The effects of syntactic structure on discourse comprehension in patients with parsing impairments

The effects of syntactic structure on discourse comprehension in patients with parsing impairments

BRAIN AND LANGUAGE 39, 206-234 (1990) The Effects of Syntactic Structure on Discourse Comprehension in Patients with Parsing Impairments DAVID CAP...

2MB Sizes 13 Downloads 72 Views

BRAIN

AND

LANGUAGE

39, 206-234 (1990)

The Effects of Syntactic Structure on Discourse Comprehension in Patients with Parsing Impairments DAVID CAPLAN AND KAREN

L. EVANS

Neuropsychology Laboratory, Massachusetts General Hospital, and School of Human Communication Disorders, McGill University Syntactic complexity was systematically manipulated in sentence and discourse comprehension tasks. Aphasic patients, but not normal controls, showed effects of syntactic structure in the sentence comprehension task. Neither aphasic nor control subjects showed significant differences between the comprehension of syntactically simple and complex stories. This pattern confirms the clinical impression that some discourses can be understood by patients who cannot utilize syntactically based sentence comprehension mechanisms. o 1990Academic press. ~ttc.

INTRODUCTION

Many researchers in psycholinguistics have claimed that normal sentence comprehension involves a syntactic analysis of incoming words. According to these models of comprehension, this syntactic analysis is accomplished by a special-purpose device known as a “parser.” A parser is a device that assigns syntactic structure and uses that structure to determine aspects of sentence meaning such as thematic roles (who did what to whom), coreference, etc. Many psycholinguistic studies have explored the nature of this syntactic route to sentence meaning (see Frazier, 1987, for review). The parser takes as input information about the syntactic categories of lexical items (whether a word is a noun, verb, preposition, etc.) and possibly syntactic combinatorial relationships of each word (so-called “subcategorization” facts, such as the fact that the This work was partially supported by the Medical Research Council of Canada (Grant MA-9671) and by a Chercheur-Boursier Award from the Fonds de la Recherche en Sante du Quebec to Dr. Caplan. Reprint requests should be addressed to David Caplan, Neuropsychology Laboratory, Vincent Bumham 725, Massachusetts General Hospital, Fruit Street, Boston, MA 02114. 206 0093-934x/90 $3.00 Copyright AI1 tights

8 1990 by Academic Press, Inc. of reproduction in any form reserved.

SYNTACTIC

STRUCTURE

AND DISCOURSE

COMPREHENSION

207

verb hit obligatorily requires a direct object) and constructs more elaborate syntactic structures. The evidence relevant to the existence and nature of parsing operations consists of the demonstration that variation in syntactic structure affects subjects’ abilities to understand a sentence, the time it takes them to do so, and/or their ability to accomplish a second task while engaged in understanding a sentence. Studies of parsing have used reaction times and error rates to entire sentences in comprehension tasks (Frazier, Clifton, & Randall, 1983), reaction times in anomaly detection tasks during sentence comprehension (Stowe, Tanenhaus, & Carlson, 1985; Tanenhaus, Stowe, & Carlson, 1985; Tanenhaus & Carlson, 1989), eye movement recordings (Rayner, Carlson, & Frazier, 1983), cerebral evoked potential recordings (Garnsey, Tanenhaus, & Chapman, in press), and other techniques. We cannot review all these data here, but shall present one line of argument regarding the nature of the parser, developed by Frazier (1987), that is based upon data from these various sources. Frazier’s case is based on the existence of preferences in interpretation of locally ambiguous syntactic structures. When these preferences lead to an overt misinterpretation of part of a sentence, they go by the name of “garden path” effects. For instance, subjects misanalyze sentences such as (1) (Bever, 1970): (1) The horse raced past the barn fell. Subjects analyze (1) as containing a main clause, the horse raced past the burn, and then do not find a grammatical role for the last word. In fact, the “correct” (i.e., grammatical) structure of (1) is the same as that of (2): (2) The horse racing past the barn fell. Sentence (2) is related to (3), with the relative pronoun and auxiliary deleted: (3) The horse that had been raced past the barn fell. Frazier (1987) argues that the difficulty in structuring (1) is due to the existence of syntactic operations that automatically misanalyze the first words in the sentence. She argues that the parser uses two principles in incorporating each new word in a sentence into a syntactic structure: Minimal Attachment and Late Closure. Minimal Attachment specifies that the parser does not postulate any potentially unnecessary nodes. Late Closure specifies that new items are attached to the phrase or clause being processed, if grammatically possible. Minimal attachment leads to the structure (a) in Fig. 1 for (l), rather than the ultimately correct structure (b). Minimal Attachment and Late Closure are general principles that lead to systematic preferences regarding the higher-order syntactic structures assigned to sequences of words. Both are “natural” principles, in the sense that they make first-pass parsing simpler. They preclude the cre-

