ارتباطات عصبی واژگان و دستور زبان: شواهدی از تولید، خواندن و قضاوت صرف فعل در زبان پریشی

Are the linguistic forms that are memorized in the mental lexicon and those that are specified by the rules of grammar subserved by distinct neurocognitive systems or by a single computational system with relatively broad anatomic distribution? On a dual-system view, the productive –ed-suffixation of English regular past tense forms (e.g., look–looked) depends upon the mental grammar, whereas irregular forms (e.g., dig–dug) are retrieved from lexical memory. On a single-mechanism view, the computation of both past tense types depends on associative memory. Neurological double dissociations between regulars and irregulars strengthen the dual-system view. The computation of real and novel, regular and irregular past tense forms was investigated in 20 aphasic subjects. Aphasics with non-fluent agrammatic speech and left frontal lesions were consistently more impaired at the production, reading, and judgment of regular than irregular past tenses. Aphasics with fluent speech and word-finding difficulties, and with left temporal/temporo-parietal lesions, showed the opposite pattern. These patterns held even when measures of frequency, phonological complexity, articulatory difficulty, and other factors were held constant. The data support the view that the memorized words of the mental lexicon are subserved by a brain system involving left temporal/temporo-parietal structures, whereas aspects of the mental grammar, in particular the computation of regular morphological forms, are subserved by a distinct system involving left frontal structures.

In the study of language, a fundamental distinction is drawn between the “mental lexicon” and the “mental grammar.” The lexicon contains memorized pairings of sound and meaning. It must contain at least those words whose phonological forms and meanings cannot be derived from each other, such as the non-compositional word cat. It may also contain other non-compositional forms, smaller or larger than words: bound morphemes (e.g., the -ed past tense suffix, and the root nomin in nominal and nominate) and idiomatic phrases (e.g., kick the bucket). The grammar encompasses rules or constraints that govern the sequential and hierarchical combination of lexical forms into predictably structured complex words, phrases, and sentences. That is, the grammar subserves the computation of compositional linguistic forms whose meaning is transparently derivable from their structure. For example, a mental rule which specifies that English past tense forms are derived from the concatenation of a verb stem and an -ed suffix would allow us to compute past tenses from new words (e.g., fax + -ed → faxed) and from novel forms (e.g., blick + -ed → blicked). Rule-derived forms can thus be computed in real-time, and so do not need to be memorized—although even compositional linguistic forms (e.g., walked) could in principle be memorized in the lexicon ( Berko, 1958, Chomsky, 1965, Chomsky, 1995 and Pinker, 1994). These two language capacities have been explained by two competing theoretical frameworks. “Dual-system” theories posit distinct cognitive or neural components for the two capacities (Chomsky, 1965, Chomsky, 1995, Damasio and Damasio, 1992, Fodor, 1983 and Pinker, 1994). On this view, the learning, representation, and/or processing of words in a rote or an associative memory is subserved by one or more components, which may be specialized and dedicated (“domain-specific”) to these functions (Bloom, 1994, Chomsky, 1965, Chomsky, 1995, Fodor, 1983, Forster, 1979, Levelt, 1989, Levelt, 1992, Markman and Hutchinson, 1984, Pinker, 1994, Seidenberg, 1985, Swinney, 1982 and Waxman and Markow, 1996). The use of stored words has been posited to depend especially on left posterior regions, particularly temporal and temporo-parietal structures (Damasio, 1992, Damasio et al., 1996, Dejerine, 1901, Geschwind, 1965, Goodglass, 1993, Lichtheim, 1885, Luria, 1966 and Wernicke, 1874). The learning, knowledge, and/or processing of grammar are subserved by one or more components that are specialized and dedicated to their linguistic functions, and that have been posited to be innately specified (Chomsky, 1965, Chomsky, 1995, Fodor, 1983, Frazier, 1987 and Pinker, 1994). The use of grammar has been claimed to be dependent on left frontal cortex, particularly Broca’s area (the inferior left frontal gyrus, which contains the cytoarchitectonic Brodmann’s areas 44 and 45 (Damasio, 1992)) and adjacent anterior regions (Bradley et al., 1980, Caramazza et al., 1981, Damasio, 1992, Grodzinsky, 2000 and Zurif, 1995), although this has been controversial, in particular regarding the comprehension of syntax (e.g., Hickok, 2000). In contrast, “single-mechanism” (single-system) theories posit that the learning and use of the words and rules of language depend upon a single computational system that has a relatively broad anatomic distribution (Bates