Carnegie Mellon University researchers have built a computer model that can predict which concrete noun (things that one can see, hear, feel, taste or smell) one is thinking about based on brain scans.

The model was trained to look at how words relate to sensory and movement information (“hammer” with movement areas), and how nouns are associated with 25 basic verbs (“celery” but not “airplane” with “eat”).

Volunteers were scanned with fMRI as they looked at nouns. After the model was trained, it could predict which pattern of brain activity was associated with two nouns (after being trained on 58 other nouns). Later, given a pattern of brain activity, it could also rank which nouns matched the pattern. Its performance was better than chance in both tests.

A “brain-reading” technique reported earlier in 2008 could predict what picture a person was seeing from a selection of more than 100. The new model is different in that it has to look at the meanings of the words, rather than just lower-level visual features of a picture.

See Also Mind Reading with Functional MRI