208

CAPLAN

(6)

AND

EVANS

lS\

VP

;‘\ NP -A

The horse

7 raced

PP ( past the barn

I fell

FIG. 1. The syntactic structure of the sentence: The horse raced Structure A shows the misanalysis usually imposed, in which raced verb. Structure B shows the grammatically licensed structure.

past

the burn

fell.

is taken as a main

ation of potentially unneeded nodes and keep attachments local, at least initially, thereby presumably reducing both the number of structures created by the parser and the number of times the parser must move from place to place in the syntactic tree it is creating. Thus, both principles reduce the workload of the parser. One or the other, or both, create a whole host of preferences and garden path situations. A partial list, given by Frazier (1987), includes sentences (4)-(7). (4) John hit the girl with a book (preferred reading attaches with a book to hit, not to the girl). (5) Ernie kissed Marcie and her sister laughed (temporary misanalysis yields the interpretation that Ernie kissed Marcie and her sister). (6) Since Jay always jogs a mile seems like a short distance to him (temporary misanalysis yields the interpretation that Jay always jogs a mile). (7) Joyce said that Tom left yesterday (preferred reading attaches yesterday to left, not to said. This is the opposite pattern to that seen in (4) because Late Closure operates here while Minimal Attachment operates in (4)). These phenomena, and others, suggest that sentence interpretation involves syntactic parsing. In addition, these and related results suggest that parsing is automatic: it must occur in the presence of appropriate, attended, input. Sentence (1) must lead to a garden path effect; it cannot be attended to without that effect occurring (see Fodor, 1983, for discussion). ’ ’ There are alternative analyses of the facts just presented. One alternative to Frazier’s analysis is the claim that the phenomena described above are not due to the operation of

SYNTACTIC STRUCTURE AND DISCOURSE COMPREHENSION

209

Though there is good evidence for the existence of a parser that is automatically engaged during the sentence comprehension process, it is not clear that parsing is necessary for sentence comprehension to occur. In addition to using this syntactically based sentence comprehension mechanism, listeners can assign aspects of sentence meaning by making inferences based on the meanings of individual lexical items, by assigning certain aspects of meaning by nonsyntactic heuristics, and even by guessing as to the meaning of a sentence (Caramazza & Zurif, 1976; Ansell & Flowers, 1982; Wilcox, David, & Leonard, 1979; Boller, Cole, Vrtunski, Patterson, & Kim, 1979; Caplan & Hildebrandt, 1988; Hildebrandt, 1988; see also Grodzinsky & Marek, 1988; and Caramazza, 1988 for discussion). These nonsyntactic mechanisms are invoked in parallel with a syntactically based interpreter by normal subjects (Slobin, 1966; Caplan & Hildebrandt, 1988). The existence of both syntactic and nonsyntactic mechanisms for sentence comprehension raises questions regarding the role of syntactic comprehension in understanding multisentence context; i.e., discourse. Most models of discourse processing assume that sentence-level comprehension processes apply to yield various aspects of the propositional meaning of sentences, and that these meanings are entered into a mental model of the developing discourse (Kintsch & van Dijk, 1978). If the parser is not operating normally and the listener must rely on the nonsyntactic routes to understand sentences, it may be that the sentence comprehension process is slowed down or negatively affected in some way. If so, discourse comprehension may suffer. On the other hand, a parser but to lexical factors. There is considerable literature regarding the effects of verb structure and meaning on attachment (e.g., Ford, Bresnan, & Kaplan, 1982; McClelland, St. John, & Taraban, 1990), but the data supporting alternative theories are inconclusive (see Ferreira & Clifton, 1986; Holmes, 1987; Tyler, 1989; Frazier, 1989; Tanenhaus & Carlson, 1989 for discussion). Another analysis that has been contrasted with a parsing explanation of preference and garden path effects is based upon the existence of context effects upon garden path and preference phenomena. Crain and Steedman (1985) Altmann and Steedman (1988), and others have argued that garden paths are artifacts of presenting sentences out of context. They argue that biases such as those seen in (1) and (4) are not due to attachment preferences but rather to referential success. In (4), for instance, attaching fhe book to the girl implies that the context has identified more than one girl and that the sentence is referring to the girl with the book (of the girls mentioned in the context). Since presenting the sentence in isolation does not establish a context with more than one girl, the sentence is preferentially interpreted as attaching the book to hit. Evidence supporting this analysis would come from showing that context reverses preferences such as those seen in sentences like (I) and (4) in isolation. There is evidence that context changes many of these effects, and does so while sentences are being processed (Altmann & Steedman, 1988). However, that evidence is as yet controversial. Ferreira and Clifton (1986). Clifton and Ferreira (1987), Frazier (1987), and others have argued that at least some garden paths remain in effect despite strongly disambiguating context. Moreover, not all preferences can be explained this way; those in (S)-(7), due to Late Closure, cannot.

