Decoding intentions of self and others from fMRI activity patterns

Previous studies using multi-voxel pattern analysis have decoded the content of participants' delayed intentions from patterns of fMRI data. Here we investigate whether this technique can be used to decode not only participants' own intentions, but also their representation of the intentions held by other people. In other words: if Sam is thinking about Hoki, can we decode the content of Hoki's intention by scanning Sam's brain? We additionally distinguished two components of intentions: action-plans versus goals, and included novel control analyses that allowed us to distinguish intending an outcome from simply expecting it to occur or simulating its consequences. Regions of frontal, parietal, and occipital cortex contained patterns from which it was possible to decode intentions of both self and other. Furthermore, crossclasification between self and other was possible, suggesting overlap between the two. Control analyses suggested that these results reflected visuo-spatial processes by which intentions were generated in our paradigm, rather than anything special about intentions per se. There was no evidence for any representation of intentions as mental states distinct from visuospatial processes involved in generating their content and/or simulating their outcomes. These findings suggest that the brain activity patterns decoded in intention-decoding fMRI studies may reflect domain-general processes rather than being intention-specific.