Skip to main content

Michael Ullman


Director, Brain and Language Laboratory; co-Director, Center for the Brain Basis of Cognition




Our research program probes the neural, computational and psychological bases of language and memory. Most of the research focuses on two fundamental capacities of language: the “mental lexicon” (mental dictionary) of memorized words, and the “mental grammar,” which underlies the rule-governed sequential and hierarchical combination of lexical forms into complex words (e.g., walk + -ed), phrases and sentences (e.g., Clementina excoriated the pachyderm). Previous neurocognitive models of lexicon and grammar (and of language more generally) have had difficulty integrating knowledge across the relevant disciplines, in particular neuroscience, psychology, and linguistics. According to the “declarative/procedural” model that we have proposed, lexicon and grammar depend respectively upon two well-studied brain memory systems. Lexical memory depends on declarative memory, which underlies the learning and use of fact and event knowledge. Declarative memory may be computationally specialized for learning arbitrary relations, and is rooted largely in temporal lobe structures. Aspects of grammar involve procedural memory. This distinct brain system subserves the acquisition and expression of motor and cognitive skills (e.g., riding a bicycle), may be specialized for sequences, and is rooted largely in frontal/basal-ganglia structures.

We have examined the predictions of this novel model, in comparison to those of competing models, with numerous approaches. (1) Psycholinguistic and other behavioral experiments. These studies examine the effects on language processing of a variety of factors, such as word frequency, phonological neighborhoods, working memory capacity, and the interference of non-linguistic declarative and procedural memory tasks. (2) Neuropsychological studies. These test patterns of spared and impaired linguistic and non-linguistic function in developmental and adult-onset disorders, including Williams syndrome, Specific Language Impairment, phenylketonuria, autism, Tourette syndrome, anterior and posterior aphasia, cerebellar degeneration, and Alzheimer’s, Parkinson’s and Huntington’s diseases. (3) Neuroimaging experiments, primarily using Electroencephalography/Event-Related Potentials (EEG/ERPs)functional Magnetic Resonance Imaging (fMRI). These studies probe the neurocognitive correlates of various domains of language, including morphology (the structure of words), syntax (the structure of phrases and sentences), and compositional semantics (interpretive aspects of the composition of words into complex structures). We examine these domains in both children and adults and in multiple languages, including English, Italian, Spanish and Japanese. We have argued that converging evidence supports the declarative/procedural model, providing a new neurocognitive framework for the study of language.

Other projects, three of which are described here, suggest extensions of the declarative/procedural model with potentially important educational and clinical implications.

Sex differences. Previous studies have shown that women are superior to men at remembering new words, and that this ability likely depends on declarative memory and estrogen. We predicted and have found that women tend to store, in lexical/declarative memory, complex linguistic forms that men normally compute on-line in the grammatical/procedural system (e.g., walked). Moreover, our evidence suggests that this effect is modulated at least in part by estrogen. Converging evidence has thus far been obtained from seven experiments, using psycholinguistic, neuropsychological, and neuroimaging techniques, including a hormone replacement study, in English and Spanish, testing morphology and syntax, in both expressive and receptive language.

Second Language. Linguistic abilities, in both first and second language (L1 and L2), are sensitive to the age of initial language exposure. However, late exposure may affect grammatical more than lexical functions. We argue for a compensatory shift: at least initially (at lower experience and proficiency), complex linguistic representations that may be compositionally computed by the grammatical/procedural system in L1 tend to depend on lexical/declarative memory in L2, likely due to age-related attenuation of the grammatical/procedural system. However, data suggest that at least some grammatical abilities that depend on the grammatical/procedural system in L1 can be learned and used by this system in high-proficiency L2 learners. This may correspond to the finding that practice on procedural memory-based motor skill tasks leads to improved performance in adults. In sum, we posit that at least certain complex representations that may be composed in real-time by the grammatical/procedural system in L1 tend to be similarly computed in high-proficiency L2, but stored in lexical/declarative memory in low-proficiency L2. A range of psycholinguistic, neuropsychological and neuroimaging studies support these predictions.

Recovery of function. A dysfunction of the grammatical/procedural system might lead to the subsequent lexical/declarative memorization of complex forms (e.g., walked). We have observed such functional reorganization in two distinct populations of the developmental disorder of Specific Language Impairment, and in the adult-onset disorder of anterior aphasia. These findings lead to intriguing neuro-pharmacological implications. Given that the biochemical bases of the two brain memory systems are relatively well understood, and that existing pharmacological agents have been shown to affect them, it may be possible to treat certain “language” disorders with these pharmacological agents.