and MacWhinney, 1989, Elman, 1996, MacDonald et al., 1994, MacWhinney and Bates, 1989, Rumelhart and McClelland, 1986 and Seidenberg, 1997). This system is general-purpose (“domain-general”) in that it also subserves non-language functions. There is no categorical distinction between non-compositional and compositional forms on this view. Rather, rules are only descriptive entities, and the language mechanism gradually learns the entire statistical structure of language, from the arbitrary mappings in non-compositional forms to the rule-like mappings of compositional forms. Modern connectionism has offered a computational framework for the single system view. It has been argued that the learning, representation, and processing of grammatical rules as well as lexical items takes place over a large number of inter-connected simple processing units. Learning occurs by adjusting weights on connections on the basis of statistical contingencies in the environment (Elman, 1996, Rumelhart and McClelland, 1986 and Seidenberg, 1997). Single and double dissociations which differentially link the lexicon to left posterior regions and aspects of grammar to left anterior regions suggest that these regions contain distinct neural underpinnings which play different roles in the knowledge or processing of the two capacities, as predicted by a dual system view. Such dissociations have been revealed by several experimental approaches. Aphasia. There are at least two fundamental types of aphasia. These constitute an empirically demonstrated categorical distinction with respect to several behavioral and neuroanatomical dimensions. The dichotomy has variously been described as receptive/expressive, fluent/non-fluent, and posterior/anterior. Each label focuses on a different dimension of the aphasic impairment, such as whether it primarily affects input or output, how it affects speech production, and whether its associated lesions are in anterior or posterior portions of the left hemisphere ( Alexander, 1997, Caplan, 1987, Caplan, 1992, Dronkers et al., 2000, Goodglass, 1993 and Goodglass et al., 1964). Fluent aphasia involves speech that is facile in articulation and relatively normal in phrase length. It is associated with “anomia”—impairments in the production and reading of “content” words, such as nouns and verbs—and with deficits in the recognition of content word sounds and meanings. Fluent aphasics’ lexical difficulties can be contrasted with their tendency to omit neither morphological affixes (e.g., the past tense -ed suffix) or “function” words, such as articles and auxiliaries, in their speech and reading. They also generally produce sentences whose syntactic structures are relatively intact. Fluent aphasia is strongly associated with damage to left temporal and temporo-parietal regions. Non-fluent aphasia involves speech that is effortful, with a reduction of phrase length and grammatical complexity. This “agrammatic speech” in non-fluent aphasia is strongly associated with impairments at producing appropriate morphological affixes (e.g., -ed) and function words. Non-fluent aphasics also often have difficulties using syntactic structure to understand sentences, and may have deficits at judging the grammaticality of sentences involving particular types of structures. In contrast, non-fluent aphasics are relatively spared in their use of content words, particularly in receptive language. Non-fluent aphasia is associated with damage to left frontal structures ( Caplan, 1992, Caramazza et al., 1981, Dronkers et al., 2000, Goodglass, 1993, Goodglass and Wingfield, 1997, Grodzinsky, 2000 and Grodzinsky and Finkel, 1998). Electrophysiology. Event-related potential (ERP) studies seem to be consistent with the dissociations noted in aphasia. The “N400” is a central/posterior negative component which is associated with manipulations of word sounds and meanings ( Hagoort and Kutas, 1995, Kutas and Hillyard, 1980 and Kutas and Hillyard, 1983), and has been linked to left temporal lobe structures ( Nobre et al., 1994, Papanicolaou et al., 1998 and Simos et al., 1997). In contrast, disruptions of syntactic processing can yield early (150–500 ms) left anterior negativities ( Friederici et al., 1993, Hagoort et al., 2003 and Neville et al., 1991)—i.e., “LANs.” These have been linked to rule-based automatic computations ( Friederici et al., 1996 and Hahne and Friederici, 1999) and left frontal structures ( Friederici et al., 1998 and Friederici et al., 1999). LANs have been elicited cross-linguistically by violations of syntactic word-order ( Friederici, 2002, Friederici et al., 1993, Neville et al., 1991 and Newman et al., 1999) and morpho-syntax ( Coulson et al., 1998, Friederici and Frisch, 2000, Kaan, 2002, Kutas and Hillyard, 1983, Münte et al., 1993, Osterhout and Mobley, 1995 and Rosler et al., 1993). However, not all studies examining these types of violations have reported LANs ( Ainsworth-Darnell et al., 1998, Allen et al., 2003, Hagoort