Meeting of the Mind

Brain scan to text

Thanks to “Matching Images of Brain Activity with Complex thought,” a post in Science debate where we discuss science, which points to a PDF in Frontiers in human neuroscience, we have a report of Princeton psychology researchers generating one-word text from fMRIs of subjects reading words representing concrete items (e.g., “door,” “window,” “apartment.”) Given the predilection of the developing brain for single-word concrete expressions, this would seem an appropriate occasion in developing research. Given the predilection of library patrons for single-word searches, perhaps a neuropsychological library application is not that far off.

Reconstructing images from brain scans of the viewer is approached with a “coding” model at the Gallant Lab at the University of California at Berkeley (the “Research” section is a good “about” page, but their publications are of course more detailed.)  Instead of trying to extend the interpretation process from brain scan to text, these researchers show subjects a movie clip, and then see what sense they can make of the fMRI scan.  This seems more “honest” about the capabilities of the technology than trying to discern words from the process of subjects’ reading.  Their Web site points out that fMRI measures hemodynamic, not neural, activity, so detailed reading seems unlikely.  There is a YouTube video of an example of the lab’s work.