L.K. Obler & K. Gjerlow. 1999. “Language Organization”, in Obler, L.K. & K. Gjerlow. Language and the Brain. CUP.
Language organization
Up to this point in the book, we have been discussing what we know about the individual populations from whom we have gleaned knowledge about neurolinguistics. At the end of each chapter we have provided a conclusion about the brain areas that seem to be impaired for the group studied in that chapter and how the brain damage is related to the language phenomena characteristically seen. In this chapter we pull together information provided in those chapters to focus on the linguistic constructs that have a "psychological reality" apart from their abstract value.
Since Noam Chomsky's 1957 book Syntactic Structures, a major goal for modern linguistics has been to construct a model for the grammar of an "ideal" native speaker of a language. The primary goal is to create an abstract model that "generates" – that is, accounts for – all of the grammatical sentences in the language without generating any ungrammatical strings. These models are not intended to reflect processes that native speakers use to actually create sentences each time they speak, of course. Some linguists may see the generative grammar as an essential part of a production model; that is, they see the the grammar as somehow "driving" actual speech production or comprehension. Many others see the generative grammar as potentially divorced from the speech process. For them, the grammar describes the abstract patterns of the language, while another set of rules that one might call a production grammar tells speakers how to actually speak.
We prefer to assume that our abstract grammar does have some connection to the physical manifestations of speech and language processes. As a result, we would expect that autonomous aspects of analysis could be linked to dissociable kinds of impairment in brain-damaged patients. If performance on one set of linguistic items (e.g. functor words) is markedly impaired but that on related items (e.g. nouns and verbs) is quite spared, this dissociation in performance suggests that the two sets of items are at least psychologically distinct. That is, independent psychological mechanisms must exist for processing them. What is meant by "psychological reality" of an abstract grammatical concept, then, is that it is not only an abstract linguistic concept; it can be seen to operate during speech production and/or comprehension.
This notion of "psychological reality" is an important one for determining the relation between abstract theoretical notions that linguists have deduced and how the brain actually functions. It is conceivable that these abstract notions in fact have their reflections in brain representations and processes. It is, we grant, also possible that the abstract notions describe a pattern that exists in language as an abstract entity, but has no meaningful relation to how language is actually processed. Consider how little relation there is between the mathematical equation that describes the spiral-shape pattern of the seeds at the center of a sunflower, and whatever biological rules govern the seeds patterning into that shape.
Arithmetic is a system in some ways similarly abstract to language. While we may expect some brain representation for the math-facts that we have automatic access to (e.g. 5 + 4 = 9), we may question whether such abstract principles as commutativity (the rule that tells us that if we add A plus B first and then add C to it, we get the same result as if we add B plus C first and then add A to the sum) are represented as a rule in a distinct location, or even a distinct neural pathway in the brain. Certainly very few of us can articulate that rule (or remember its name), although most adults operate on the understanding that it is true without being able to specify it. We may conclude, based on our real-world use of the principle, that commutativity has psychological reality, but we would not expect it to have a brain location associated with it.
How then does the notion of "psychological reality" speak to the relation between competence and performance? While most linguists are interested in studying competence, as we mentioned above, that is, the elegant underlying grammar of language as we adults know it, we may well ask how our flawed performance relates to this competence. Is there a simple one-to-one relation between the rules of the grammar and the rules we use for speaking or understanding (parsing) what we hear? Or is there an entirely different grammar built for producing language and understanding it, one that may take into account such factors as word-frequency or the likelihood that certain syntactic structures will be employed? Indeed it is conceivable that computer-programming types of principles will predominate in giving us performance grammars that work for production and comprehension and that are substantially distinct from the elegant structures of a competence grammar.
In earlier chapters, we have considered how linguistic notions have proved useful in studying patterns of behavior in braindamaged individuals. Here we ask how neurolinguistic data buttress aspects of linguistic theory. Most theoretical linguistic discussion arises from linguists' thinking about what appears to be going on and testing out their ideas against native-speaker intuitions of which sentences are grammatical and which are not. Neurolinguistic data, like data from other "exceptional" language performance such as pidgin language or literary language or normal speech errors (Obler and Menn, 1982) permit us a complementary measure to evaluate the linguistic constructs.
