Current Projects

Mark Seidenberg

My research probably appears to have a bit of a split personality. One of my main foci is language: what kinds of knowledge underlie the use of language, how this information is acquired, and how it is represented in the brain. I've been interested in these issues for a long time: my first journal publication was a critique of the original chimpanzee language projects (anyone remember Washoe? Cognition, 1979); more recently I've been interested in the nature of morphological knowledge (inflectional and derivational), and lately I've been engaged in debates about "statistical" vs. "grammatical" mechanisms in language learning.

But, a lot of my work has been about reading, a particular use of language. Many people who study language don't think reading is very interesting because it's derivative of spoken language; people are born with UG, it's said, not knowing how to read. It's therefore a little ironic that much of what we know about the structure of language comes from...studies of reading. I think reading is interesting in its own right; we can argue about whether animals have any of the components of language and to what degree, but none of them have anything remotely like reading. Reading and writing are relatively recent inventions, which made use of existing capacities, and so reading provides an interesting domain in which to explore questions about learning, perception, memory and the rest of cognition. Finally, reading is interesting because of its importance to society. If we understood reading better perhaps we could develop better ways of teaching it and helping individuals who find learning to read difficult.

I see these aspects of language as intimately related; they are governed by the same principles concerning knowledge representation, processing, and learning. So I don't draw as sharp a line between the areas as others might.

At the moment, I am focusing on the following specific topics, in collaboration with students in the lab and researchers at other institutions:

• Dyslexia. A great deal has been learned about dyslexia over the past 25 years or so, and yet basic questions about etiology, co-morbidity, and variability remain unanswered. Most dyslexics exhibit impaired representation of information at the phonemic level, and this has a definite impact on learning to read (see the Harm & Seidenberg 99 model for evidence). But, what is the cause of the phonological impairment? A subtle problem with speech perception? A visual processing problem that affects the development of phonemic representations? Noisy neural processing in brain areas relevant to reading? We don't really know. My own interest is in trying to link behavioral aspects of dyslexia to underlying causes specified at different levels, ranging from genetic polymorphisms to brain development, to computational effects, to observed behavior. Some of this work is being conducted with the Pugh-Rueckl-Mencl research group at Haskins Labs.


• Plasticity and the bases of age-dependent constraints on learning. This research examines possible computational bases for changes in the capacity to learn as a function of age. We have developed a new account of phenomena thought to support the existence of a biologically-determined critical period for language acquisition. In the Zevin and Seidenberg (in press) article we discuss a theory in which "entrenchment" of a first language interferes with learning a second language; these effects occur for specifiable technical reasons in neural networks.


• Statistical bases of language learning and processing. It's now clear that babies engage in statistical learning, and that this is a general capacity not limited to language. Questions abound, however, about the nature of statistical learning: which statistics are computed and why? Does grammar start where statistics stop, as we asked in a Science article? Or is it statistics all the way down (to the perceptual bottom)? Are babies Bayesian learners, or do they just seem that way sometimes? Students are conducting both empirical studies with babies (and adults) and linking them to computational models that explore the mechanisms underlying learning.


• The "achievement gap". Poor and minority children (including African Americans, Latinos, Native Americans and others) are at high risk for poor performance in school. The gap between these groups and other, more advantaged groups has been given a name: the achievement gap. Our research group, with funding from the Wisconsin Institutes for Discovery, is focusing on factors that work against African American children's success in school, particularly in the early grades and in learning to read. Many factors contribute to these children's relatively poor performance; an important one is the mismatch between their native dialect (African American English) and the dialect used in school (a version of "Standard" or "Mainstream" American English. This match makes it harder for the African American child to benefit from school experiences, and it makes tasks such as learning to read literally more difficult. Our project is examining young, preschool children's knowledge of the alternative dialects and ways to minimize the dialect clash before it can occur. Progress in this area would be of enormous importance to the children and to American society; the research also provides a way of examining questions about neural plasticity for language and how children learn to read.


• Neural mechanisms of language comprehension. With Jeff Binder and his research group at the Medical College of Wisconsin, I am collaborating on studies of the brain bases of language comprehension, using fMRI. We are examining issues about the development of lexical expertise, the relationship between literal and figurative uses of language, and orthographic representations.