How the Brain Computes Language
Gina R. Kuperberg, MD, PhD, joined the Department of Psychology in 2006. Kuperberg’s research focuses on the cognitive neuroscience of language, thought, and meaning, and on how these processes may go awry in schizophrenia—a major, disabling mental illness. Her research group studies volunteers with and without schizophrenia using the complementary techniques of event-related potentials (ERPs), which tell you when specific activity is going on in the brain, and functional magnetic resonance imaging (fMRI), which tells you where specific activity is going on in the brain.
Kuperberg earned her MD at St. Bartholomew’s Medical School, London, and completed an internship at St. Bartholomew’s Hospital and a residency and fellowship in psychiatry at the Maudsley Hospital and Institute of Psychiatry, London. She earned her PhD in psychology and cognitive neuroscience at King’s College, University of London. Kuperberg completed research fellowships in neuroimaging and cognitive electrophysiology at the Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, and the NeuroCognition Lab, Tufts University. She is currently an associate professor in the Department of Psychology at Tufts. She also holds positions in the psychiatry departments of Massachusetts General Hospital and Tufts Medical Center.
Kuperberg collaborates closely with colleagues in the NeuroCognition Lab and the Martinos Center for Biomedical Imaging. “I’m really interested in how language is computed by the brain—how you’re extracting the gist of what I say,” comments Kuperberg. “I’m interested in the structure of the language as a means to get to the endpoint—the meaning. And I’m interested in applying that knowledge to the disorder of schizophrenia, which, in my mind, is a disorder of thought.”
“We have multiple projects in our lab, but there are some overarching themes,” says Kuperberg. “How do we make sense of language? How are the meanings of words computed? How are they combined to form sentences, and then how are those sentences combined in real time to form discourse, or whole meaning structure? How do we travel through time and space in language? How do we compute the emotional meaning of language?”
Many of the research projects undertaken in Kuperberg’s lab require the creation of very specific visual or auditory stimuli. The test and control stimuli must be as similar as possible except for the question under investigation. “Suppose we’re investigating the emotional meaning of language,” says Kuperberg. “We have to make sure we get all these emotional words, and embed them into sentences which don’t differ on anything other than emotionality.” Test participants are presented with the stimuli, and ERP or fMRI are recorded, analyzed, and interpreted.
Kuperberg is also extending some of the principles of language comprehension to real-world visual comprehension. One research project, conducted in collaboration with lab member Tatiana Sitnikova, compared the ERPs of volunteers with and without schizophrenia while they were viewing congruous or anomalous silent video clips. A congruous video clip might be a man putting shaving cream on his face and then shaving, whereas an anomalous video clip might be a man putting shaving cream on his face and then stroking his face with a rolling pin. The ERP data suggest that when participants without schizophrenia viewed the anomalous video clip they first attempted to match what they saw with their stored knowledge, and when that didn’t make sense, they undertook a second-pass analysis to understand the clip. Participants with schizophrenia did not appear to take the second-pass analysis; they relied solely on their stored knowledge. Dr. Kuperberg has also observed this lack of reanalysis in schizophrenia during language comprehension, and has hypothesized that it might lead to the disorganized, associatively driven language of schizophrenia. In real-world comprehension, it may lead to misinterpretation and possible delusions.
In addition to her close collaboration with Phillip Holcomb in the Department of Psychology, Kuperberg is also currently collaborating with Ray Jackendoff of the philosophy department. Some of Kuperberg’s ERP work lends itself to interpretations that would be consistent with Jackendoff’s theory of the parallel architecture of language (described in Jackendoff’s profile in this issue of Research News). One collaboration with Jackendoff is on semantic coercion, which describes how the brain fills in the blanks with meaning. For example, in the phrase the woman began the book, our brains compute the most logical meaning by adding the missing word reading, making the understanding of the phrase the woman began reading the book. In linguistics terminology, the word began is semantically coerced such that its meaning is shifted to began reading. Kuperberg’s research group is comparing ERPs of study participants reading coerced and noncoerced sentences to determine such things as time costs for neural processing of coerced sentences.
Much of Kuperberg’s work since joining the Department of Psychology has been on exploring normal language processing. To expand her work in schizophrenia, she would like to partner with a clinician. Kuperberg expects that the results of her research will have long-term clinical applications, such as in the discovery of brain-based biomarkers that will facilitate the development of drugs that specifically target thought abnormalities in schizophrenia, and, possibly, in the identification of genes contributing to schizophrenia.
The NeuroCognition Lab on the Medford/Somerville Campus has three facilities for measuring ERPs. For fMRI the research group uses facilities at Massachusetts General Hospital. Tufts researchers interested in learning more about these techniques should contact Kuperberg at Gina.Kuperberg@tufts.edu.
For more information on Gina Kuperberg and her research, please go to