210

CAPLAN

AND

EVANS

discourse has structures and properties that are greatly influenced by pragmatic knowledge and thus the semantic interpretation of sentences in discourse is usually highly constrained (Ulatowska, 1981). Thus it is possible that, in context, sentences can be efficiently understood by entirely nonsyntactic means. Overall, it is unclear whether the relatively high degree of pragmatic constraint found in most discourses is sufficient to make syntactically based comprehension completely unnecessary for sentences in discourse. The research reported here investigates the question of whether the parser must be used for discourse to be understood normally. We studied brain-damaged patients with disorders of parsing. If disturbances of parsing affect subjects’ abilities to understand discourse, these patients should have disturbances at this level of comprehension. The purpose of this study was, therefore, to determine whether aphasic patients who have parsing disturbances also have disturbances of discourse comprehension, and, if so, whether the discourse comprehension disturbance arises because of the syntactic comprehension impairment. PREVIOUS STUDIES

An exhaustive review of the literature is not intended here. The purpose of this selective review is twofold: to establish that aphasia can result in specific deficits in syntactic comprehension and to examine some of the strategies individuals might use to circumvent syntactic comprehension in the discourse comprehension process. (a) Syntactic Comprehension Deficits and Nonsyntactic Sentence Comprehension Mechanisms in Aphasia Many studies document the existence of syntactic comprehension impairments in aphasic patients. The Token Test (DeRenzi & Vignolo, 1962) can be seen as a means to examine syntactic comprehension in aphasia. Performance on this series of context-free syntactic comprehension tasks has been shown to reliably discriminate between aphasic and nonaphasic individuals (DeRenzi & Faglioni, 1978). The number of modifier-noun pairs in a command affects patients’ performance, probably reflecting memory limitations in aphasia, and patients fail more markedly on the last section of the test in which a number of syntactic structures are represented. Other instruments have chosen a number of different syntactic features to study: negative and passive “transformations” (Shewan & Canter, 1971; Lasky, Weidner, & Johnson, 1976); prepositions signaling syntactic reversibility (Schwartz, Saffran, & Mat-in, 1980); idiomatic use of pronouns (Goodglass, Berko-Gleason, & Hyde, 1970); active vs. passive sentences (Brookshire & Nicholas, 1980; Schwartz, Saffran, & Marin, 1980); sentences in canonical form vs. noncanonical form and sentences

