Carnegie Mellon University
Skip navigation and jump directly to page content

Jan. 28: Carnegie Mellon Researchers Receive $1.1 Million From Keck Foundation To Pursue New Breakthroughs in Learning How the Brain Works

Contact:

Anne Watzman                       
412-268-3830                       
aw16@andrew.cmu.edu

Byron Spice
412-268-9068
bspice@cs.cmu.edu

Carnegie Mellon Researchers Receive $1.1 Million From Keck Foundation
To Pursue New Breakthroughs in Learning How the Brain Works

Research To Focus on How Brain Sees Abstract Concepts Like Truth, Beauty, Justice

PITTSBURGH — Carnegie Mellon University Cognitive Neuroscience Professor Marcel Just and Computer Science Professor Tom M. Mitchell have received a three-year, $1.1 million grant from the W. M. Keck Foundation to pursue new breakthroughs in the science of brain imaging. Ultimately, this research could shed light on brain disease or conditions like autism, dyslexia or depression.

justIn the first phase of their research, funded in part by an earlier grant from the Keck Foundation from 2005 to 2007, Mitchell, Just and their colleagues combined information from functional magnetic resonance imaging (fMRI) scans with machine learning algorithms that discern patterns of brain activity to show for the first time where thought processes about a particular object originate in the human brain.

They also were able to show that when people think of familiar objects like tools or dwellings, their brains activate in several places rather than just one. In addition, when they compiled information from brain activation studies of a dozen people over a two-year period, they found that the parts of the brain activated when thinking about a particular object were very similar across different subjects

Based on these results, which were published Jan. 2 in the journal PloS One, Just and Mitchell will now seek to develop the first scientific theory ever conceived for producing testable detailed predictions of observable fMRI neural representations for every concrete noun in common English, as well as many abstract nouns and verbs.

"We've learned how the brain represents concepts that we describe by concrete words," said Mitchell, who heads the Machine Learning Department in Carnegie Mellon's School of Computer Science. "The next step is to learn how the brain thinks about abstract subjects, like art, beauty or truth. We will be looking at how the neural representation of these objects builds out of their component features."

"We're starting to crack the brain's code for concept representations," added Just, who directs the university's Center for Cognitive Brain Imaging. "We're looking at the content of thought."

mitchellJust and Mitchell will use several approaches to develop and validate their theory of neural representation of words in the English language. They plan to do computational studies of what Mitchell describes as the trillion word corpus of English text available on the World Wide Web, as well as brain imaging (fMRI) studies of the processing of words, machine learning studies of brain activation and cognitive modeling of the psychological representation of concrete concepts.

The researchers point out that the advent of several relatively new technologies makes their goal feasible. Brain imaging with MRI scanners has provided a way to non-invasively study brain activity associated with various kinds of thinking. Machine learning, a new branch of computer science, can discover subtle patterns in large data sets of these brain images. The World Wide Web provides insight into how people use language, and unified theories of cognition take brain architecture as their point of departure. Furthermore, analyses of computational linguistics have provided new ways to characterize word meaning.

"It is the convergence of these technologies at this time in history that make our goal feasible," Just said. "We are extremely grateful to the Keck Foundation for showing confidence in the value of our work."

###
 
Top photo: Marcel Just; Bottom photo: Tom Mitchell