Rapid Modification of Populational Neuronal Response Onset Times Via Hebbian
Learning: a Non-Invasive Single-Trial Analysis of MEG Data.

A.C. Tang, S.A. Carter, B.A. Pearlmutter, N.A. Malaszenko, L.K. Anderson, C.J. Aine and R. Christner.
Abstr. Cog. Neur. Soc. 2001. 123C

We studied Hebbian learning in humans via coactivation of the visual and auditory
pathways. The experiments have three phases: baseline, learning, and extinction.
During baseline (3 blocks) and extinction (2 blocks), faces and tones are presented
alone in a random sequence (50 face and 50 tone presentations per block). During
learning (1 block), 50 paired face-tone were added randomly to the above described
standard block of single faces and tones. Using a noninvasive functional imaging
method (MEG) and a blind source separation method (SOBI), we were able to monitor,
during this Hebbian learning millisecond temporal resolution from localized brain
regions (a few mm spatial resolution). If Hebb's rule holds, co-activation of the
visual and auditory pathways should result in changes in the populational neuronal
response along each pathway alone. In 6 out of 7 subjects, we found learning induced
changes in the timing of the single-trial populational response onset. Majority of
such neuronal populations appeared to be along the dorsal visual processing pathway
in all 6 subjects. Hebbian learning is usually thought of as a change in response
amplitude. Our results offer evidence for rapid induction of populational response
timing change as a redult of a Hebbian learning protocol. Methodologically, these
results show that it is now possible to non-invasively study repidly-occuring learning
with millisecond resolution in humans.