of one’s attention and the goal of one’s motor intentions. These results suggest that the information contained within these areas during delay periods is not dependent on working memory, attention, or intentions. Instead, it argues that these areas implement a common mechanism that supports a variety of spatial cognitions.

From: Riggall, A.C., Ikkai, A., Srimal, R. & Curtis, C.E. (2008). Predicting the direction of spatial attention, working memory, and motor intentions using multivoxel pattern classification. Society for Neuroscience, Washington DC.