SYNTACTIC

STRUCTURE

AND DISCOURSE

COMPREHENSION

211

with one verb vs. two verbs (Caplan, Baker, & Dehaut, 1985); and sentences containing logical referential dependencies (Hildebrandt, Caplan, & Evans, 1987; Caplan & Hildebrandt, 1988). Sentence-picture matching or object manipulation tasks were chosen to eliminate oral production demands on the aphasic subjects. In all of these studies the aspects of syntactic complexity tested were shown to affect the performance of the aphasic subjects more than the nonaphasic subjects. While documenting disorders of syntactic comprehension, researchers have shown that there are situations in which aphasic individuals understand sentences without basing comprehension on the syntax of a sentence. Aphasic patients can employ a pragmatic knowledge-based strategy in a sentence comprehension task, and can thus understand “irreversible” sentences with syntactic structures that they cannot decode in reversible sentences (Caramazza & Zurif, 1976; Ansell & Flowers, 1982; Leslie, 1981). A pragmatic strategy would assign aspects of sentential semantic meaning such as thematic roles on the basis of the meanings of lexical items and real-world knowledge about the possible relations between items. Thus, this strategy would infer that the buy is the Agent and the cake the Theme in a sentence such as The cake was eaten by the boy, because boys are animate and can eat and cakes are inanimate and can be eaten, and not vice versa. A second means whereby patients who cannot utilize syntactic structures to understand sentences appear to arrive at meanings for sentences is to base their understanding of a sentence on heuristics applied to the linear sequence of nouns and verbs in a sentence (Caplan & Futter, 1986; Grodzinsky, 1986; Caplan & Hildebrandt, 1988). The use of such a strategy would lead to consistent misinterpretations of certain syntactic structures; e.g., a sentence such as The boy promised the man to shave could be consistently misinterpreted as indicating that the man is the Agent of shave. Finally, in certain tasks, patients who cannot assign sentential meanings via syntactically based comprehension mechanisms appear to guess at the meanings of sentences, and sometimes guess correctly (Hildebrandt, 1988). (6) Discourse Comprehension Comprehension in Aphasia

and Its Relationship

to Syntactic

All theories of discourse comprehension are based upon a multilevel model of meaning, and include a means of describing how semantic information is extracted from the presented text and the preexisting knowledge the comprehender brings to the task. The theories presuppose that linguistic and nonlinguistic analyses occur simultaneously at many different levels and interact with each other during comprehension (Frederiksen, 1986). The research on discourse comprehension in aphasia has drawn upon these theories to characterize disorders of this level of processing. However, it has not satisfactorily separated out syntactic effects

212

CAPLAN

AND

EVANS

at the sentence level from other variables or determinants of discourse comprehension performances. Stachowiak, Huber, Poek, & Kerschensteiner (1977) and Waller & Darley (1978) proposed that aphasics could comprehend spoken discourse through a process of contextualization that involves recruiting internal knowledge structures that guide a listener’s semantic construction of a discourse. They suggested that contextualization may remain intact in aphasia and be able to compensate for disorders in basic linguistic abilities. These contextualization mechanisms appeared to work despite a disrupted syntactic comprehension system as tested by the Token Test. However, the authors did not systematically manipulate the sentential material in their stories to determine if their aphasic subjects’ ability to use contextualization may have been affected by syntactic structure. Huber and Gleber (1982) did manipulate the text-level features in their study of aphasics’ abilities to access and apply higher level knowledge structures to aid comprehension. They designed stories which differed according to linguistic cohesion. They defined low cohesion to include simple syntactic structures, the use of only familiar content words, and no use of attributive and adverbial specifications. Their high-cohesion stories were characterized by complex syntax and many attributive and adverbial specifications. The high cohesion versions were also considerably longer than the low cohesion versions. Aphasic subjects had more overall difficulty with the tasks than controls but the degree of linguistic cohesion had no significant influence on the performance of either group. This may reflect the nonuse of syntactic structure in discourse comprehension in aphasia, but this is not a rigorous test of this hypothesis because the syntactically complex stories were more cohesive than the syntactically simple ones, and because the measures of syntactic complexity were not well defined. Brookshire and Nicholas (1984) and Wegner, Brookshire, and Nicholas (1984) also manipulated the coherence of discourse to determine if they could disrupt aphasic subjects’ abilities to establish a macrostructure. The authors found that aphasic subjects’ overall paragraph comprehension scores were worse than those of the control group or a right braindamaged control group, and that there was no relationship between subjects’ sentence comprehension abilities, as assessed by the Token Test, and their paragraph comprehension scores. These results support the previous studies’ claims that the ability to extract semantic information from discourse can be relatively preserved in aphasia, and that this cannot be predicted from performance on available sentence comprehension tests. However, the authors did not attempt to vary the syntactic complexity of their stories, and it is not known whether syntactic processing demands had an effect on aphasic subjects’ abilities to understand discourse.

