Michael T. Ullman - US grants

Human language depends upon a dictionary of words (the mental lexicon) and rules which specify how those words are put together into larger words, phrases, and sentences (the mental grammar). The computational and brain bases of these two fundamental language capacities are still not well understood. This project investigates the novel view that the learning and use of the two language capacities depends upon two more general-purpose memory systems whose neural foundations are relatively well studied. Specifically, it is hypothesized that the acquisition and use of words depends upon "declarative memory" temporal-lobe brain circuits previously shown to underlie the learning and use of fact knowledge, such as what a zebra is and where it lives. It is also hypothesized that the acquisition and use of grammatical rules depends upon "procedural memory" left frontal/basal-ganglia brain circuits previously shown to underlie the learning and expression of motor and cognitive "skills," such as riding a bicycle. On this "declarative/procedural" view of language, lexicon and grammar are linked to distinct general-purpose systems, each responsible for a particular set of language and non-language functions. The declarative/procedural view contrasts with the two main classes of previously proposed theories regarding the computational and neural mechanisms of language. It shares the perspective of traditional "dual-system" theories that lexicon and grammar depend upon distinct systems, but it diverges from their claim that these systems are responsible for language alone. Conversely, while the declarative/procedural view agrees with "single-system" theories that the two language capacities depend upon general-purpose brain circuitry, it diverges from their claim that both capacities are linked to a single system with a broad distribution in the brain.

To distinguish the three views, the project will focus on a simple distinction in language: between regular and irregular transformations in past tense and in other aspects of language. In this distinction lexicon and grammar can be contrasted, while other psychologically relevant factors are held constant. Irregular transformations are at least partially arbitrary (e.g., sing-sang, bring-brought), and therefore must depend upon memory, whereas regular transformations are predictable (e.g., verb + -ed), and so could be computed by a grammatical rule. Thus links are predicted among irregulars (lexicon), facts, and temporal-lobe circuits, and among regulars (grammar), skills, and frontal/basal-ganglia circuits. Patients with brain damage limited to particular temporal-lobe areas (e.g., "posterior aphasics") should have more trouble with irregulars than regulars, and with facts than skills. Patients with damage limited to particular left frontal/basal-ganglia structures (e.g., "anterior aphasics") should show the opposite pattern.

If indeed the declarative and procedural memory systems play a role in lexicon and grammar, then data from numerous investigations of the neural, computational, and developmental underpinnings of the two brain memory systems in humans and in animals will contribute to our understanding of language. Conversely, new approaches to the study of the two memory systems will become possible on the basis of our knowledge of language structure, processing, and acquisition. Tests of lexicon and grammar, such as those used in the proposed project, may become useful probes of the two brain systems. In addition to the advancement of fundamental knowledge in cognitive science, the contributions of animal and human studies of the two brain systems to our understanding of language should lead to new clinical approaches for the diagnosis and treatment of acquired and developmental language disorders.

The long-term objective of this project is to understand the brain bases of the mental lexicon, which contains memorized words, and the mental grammar, which contains rules that combine lexical forms into larger words, phrases, and sentences. We propose that the memorization and use of words is subserved by temporal-lobe circuits previously implicated in the learning and use of fact knowledge, whereas the acquisition and use of grammatical rules is subserved by frontal/basal-ganglia circuits previously implicated in the learning and expression of motor, perceptual, and cognitive "skills," such as riding a bicycle. Thus we posit that lexicon and grammar are linked to distinct brain systems, each of which is domain-general in that it subserves non-language as well as language domains. This novel view contrasts with the two main competing theoretical frameworks. Although we share the perspective of traditional dual-system theories in positing that lexicon and grammar are subserved by distinct systems, we diverge from these theories where they assume components dedicated (domain-specific) to each of the two capacities. Conversely, while we share with single-system theories the view that the two capacities are subserved by domain-general circuitry, we diverge from them where they link both capacities to a single system with broad anatomic distribution. To distinguish our theory from the other two, we will probe the brain bases of irregular and regular word transformations, in which lexicon and grammar can be contrasted, while other factors are held constant. Irregular forms (e.g., dig-dug) are retrieved from memory, whereas regular forms (e.g., look-looked) require a suffixation rule. We predict, and have found in our preliminary studies, links among irregulars (lexicon), facts, and temporal- lobe circuits, and among regulars (grammar), skills, and frontal/basal-ganglia circuits. Single-system models do not make this set of predictions, and traditional dual-system theories do not predict the links with fact and skill use. Patients with either temporal-lobe or frontal/basal-ganglia damage will be given tasks probing the production and judgment of irregular and regular past tense inflection, plural inflection (mice, bees), and derivational morphology (solemnity, awkwardness), as well as measures of fact and skill use. Our specific aims are to test three hypotheses by probing for double dissociations between irregulars and regulars, and between facts and skills: (1) Lexicon is linked to temporal-lobe circuits, and grammar to frontal/basal-ganglia circuits. (2) These circuits also subserve fact and skill use, respectively. (3) The basal ganglia play a similar role in motor activity and grammatical rule use.

This dissertation project investigates specific aspects of the mental computation of verbal morphology in speakers of Spanish as a first language, and speakers of English as a second language. The project is framed within a research program directed by Michael Ullman. The program investigates the computational and neural underpinnings of two capacities that characterize language: the mental lexicon of memorized words, and the mental grammar of rules that specify how lexical forms combine into larger sequences. It tests the 'dual-system' hypothesis that the lexicon is subserved by associative memory, whereas a symbol-manipulation system underlies grammatical combinations. In contrast, 'single-system' theories posit that all words and rules depend upon associative memory.

Evidence supporting a dual-system view has come from psycholinguistic and neurolinguistic studies of morphology. These have shown that irregular forms (e.g. 'dug') are lexically memorized, while regular forms (e.g. 'walked') are computed by grammatical rules. Most studies have examined this distinction in English past-tense inflection, in native English speakers. This project will test the generality of the results in (1) Spanish verbal morphology, which differs substantially from English morphology, in native Spanish speakers; and (2) English past-tense morphology, in Spanish and Chinese native-speaking adult learners of English as a second language.

DESCRIPTION (provided by applicant): The long-term objective of this project is to further elucidate the neurocognitive underpinnings of the distinction in language between idiosyncratic and rule-governed mappings. The project examines four issues - separability, computation/representation, domain-specificity/-generality, and neural correlates - in the context of competing neurocognitive models of language. Dual-system models claim that idiosyncratic mappings are stored in the mental lexicon, while the mental grammar underlies the rule-governed composition of complex linguistic representations. One such view - the declarative/procedural model - argues that the lexicon depends on the temporal-lobe- declarative memory system, whereas grammar involves the frontal/basal-ganglia procedural memory system. Single-system models posit one computational system, denying a categorical distinction between the two types of mapping. Previous psycholinguistic, developmental, neurological and neuroimaging studies, largely on regular/irregular morphology, have begun to tease apart the models. We propose to replicate and extend the research in two dimensions. (1) Linguistic domains: We will examine morpho-phonology (in regular/irregular past tense, past participle, plural, and derivational morphology), syntax (in the complement/adjunct distinction, and in the hierarchy of morphosyntactic functional categories), and compositional semantics (in the definiteness effect, and negative polarity items). (2) Subject groups: Preliminary data suggest that men and women differ in their relative reliance on the two brain systems: Whereas men tend to compose complex linguistic representations, women tend to memorize them. Thus we introduce sex as an important design factor. Specific aims: We will test the predictions of the competing models with two complementary studies: (i) a neuroimaging study of cognitively unimpaired adults, using ERPs, fMRI and dipole modeling;and (ii) a psycholinguistic and neurological study of healthy adults and of adult-onset brain-damaged patients.

Is it possible for adults to learn a second language as well as a native speaker? Can they learn both vocabulary and grammar equally well? How does the way adults learn second languages differ from how children learn language? Do adults learn second languages differently in different situations, such as in a classroom or a foreign country? Previous research from cognitive neuroscience (that is, the integration of neuroscience and cognitive psychology) suggests that adults learn second language differently than children, but that adults may be capable of achieving high levels of proficiency with sufficient practice. Research by applied linguists has established that the learning situation can affect at least how quickly an adult is able to learn a second language.
With support from the National Science Foundation, Dr. Michael Ullman and Kara Morgan-Short, M.A.T., will use some innovative methods drawing from these two related fields of research in order to further explore such phenomena. The project will examine which brain systems are involved when adults learn second languages, if these brain systems are similar to or different from the brain systems that children use, and if they are affected by explicit or implicit learning situations (that is, taught directly or picked up unconsciously). The research has broad impacts, including implications for second language teaching. The research is expected to benefit government, academic, and private institutions that are concerned with successfully training adults to become highly proficient in a second language, a vital need given today's international socio-political situation. This project will take place in the context of Ms. Morgan-Short's dissertation studies and will provide interdisciplinary educational opportunities for students who participate in conducting research.

It is frequently observed that children learn languages better than adults do. Why should this be so? With National Science Foundation support, Ms. Harriet Bowden will investigate several interconnected issues related to this question for her doctoral dissertation under the direction of Dr. Michael Ullman. Is all of language equally affected by late acquisition, or are certain parts of language (such as "irregular" as compared to "regular" forms in language, or grammar versus vocabulary learning) more susceptible than other aspects of language? Does the brain use the same memory systems in adult language learning as in learning during childhood? How does language experience (that is, proficiency or amount of exposure) affect the way the brain processes language? Do very proficient late learners process their second language like native speakers of that language? How does a later age of acquisition affect language learning and processing (that is, what are the effects of a so-called critical or sensitive period)? A combination of language production measures and neuroimaging techniques will be used to probe the processing of late-acquired language. The research project will compare performance levels and neurocognitive bases of native Spanish with those of late-learned Spanish at three levels of language experience/proficiency (low, medium and very high). This will be the first project to investigate the behavioral and neurocognitive bases of late-acquired language across different levels of experience and proficiency in this manner.