In this chapter we will use the basic linguistic Levels (phonology, morphology, syntax, semantics) to structure the discussion. After discussing the linguistic notions of competence and performance, we address several phonological concepts (the phoneme, the distinctive feature, morphophonological constraints on word structure, syllable, and the suprasegmentals tone and intonation). We then turn to morphological notions such as the word-group classes, morphophonological concepts such as compounding and affixation and the relation between morphology and syntax. Under syntactic notions we treat the notions of hierarchy and complexity, deep and surface syntax, traces, and thematic role. With respect to theories of how the lexicon is organized, we treat what a lexicon is itself, the notion of word, the distinction between lexical form and semantic meaning, semantic hierarchies, and semantic networks. Under pragmatics we treat conversational rules. Grice' maxims for interpreting conversation. the lexicon of emotion, and bilingual language choice. Vis-à-vis written language systems, we cover the interface between phonology and orthography, the differences between different orthographic types, even for the same language (as in the syllable and whole-word representations in Japanese), the two-route model of normal reading and spelling regularity. Of course these are not all the theoretical concepts that could be extracted from the work in this book; rather, they are a representative sample of concepts that linguists have found useful for which we have neurolinguistic evidence.
Competence and performance
This linguistic distinction between our knowledge of language and our actual use of it has been stressed by linguists to emphasize the importance of focusing on the abstract grammar. It was not intended to enable us to understand even normal speech errors, not to mention the production of brain-damaged subjects. However, we may say that the distinction is made much richer by the literature on language breakdown. It is most strikingly evident in patients whose output is agrammatic, yet who are able to perform quite well on grammaticality judgment tasks. We must assume they have the grammatical knowledge even if they do not appear to be able to use it in their speech. Similarly, we must assume that competence, as linguists think of it, is spared for aphasic patients who have either production or comprehension skills impaired, but not both. Likewise, patients whose only problems are in reading, or in understanding auditory language input, but not both, provide evidence that competence is spared because they have one modality that is functioning.
Phonology
Because units of linguistic analysis are not necessarily physically real, we must evaluate their "psychological reality" by seeing if they make a difference in our speech or comprehension of language.
For example, people do not knowingly speak in phonemes. Speech scientists tell us that a given "phoneme" may be pronounced markedly differently depending on the context it is in. The characteristic sounds composing a /b/ before the vowel /i/ are different from those before a /u/. However, linguists have long understood that both are perceived by the speaker as well as the listener as the "same" sound, and, at least for English speakers, this sound is to be distinguished from a similar one whose voicing starts later, the sound that is produced in similar fashion but with a later onset of vocal cord vibration, namely /p/. Indeed, these two "sounds" equally distinguish the words "cap" and "cab"' and the words "pour" and "boar." It is this possibility of recognizing distinctive phonemes that contributes to what we mean by "psychological reality." The distinction is meaningful for the speaker and the listener, and it must be represented in their brains for language processing, even if they are not able to explain the phenomenon without advanced linguistic training.
Each particular instance of a sound in a speech stream has its own collection of features. In the chapter on bilingualism, we discussed Voice Onset Time or VOT, one of the properties of "stops" (i.e. sounds made by using the articulators to "stop" the flow of air from the lungs). Recall that voice onset time refers to the time when the vocal cords begin to vibrate relative to the time the articulators close. The English sounds "p" and "b" for example, are distinguished only by VOT. However, there is no one single VOT associated with each of the two sounds. Instead, native speakers recognize a range of VOTs starting from around sixty milliseconds after the release of the stop as "p" and another range ending around five milliseconds before release of the stop as "b." Distinctions that can be readily discerned by machines, by prelingual infants and by speakers of languages that distinguish three phonemes in this range (e.g. Thai), are routinely ignored by native speakers of English. This "categorical perception" or sounds as phonemes is one of the main arguments from unimpaired speakers in favor of the psychological reality of the phoneme. Further evidence from normals for the meaningfulness of phonemes is that cueing a forgotten word by giving the speaker its first phoneme is often quite successful in eliciting the word.
In the literature we have covered in this book, the psychological reality of phonemes is evidenced most vividly in the phonemic paraphasias of aphasics. Regularly, in this sort of substitution error, it is only one phoneme that has been substituted for or omitted (e.g. spill for spell). Recall that the equivalent can also be seen in the visual-gestural system, as elements equivalent to phonemes are seen to be substituted for in certain aphasics who are speakers of signed languages.
The fact that some bilinguals often evidence a "foreign accent" in one of their languages, also indicates that specific phonemes have a strong, psychologically real, phonetic identity. The phonetic constraints on the phoneme are so strong for the individual that related ones in a foreign language are "heard" and "produced" as the native-language phoneme when the non-native speaker speaks the new language. For example, Arabic has the phoneme /b/ but no phoneme /p/. When speakers of Arabic are not fully proficient in English, they regularly pronounce the English /p/ as a /b/. Similarly, Arabic, unlike English, has several sounds in the "h" range (/h/, /h,/, and /x/). In spontaneous speech and conversation, native speakers of English who are learning Arabic often produce and hear all three as /h/. Aphasics too may suddenly appear to have what sounds like a foreign accent as the result of brain damage. This illustrates that the phonetic distortions associated with aphasia can be quite systematic in such a way that certain phonemes (or suprasegmentals) are particularly impaired, causing the "accent."
The foreign accent data also apply to notions of morphophonology, as words that are morphophonologically not permitted in the first language are particularly difficult to produce in the second language and are often betrayed by foreign accent at the impermissible points. For example, because words cannot start with a sibilant-stop-resonant cluster in Spanish, Spanish speakers will regularly add a vowel at the beginning of English words (e.g. "spring" is produced /ɛsprlŋ/).
Recall that when phonemic substitutions are made by aphasics, the substitutions are markedly more likely to differ from the phoneme that they substitute for by only one feature (such as voicing or frication) and markedly less often by two or three. Thus aphasic pronunciation of the word "cat" is more likely to be /kæd/ or /kIt/ than it is to be /kls/. The notion of distinctive features – the linguistic constructs such as manner or place of articulation that distinguish among similar phonemes – is supported by such a pattern of breakdown.
The syllable is a familiar unit of theoretical analysis. The psychological strength of the syllable is manifested when both normal speakers and aphasics can demonstrate they remember the number of syllables in a word they cannot think of in the tip-of-thetongue phenomenon. The further breakdown of syllables into less familiar units or "onset" and "rhyme" is plausible in light of another tip-of-the-tongue phenomenon, whereby aphasics, like normals, can sometimes remember the initial phoneme of a word they cannot remember in its entirety.
Suprasegmental phenomena are also attested to in the data from the populations we have considered. Recall that linguistic tone in speakers of tone languages appears to be easily discriminated from pitch patterns that are not linguistic. The evidence comes not only from laterality studies in normals, but also from observation of tone breakdown in patients with Broca's aphasia. Patients who have difficulty with intonation for syntax but not with linear pitch patterns provide evidence for the psychological reality of intonation patterns that serve syntactic functions.
The rule-governed relationships between the phonological forms produced by children with Specific Language Impairment (SLI) and those of normally developing children likewise testify to the notion of word-form principles in language acquisition. While age-matched children have already acquired a certain type of consonant cluster in a certain position, a child with SLI may systematically produce a reduced form of that consonant cluster (e.g. word-final /sp/ as in the word "clasp”). Both the ability of the normal children to produce the consonant cluster and the contrasting delay in the child with a specific language impairment (for whom the word "clasp" might be produced like the word “class") testify to the brain-based "psychological reality" of the constraints on lexical shape.
The fact that fluent aphasics and demented patients produce neologisms that invariably conform to the morphophonological constraints of the speaker's language is evidence for such morphophonological rules. A related set of evidence for real-word morphophonology derives from the fact that the demented patient HCEM (see chapter 8) would spontaneously correct deliberately inserted phonological errors in repeating back the words and sentences of her examiner.
Morphology
A certain psychological reality for the notion of word is seen in the fact that many agrammatic aphasics, when they delete affixes, only do so in such a way that real words in their language will be the outcome.
Word classes are seen as meaningful in production in several different ways. The distinction between functors and substantives is clearly manifested in the breakdown of the former but not the latter among agrammatic and paragrammatic patients. That verbs and nouns have special, independent status is seen in the relative dissociability of nouns and verbs among numbers of aphasics. In the aphasia chapters we discussed bow some aphasics have nouns preferentially spared while others have nouns particularly impaired and verbs relatively spared.
The distinction between affixes and word stems is seen in their differential breakdown in agrammatism. Within affixes, distinctions between inflectional and derivational classes can be seen as well, not only in agrammatism but also in alexia (particularly deep dyslexia) and in dysgraphia (particularly in the errors of demented patients). As all these subjects produce word stems better than their affixes, we might assume that stems are simply better represented than affixes in the lexicon. Recall that fluent aphasics, by contrast, reproduce nonsense words with "appropriate" affixes so the dissociation is reciprocal; it is not simply the case that stems are more strongly represented and easier to access; rather the two types of elements are differentially organized.
People who acquire their first language in normal fashion often demonstrate the psychological reality of morphophonological rules by non-standard instances of adherence to them during their early stages of acquisition. For example, the child who overregularizes the past tense rules producing "doed" for "did" demonstrates internalization of that rule. Children's creative morphologizing can be seen as evidence for their linguistic analysis of the language they hear. For example, the child who asks (about what a magnet will attract) "What does this [mæg NET] to?." putting stress on the second syllable, shows that the stress assignment rule that distinguishes between pairs like project (N) and project (V) bas been acquired and generalized.
In familial language impairment, the importance of morphophonological rules of affixation is made visible in that they break down. Recall that Goad and Rebellati pointed out how plural markers were realized by compounding rather than affixation in such individuals, and, as Gopnik has demonstrated, their SLT subjects can articulate rules for affixation but not apply them fully appropriately and automatically. Also, the general rules of compounding can break down in aphasic speakers as was reported of both German (Hittmair-Delazar, 1994) and Chinese (Packard, 1993).
That morphology itself can be dissociated from syntax is of interest. Recall the subjects with familial language impairment who had great difficulty with morphological markers of number, person, tense, and aspect but no problems with argument structure. Also, Miceli et al. (1983) have reported two Italian-speaking agrammatics, one of whom was impaired on morphosyntactic affixation but not on word order, whereas the other was impaired on word order but not on morphosyntactic affixation.
Syntax
The syntactic notion of hierarchically organized phrasal units appears to be determining where code-switching can take place in bilinguals. Zurif and Caramazza (1976) demonstrated that the ability to construct phrase-structure trees breaks down in agrammatics. Also, such concepts as our ability to appreciate syntactically ambiguous sentences (of the "Flying planes can be dangerous" type) are highlighted by the apparent breakdown of the ability to identify both interpretations of them in right-brain-damaged patients.
The importance or some notion of complexity of syntactic structure is seen by numerous reports across the populations studied that more complex syntax provides particular difficulties. This is seen in left-brain-damaged children as they recover, in Genie with respect to passive sentences as compared to active sentences, for example, and in agrammatics whose only comprehension problems, when any can be documented at all, are for more complex syntactic structures. The complexity of syntactic structures is often related to movement of constituents and the "traces" of that movement as we discussed in chapter 5. The "trace" itself has neither phonetic nor orthographic realization. In linguistic analyses the notion of trace has proved invaluable in explaining reference phenomena and the acceptability or unacceptability of many constructions. One kind of evidence for the reality of traces comes from the effects of this construct, posited for syntactic reasons, on phonology. For example, the question: Who do you want to visit? is ambiguous. The two possible answers are:
(1) I want to visit Jean. or
(2) I want Jean to visit.
However, it is only possible to use the colloquial "wanna" for sense (1) of this question. The contraction of "want to" to "wanna," is blocked on the second sense by the "trace" of the moved constituent "who":
YOU WANT TO VISIT WHO Who1 do you WANNA visit t1
YOU WANT WHO TO VISIT Who1 do you want t1 to visit
In the second underlying sentence, the trace between want and to prevents contraction. Grodzinsky has argued that it is sentences with traces embedded that pose particular difficulty for agrammatic patients to understand. If "traces" (as opposed to sentence length or number of arguments) can be shown to be the crucial factor in agrammatics' comprehension problems, this is consistent with their psychological reality.
The notion of thematic role – the relation a given noun has to the verb. e.g. subject, direct object, etc. – is also indicated in its apparent breakdown in agrammatic aphasia, as can be seen from such patients' error-prone performance on sentences with reversible passives. Agrammatic patients do not pick up the syntactic cues in a sentence like "the elephant was helped by the mouse" that indicate it was the mouse that was doing the helping.
A respected model of how sentences are produced is that of Merrill Garrett (e.g. 1983). Most crucial to that model is the notion that substantive lexical items like nouns and verbs are inserted into a sentence at a stage independent from that at which syntactic markers like inflectional suffixes are set into sentence frames. Garrett's model of processing syntax for production is given substantial support by the errors of sentence structure but not substantive lexical items in mild agrammatics, as well as the slip of the tongue of normals. That a separate level for surface syntax must exist is substantiated in the ability of demented patient HCEM to correct sentences to which errors of surface syntax have deliberately been added.
Lexicon and semantics
In linguistics there has been discussion of how words are "listed" in the internal lexicon. Are word-forms and their meanings dissociable? What other information is included with a lexical "item"? In particular it has been assumed that the characteristics of the words it must co-occur with, or those it can co-occur with are listed. Neurolinguists have, additionally, considered if there is more than one lexicon for the different modalities – one for written word forms to some extent distinct from the one for oral word forms. And psycholinguists and neurolinguists have asked how lexical items get "accessed," that is how they get "searched" when we need to "locate" them for use in a sentence.
That subjects, even aphasics, can distinguish real words in their language from words that are not in their language, is evidence for a certain psychological reality to the notion lexicon. It would appear that we indeed have at least one lexicon instantiated in our brain, with all the words we know in it.
Strong evidence for the reality of lexical search is our inability, sometimes, to find a word we are looking for. In anomic aphasics this phenomenon stands out as even more problematic. That it is the form of the word and not its meaning that is lost is evident from the fact that such patients can provide useful circumlocutions. In pure word-deafness, by contrast, patients recognize sound shapes of words as words in their language, but these "words" are divorced from the meanings. Such a double dissociation provides particularly strong evidence for the psychological reality of both systems of lexical organization.
The status of word-class categories and of affix categories as such has been dealt with above in the morphology section. It remains unclear whether affixable endings are listed in the lexicon with each of the words they can occur on or not. The current evidence suggests that inflectional affixes (like past tense ed) are unlikely to be listed with the words they may be appended to while derivational affixes (both prefixes like ex and pre and suffixes like tion and like in English) are likely to be.
Subtle differences in aphasic patients' abilities to name different categories of items have led to hypotheses about how the lexicon is organized in the brain. Semantic notions such as hierarchies of basic level words, particularly subordinate and superordinate words, may be seen to break down differentially, the reader will recall, in dementia. Severe dissociations between proper nouns and common nouns have been reported, between abstract and concrete nouns, and the related categories imageable and non-imageable nouns, and between mass nouns and count nouns (Grossman et al., 1993). A few highly educated aphasic patients have been reported to have formal speech spared, while informal, colloquial speech is more problematic. This is due, in part, to their lexical selections. Some theorize that it is word frequency effects that bring about these differences, but recently others have suggested it is the age of acquiring words, rather, that is reflected in such differential abilities.
The notion of semantic networks whereby related items are interlinked is evidenced by numerous forms of dissociated breakdown. Goodglass (1993), for example, has reported that aphasic patients often perform particularly poorly in comprehending bodypart names. Hart, Berndt, and Caramazza (1985) have reported a patient with brain damage suffered after encephalitis, who had particular anomia for fruit and vegetable names. Specific anomia for animal names has been reported by Hart and Gordon (1992). Further evidence for semantic linking comes from the semantic errors of deep dyslexia and of deep dysgraphia (recall the patient who wrote "scissors" when asked to write "stapler"). That some type of form-based network also links words is evident by the fact that cueing by the first phoneme of a word helps access it (or in the case of demented patients, access other words that begin with that phoneme). Further evidence comes from the finding that realword substitutions in aphasics' language are often phonologically similar to the target words, both in speech errors and in dysgraphia (e.g. eel or snake for snail).
Pragmatics
Pragmatic phenomena break down in dissociable ways in patients with right-brain damage and those with dementia. The ability to appreciate non-literal aspects of language (such as humor or statements requiring inference) breaks down while literal aspects of language pose no problem for such patients. The ability of normals but not these patients to appreciate ambiguity in sentences and homonymy in spoken words suggests the dissociable breakdown of these language skills.
All sorts or conversational rules are "proved" in the breaking of them by demented patients. These include error repair, topic maintenance, and the like. Certain ones, by contrast, are maintained until relatively late, such as the ability to take turns and to maintain eye contact. The improper use of anaphoric pronouns among such patients is a striking example of their particular impairment in the ability to monitor what the interlocutor knows.
Grice's first three conversational maxims (quantity, relevance, and manner; see chapter 7) are vividly exempliflied in their violation by demented patients who speak too much, tangentially, and in disorganized fashion, taking little account of what the interlocutor may fail to comprehend. Grice observed that the maxims reflect what listeners expect. When listening to patients in the middle-tolate stages of Alzheimer's disease, listeners find themselves befuddled by the breakdown of the conversation despite their best efforts. Only with a linguist's analysis (like that of Hamilton, 1994 a and b) can the specific components of breakdown be extracted.
With right-brain-damaged patients. by contrast. it is another set of pragmatic phenomena that strike the interlocutor as problematic, and thus strike the researcher as "real." The particular interaction of emotional content with discourse is evident in its problematic nature for right-brain-damaged patients, suggesting another distinct paralinguistic level that cuts across language, at least across lexicon, maturation, and discourse.
Language choice in bilinguals may be seen as a pragmatic ability that can independently break down. While this is rarely seen in aphasic bilinguals, it is particularly evident in demented patients who will inappropriately address an interlocutor in a language the interlocutor cannot understand.
Language structure inherent in written language systems
We believe that different types of orthographic systems will be related to differential use of brain space for written language representation. The differential breakdown of different orthographies in bilingual alexics is evidenced, for example, in the Hinshelwood study on the educated English man for whom reading classic Greek was best spared. The interface between phonology and those orthographic systems that link to it is made patent by the differential sparing/impairment of each. The Japanese orthographic system offers the opportunity, which some alexic patients demonstrate, to have either syllabic or logographlc elements particularly impaired. The existence of such patients is evidence that the two systems make use of different mechanisms in reading.
The fact that some children who are learning to read find decoding words particularly difficult while others find whole-word reading particularly difficult (and similar breakdown can be seen in some adult brain-damaged patients) is evidence for two distinct routes in normal reading.
The importance of spelling regularity in a language like English (whose spelling system is ideal only for historical information about words and documenting such dissociations in breakdown!) can be inferred from the data from surface dyslexics in whom the distinction is manifest. Even demented patients have more difficulty with their spared reading aloud of irregularly spelled (as compared to regularly spelled) words, und will make regularization errors in their writing of them.
Conclusion
In this chapter, we have highlighted the various ways in which the dissociation of abilities in language breakdown parallels the division of linguistic and psycholinguistic phenomena in to components for analysis. We have touched on some particular linguistic constructs and tied together evidence for their psychological reality from normal populations with new pieces of evidence from the language behavior of people with brain damage and from the normal populations mentioned in this book.
The frequent "fit" of linguistic theory with language performance in breakdown would suggest a well defined "map" of language in the human brain. However, the many articles cited in our chapters leave open countless questions for further research. Also, we must look for instances where the details of the data refuse to fit nicely into any theory. It is markedly easier to locate instances where linguistic categories fit the data than it is to claim that one has data which pattern in a way for which no linguistic theory to date provides an appropriate explanation.