and Brown, 1999, McKinnon and Osterhout, 1996, Osterhout, Bersick et al., 1997 and Osterhout and Mobley, 1995). It is not clear at this point why LANs have been found by some studies and not by others, even for the same types of violation ( Osterhout & Mobley, 1995). Neuroimaging. Positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) have also revealed dissociations between lexicon and grammar. Posterior activation in left temporal and/or temporo-parietal regions has been associated with a variety of lexical and semantic tasks (for a summary, see e.g., Ullman, 2004), such as semantic categorical judgments of auditorily presented word pairs ( Wise, Chollet, Hadar, Friston, & Hoffner, 1991), naming colors, faces, animals, and tools ( Damasio et al., 1996, Martin et al., 1995 and Martin et al., 1996), and same/different judgments on sentence pairs containing identical syntax, but differing in one word ( Bookheimer, Zeffiro, Gaillaird, & Theodore, 1993). In contrast, preferential activation in portions of Broca’s area has been elicited by a variety of tasks designed to probe syntactic processing ( Bookheimer et al., 1993, Caplan et al., 1998, Dapretto and Bookheimer, 1999, Embick et al., 2000, Friederici, Ruschemeyer, et al., 2003, Indefrey et al., 1999, Indefrey et al., 2001, Kang et al., 1999, Knoesche et al., 2000, Moro et al., 2001, Ni et al., 2000 and Stromswold et al., 1996). For example, Broca’s area activation has been found when subjects gave acceptability judgments to syntactically more complex sentences, as compared to syntactically less complex sentences ( Caplan et al., 1998 and Stromswold et al., 1996), or same/different judgments to sentences differing in word order, but having the same meaning and containing identical words ( Bookheimer et al., 1993). However, the picture is by no means crystal clear. Thus there is also evidence suggesting that posterior regions may play a role in certain grammatical abilities, and that frontal areas play a role in certain lexical abilities. Aphasia. Fluent aphasics can have “paragrammatic” speech, characterized by the incorrect use of morphological affixes, particularly the substitution of one affix for another. Fluent aphasics have also been shown to have trouble using syntactic structure to understand sentences in standard off-line measures, and can be impaired in judging their grammaticality (although on-line measures designed to capture real-time language processing suggest that fluent aphasics have normal syntactic reflexes). Non-fluent aphasics usually have trouble retrieving content words in free speech (although they are relatively spared at recognizing such words). Moreover, they may retain the ability to make grammaticality judgments about certain syntactically complex sentences ( Alexander, 1997, Caplan, 1987, Caplan, 1992, Dronkers et al., 2000, Goodglass, 1993, Grodzinsky and Finkel, 1998, Linebarger et al., 1983, Love et al., 1998 and Swinney et al., 1996). Electrophysiology. A posterior positive ERP component, usually maximal over parietal areas and bilaterally symmetric (the “P600”), is associated with syntactic processing difficulties ( Hagoort and Kutas, 1995, Kaan et al., 2000 and Osterhout, McLaughlin et al., 1997), in particular with aspects of controlled processing ( Friederici et al., 1996, Friederici et al., 2001 and Hahne and Friederici, 1999). P600s do not appear to depend on frontal brain structures, and may involve both basal ganglia and posterior regions ( Friederici and Kotz, 2003, Friederici et al., 1998, Friederici et al., 1999, Friederici, Kotz, et al., 2003 and Ullman, 2001b). Neuroimaging. Syntactic processing has been linked to anterior superior temporal cortex ( Dapretto and Bookheimer, 1999, Friederici, Ruschemeyer, et al., 2003, Meyer et al., 2000, Newman et al., 2001 and Ni et al., 2000). Additionally, increasing the syntactic complexity of visually presented sentences has been found to yield increased bilateral frontal and temporal activation ( Just, Carpenter, Keller, Eddy, & Thulborn, 1996). Finally, activation in certain frontal regions is associated with the search, selection, or retrieval of word sounds and meanings ( Buckner and Peterson, 1996, Buckner and Wheeler, 2001 and Thompson-Schill et al., 1997). The lack of clear and consistent neuroanatomical dissociations between lexicon and grammar has kept the dual-system/single-system controversy very much alive. Testing for lexicon/grammar dissociations has been problematic because tasks probing for lexicon and for grammar usually differ in ways other than their use of the two capacities. For example, it is difficult to match measures of grammatical processing in sentence comprehension with measures of lexical memory (see Bates, Harris, Marchman, Wulfeck, & Kritchevsky, 1995). A productive approach to investigate the brain bases of lexicon and grammar may thus be to examine language phenomena in which factors other than lexical and grammatical involvement can be controlled for.