The research has broader impacts in several key areas. First, the project investigates Spanish, a minority language of importance in the United States, as both a first and second language. Empirical evidence across native and later-learned Spanish will complement the vast array of existing data on English as a first and second language. In addition, the results are expected to inform cognitive neuroscience, second language acquisition, psycholinguistics, and linguistic theory. By shedding light on the neurocognition of second language use, this research may have important implications for second language instruction, and may benefit a variety of public and private institutions interested in training proficient users of second languages.

This dissertation project tests whether adults who are bilingual (who learned two languages as children) are better at learning a foreign language than are monolingual adults. Previous research suggests that bilinguals have an advantage over monolinguals at some tasks, including creative thinking and problem solving. Recently, researchers have also claimed that bilinguals are better than monolinguals at learning another language, but this hypothesis has not yet been tested systematically.

This project will examine this hypothesized advantage by having bilingual adults learn a language that they don't already know--one that is, in fact, an artificial language, which by design they could not already have been exposed to--and comparing their performance to that of monolingual adults under the same learning conditions. The proficiency that participants attain in the language, as well as measures of their brain activity while they process the language, will be tested at two points, once early on during learning (at low proficiency) and once at the end of training (at high proficiency). The project will also evaluate the effect of the context in which the artificial language is acquired by comparing learning in a classroom to learning in an immersion context. Half of the bilingual participants will train under classroom-like conditions; whereas the second half of this group will receive training under immersion-like conditions. The corresponding data for the monolingual group, which has already been collected, indicates that by the time monolingual learners reach high proficiency, their brain patterns look similar to those of native speakers of most languages, particularly for those who underwent immersion-like training. This suggests that monolingual foreign language learners can indeed become like native speakers in how their brains learn and process the language. The data from bilinguals will be collected and compared to the data from monolinguals. The hypothesized bilingual advantage in acquiring new languages predicts that the bilingual participants will be faster and more accurate than the monolinguals in learning the artificial language. It also predicts that their brain patterns will be even more similar to patterns that are typically found when native speakers process sentences in their own first language.

Because learning foreign languages makes an important contribution to the lives of many people the world over in today's global economy, and because proficiency in foreign languages is also critical to the security needs of the government, findings from this project may have broader consequences for education, policy and security.

Learning a second language (L2) is becoming increasingly important given the globalization of modern life. Yet, L2 learning is notoriously difficult, especially in adulthood, and even after high proficiency is reached, learners face the difficult yet critical task of retaining what they have learned. We currently know very little about the neural mechanisms underlying L2 acquisition and retention. A major roadblock in investigating the neurocognitive trajectory of L2 acquisition is the time it takes a learner to acquire a full language, from initial exposure all the way to high proficiency. With support from the National Science Foundation, Dr. Michael Ullman and colleagues will use a new approach to reveal the neural processing involved in the entire trajectory from initial acquisition to high proficiency and then subsequent retention. Using a reduced version of an actual natural language, Basque, but with many fewer words and a simplified grammar, learners can achieve high proficiency in a relatively shorter time. Investigating L2 learning using this reduced language may offer insights into shared neural mechanisms underlying the learning and retention of a full language. Both behavioral and functional Magnetic Resonance Imaging (fMRI) measures will be continuously recorded while English-speaking adults learn reduced Basque to high proficiency, over several learning sessions spread out over several days, as well as again one and six months later to examine language retention. This study will provide the first fine-grained picture of the full behavioral and neural trajectories of L2 learning from initial exposure to high proficiency and then retention.

Understanding how the brain learns a second language to high proficiency, and then retains it, will pave the way to multiple beneficial outcomes. First, it may lead to effective evidence-based language learning and teaching approaches. For example, different neurocognitive mechanisms could be targeted for different aspects of language at different points in the learning trajectory. Such approaches could have significant impacts for education, security, and the economy, where learning an L2 is often important, and sometimes critical. Such approaches could also lead to advances in language (re)learning therapies for individuals with developmental or acquired disorders (e.g., Specific Language Impairment, dyslexia, autism, acquired aphasias), with substantial clinical benefits. The proposed research may additionally impact research methodology, since demonstrating the utility of this paradigm would open up major new avenues of L2 research. The project should also have institutional and educational benefits. It brings together researchers from several disciplines and institutions, strengthening existing collaborations and forming new ones. Finally, the project will be leveraged to train the next generation of scientists, through the development of an undergraduate research module, and the training of high school, graduate, and post-doctoral students, providing them with first-hand experience in cutting edge interdisciplinary research.