Data Analysis Project Presentation / 2nd Milestone

  • Gates Hillman Centers
  • Blelloch-Skees Conference Room 8115
  • MARIYA TONEVA
  • Joint Program in Neural Computation / Machine Learning Ph.D. Student
  • Center for the Neural Basis of Cognition / Machine Learning Department
  • Carnegie Mellon University
Project Presentations

Word Length Processing via Region-to-Region Connectivity

A previous magnetoencephalography (MEG) study found that many features of stimuli can be decoded around the same time but at different places in the brain, posing the question of how information processing is coordinated among brain regions. We aim to directly relate the processing of information content in one region to the connectivity of that region. We examine whether, during presentation of a word stimulus, the length of the word relates to how strongly the region that best encodes word length - left lateraloccipital cortex (LOC-lh) - connects to other regions, at different times.

Using representational similarity analysis, we find that changes in strengths of the LOC-lh connectivity network across stimuli correlate significantly with changes in word length. This significant correlation occurs at the very end of the decodability period of word length in LOC-lh between 200-220ms post stimulus onset. We further investigate whether any individual regions contribute significantly to this correlation. We find that there are only 3 regions which have connectivity strengths with LOC-lh that rank correlate significantly with word length during this time period. All of these regions are located in the left temporal lobe. Two of them -- the left temporal pole (also known as anterior temporal lobe) and the left superiortemporal gyrus -- have previously been implicated in language processing. The temporal pole has been found to be critical for semantic memory, while the superiortemporal gyrus contains Wernicke's area and has been linked to language comprehension.

The high temporal resolution of MEG combined with the set of techniques we present can enable the tracking of information flow during language processing.

DAP Committee:
Tom Mitchell, Rob Kass, Avniel Ghuman

 

For More Information, Please Contact: