Through a series of psychological and neuroimaging studies Dr. Maria M. Pinango will investigate how and where the brain decodes the information of a sentence in real-time, that is, as the sentence unfolds. Two experimental lines will be carried out: a psychological line and a neuroimaging line. Through the psychological line she will determine how the comprehension system reveals the processing of the different kinds of information contained in a sentence: syntactic, semantic and conceptual. Pinango's research model predicts that information about sentence structure (syntactic) is done immediately and with no cost, whereas information about the meaning information (semantic) takes time to develop and demands more computation. Finally, Pinango's model predicts that the process of placing the content of a sentence (conceptual information) into the larger context of what speakers know about the world is yet a costlier process as it requires the comprehension system to leave the language system and place the meaning of the sentence into the larger and richer conceptual system. Through the neuroimaging (functional magnetic resonance imaging) line it will be determined how and where in the brain this parallel decoding of syntactic, semantic and conceptual information takes place.
This research connects to important questions in linguistics, artificial intelligence, cognitive neurology and neuroscience. On the medical side, the research has implications for the study of brain pathologies as it seeks to answer the question of why on the one hand, special populations such as Alzheimer's patients, Schizophrenia patients have what appear to be intact linguistic ability yet are unable to fully comprehend sentences and, why on the other, cerebro-vascular patients such as Broca's aphasic patients have true linguistic impairments yet are able to show normal-like sentence comprehension. The project provides a unique opportunity for undergraduate students to receive intensive training in the demanding task of investigating a model of linguistic organization from the perspective of brain organization. In doing so it supports the crucial collaboration between researchers in linguistics and neuroscience at Yale University.
Objectives. This proposal had three related objectives 1) to understand the cortical distribution of distinct compositional mechanisms (syntactic vs. semantic vs. pragmatic/contextual), and 2) to glean from that understanding the structure of a neurologically viable architecture of language, particularly as it concerned its interface with the larger cognitive system. Finally, the proposal had a methodological objective, 3) to understand the potential correlations of neuroimaging and real-time (behavioral) indicators and in this way associate localizational with time-course measures. Linguistic phenomena. To address composition at these three levels we focused on six phenomena: one targeting syntactic composition (long-distance wh-dependencies), three targeting semantic composition (light verb construction, aspectual coercion and complement coercion) and two targeting pragmatic composition (lexical metonymy and reference transfer). These phenomena were chosen because they had each independently been shown empirically to be very good indicators of those respective kinds of composition. Methodological approach. In addressing the first two objectives, our starting point was the understanding (at the time) of traditional language areas Broca’s and Wernicke’s areas in connection to syntactic and semantic composition respectively, and their connection with another cognitively relevant region, the prefrontal cortex. The hypothesis of a PFC pragmatics connection was supported on the neuroanatomical connections between this area and Broca’s and Wernicke’s areas, and the long-standing understanding of the PFCs cognitive commitments in terms of executive functioning and decision making processes. To address the third objective we set out to measure the hypothesized mechanisms through a variety of tasks, some of them targeting time-course of composition such as self-paced reading, eye-tracking, and event-related brain potentials and others targeting cortical localization: fMRI and focal-lesions studies. Initial predictions. As mentioned, our initial language-brain model was built on traditional well-supported observations regarding cortical distribution of linguistic composition. It proposed the following correlations: Syntactic composition associated with the workings of the left inferior frontal cortex (LIFG), semantic composition associated with the workings of the left posterior superior temporal cortex (LpSTG), and pragmatic composition associated with the workings of the prefrontal cortex. If this hypothesized model were true it would come to support directly the natural divisions that linguistics had proposed for the language system. These fundamental distinctions would be manifested in the following (predicted) observations: syntactic composition in the form of long-distance wh-dependencies would be preferentially associated with the workings of the LIFG, semantic composition in the form of light verb construction, aspectual coercion and complement coercion would be shown to involve greater computational cost (through SPR, eye-tracking and/or ERP), and be preferentially associated with the LpSTG, and finally pragmatic composition in the form of lexical metonymy and reference transfer. Altogether, our results support the following conclusions: 1) The brain does not seem to distinguish between compositional subsystems in the way that linguistics as a field traditional assumes. Instead, the cortical patterns indicate a lexicon-composition division such that linguistic composition proper appears to occur simultaneously, in parallel depending especially on the recruitment of the workings of the LIFG. 2) The mutually exclusive Broca’s-syntax and Wernicke’s-semantics correlation traditionally assumed in neurocognitive models is not supported. Instead, results are consistent with neurocognitive maps whereby Broca’s area/LIFG is recruited for linguistic composition irrespective of kind, and Wernicke’s area is preferntially sensitive to (a subset of) lexico-semantic compositional distinctions. 3) Semantic composition beyond syntactic composition shows consistent cost irrespective of type of meaning composition. The clearest cost is observed whenever lexical meanings are underspecified, and specification is required for minimal initial interpretation. In these cases, the cost is attributed to the processing system to, as it were, mine the context (explicit or default) in order to satisffy an underspecified lexico-semantic requirement. It is this process which causes computational cost and appears to depend mostly on the workings of the LIFG (beyond traditional Broca’s area). 4) The traditional semantics-pragmatics distinction is not categorical, but rather of degree (context-modulated).This is seen in the unexpected overlapping in cortical localization and time-course for lexical metonymy and reference transfer. To our knowledge this is the first systematic attempt at studying kinds of linguistic composition (syntactic, semantic and pragmatic) simultaneously from a real-time comprehension and neurological perspectives. This approach has allowed us to shed light on the independent nature of each kind of composition, and the ways in which they connect to each other and with the rest of cognition. Accordingly, this research has direct impact on our understanding of language comprehension deficits thought to be rooted on grammatical impairments. Our research suggests that this implication could be premature. Instead our results suggest that the problem may be the way in which meaning composition is distorted given the locus of the brain damage. This new perspective has direct implications for how we understand linguistic deficits from focal-brain lesions, from degenerative disease such as Alzheimers, and from mental illness such as squizophrenia and dementia, all impairments associated with Frontal Lobe damage.