Eighteen aphasics with non-fluent agrammatic speech or with agrammatic reading were presented or reviewed in this report. Sixteen of the 18 showed a pattern of worse performance at computing regular than irregular past tense or plural forms, in production, reading, judgment, writing, or repetition tasks. The other two aphasics showed no difference in their use of regular and irregular forms. Nine aphasics with fluent speech and anomia were presented in this report. Eight of the 9 showed a pattern of worse performance at irregular than regular past tense forms, in production, reading, and judgment tasks. The remaining fluent aphasic showed no difference between regular and irregular forms. These double dissociations were maintained even when measures of a variety of other factors, including frequency, phonological complexity, and articulatory difficulty, were controlled for. The agrammatic non-fluent aphasics also had particular trouble computing over-regularizations and novel -ed-suffixed verbs. The anomic fluent aphasics had little trouble with over-regularizations and novel -ed-suffixed forms, but were impaired at novel irregularizations (e.g., crive–crove). These findings are not consistent with any previously reported connectionist models of regular and irregular morphology, including models with distinct representations for semantics and phonology (Joanisse & Seidenberg, 1999). Aspects of the data also seem to pose a challenge for Distributed Morphology (Halle & Marantz, 1993). The results support a dual-system model in which the computation of affixed and irregularized inflected forms depend upon distinct neural underpinnings. The association of non-fluent aphasia, left anterior lesions, agrammatism, apparent syntactic deficits in all three inflection tasks, and impairments of morphological affixation, suggests that morphological affixation and at least some syntactic processes are subserved by left anterior structures. An examination of the tested aphasics’ lesioned structures suggests that left frontal regions, particularly Broca’s area and adjacent frontal structures, play a particular important role in these grammatical functions. The association of fluent aphasia, left posterior lesions, lexical difficulties, and impairments of real and novel irregular morphology, including a large number of distortions on irregular verbs, suggests that left posterior brain regions subserve a lexical memory that includes the sound patterns of stored forms, encompasses irregularly inflected as well as uninflected words, and subserves the use of novel irregularizations. An examination of the tested aphasics’ lesioned structures implicates left temporal and/or temporo-parietal structures in these functions. The results obtained in the present experiments, and in the older studies discussed in some detail, are largely consistent with data reported elsewhere. As summarized in Section 1, a number of recent studies have also found regular deficits in non-fluent aphasics, in production, reading, judgment, and priming tasks (Marslen-Wilson and Tyler, 1997 and Tyler, de Mornay-Davies, et al., 2002), with the opposite pattern reported for production tasks given to fluent aphasics (Miozzo, 2003 and Ullman, Corkin, et al., 1997). To our knowledge, only one study of aphasia has argued that their data do not support the existence of reliable regular/irregular dissociations in English inflectional morphology (Bird et al., 2003). This investigation examined 10 non-fluent agrammatic aphasics on production, repetition, reading, and judgment tasks. When phonological complexity, frequency, and other factors were accounted for, the regular deficit was weakened considerably. However, controlling for these other factors did not eliminate the effect completely. First, the regular impairment in the reading task survived in all analyses. Moreover, in the repetition task, only a post hoc analysis, in which irregular past tense forms that had inconsistent voicing (e.g., felt) were excluded, eliminated the regular disadvantage. Thus even in this study, whose primary theoretical motivation was to demonstrate that the regular deficit does not survive when phonological and other factors are controlled for, it was difficult to impossible to eliminate the effect. In conclusion, we have presented a detailed analysis and discussion of the computation of English regular and irregular inflected forms in agrammatic non-fluent aphasia and anomic fluent aphasia. The findings from these studies support the view that language is a modular system—that is, language is subserved by separable neurocognitive components: at least certain aspects of the mental grammar, including certain syntactic computations as well as morphological affixation, are subserved by left frontal structures, whereas the stored words of lexical memory, including irregularly inflected forms, depend on left temporal/temporo-parietal regions. The results are consistent with the “declarative/procedural” hypothesis that aspects of the mental grammar are subserved by a frontal/basal-ganglia procedural memory system that also underlies cognitive and motor skills, whereas the mental lexicon is subserved by a temporal/temporo-parietal declarative memory system that also underlies factual knowledge about the world (Ullman, 2001c, Ullman, 2004 and Ullman, Corkin, et al., 1997).