SYNTACTIC

STRUCTURE

AND DISCOURSE

COMPREHENSION

213

Ulatowska and her colleagues applied a method of discourse analysis designed to investigate the effect of a knowledge-driven superstructure in discourse comprehension. They carried out a series of studies examining the story recall protocols of mild and moderately impaired aphasic subjects and one severely impaired aphasic using various types of discourse (Ulatowska, North, & Macaluso-Haynes, 1981; Ulatowska, Freedman-Stern, Weiss-Doyel, & Macaluso-Haynes, 1983a; Ulatowska, Weiss-Doyel, Freedman-Stern, & Macaluso-Haynes, 1983b; FreedmanStern, Ulatowska, Baker, & Delacoste, 1984). The authors have intensively analyzed their data with reference to categories defined by different narrative superstructures (Van Dijk, 1977; Van Dijk & Kintsch, 1983). Narrative superstructure categories were defined to include abstract, setting, complication, evaluation, result or resolution, and a code or moral (Ulatowska et al., 1981). Narrative superstructure is said to reflect knowledge of a conventional form of text type that guides the formation of the macrostructure. Main events of a story are assigned to the various categories listed above, thus facilitating the processing, remembering, and reproducing of stories (Van Dijk & Kintsch, 1983). In all of their experiments, the authors claimed that the aphasic subjects exhibited welldeveloped narrative superstructures. They could retell the stories and summarize them, including all the essential elements of the superstructure “explicitly, implicitly, or probably,” as judged by the experimenters (Ulatowska et al., 1983a). They claimed that their results supported the notion of a preservation of discourse structure with a selective reduction of information involving elaborative material. Performance on sentence comprehension tests did not predict how the aphasics performed on the discourse comprehension tasks. The authors conclude that aphasics can acquire information from discourse despite quite disturbed linguistic systems as tested by conventional procedures. However, it was observed in examining the written discourse of one severely aphasic subject that when the discourse required the use of complex sentence structures, his impaired syntactic ability adversely affected the success of the test. They concluded that, in the case of severely impaired language functions, the boundary separating preservation of discourse from discourse collapse and the factors involved are unclear (Freedman-Stern et al., 1984). The available studies of discourse comprehension in aphasia all argue that the ability to understand discourse can be relatively preserved in aphasia despite impaired sentence comprehension. The studies conclude that aphasics retain the ability to use context to resolve sentence level ambiguities in text (Stachowiak et al., 1977; Waller & Darley, 1978) and to impose discourse coherence by constructing a discourse macrostructure, even when coherence is disrupted experimentally (Waller & Darley, 1978; Huber & Gleber, 1982; Wegner et al., 1984) or attempts are made to obscure the macrostructure by increasing available microstructure

214

CAPLAN AND EVANS TABLE 1A CHARACTERISTICS

Age:

OF SUBJECTS

Range Mean Female Male Elementary Secondary Postsecondary

Sex: Education:

Aphasics (A’ = 16)

Controls (N = 16)

36-83 60.1 6 10 4 7 5

33-71 55.2 10 6 1 11 4

TABLE 1B OTHER CHARACTERISTICS

Aphasia type Broca Wemicke Conduction Dysarthria/ Dyspraxia Anomia Mixed/Other

OF APHASIC SUBJECTS

Severity 3 0 5 0

Mild Moderate Severe

0 8

Lesion site 4 11 1

Frontal Temporal Parietal Frontoparietal Frontotemporal Uncertain

1 1 3 1 1 9

information (Huber & Gleber, 1982). They show that aphasics preferentially recall macrostructure information whether this is implicitly or explicitly probed (Brookshire & Nicholas, 1984). Furthermore, aphasics appear to be able to utilize conventional superstructures, such as a narrative superstructure, to guide the comprehension and retrieval of discourse (Ulatowska et al., 1981, 1983; Ulatowska, 1981; Freedman-Stern et al., 1984). None of these studies, however, has specifically examined whether impaired syntactic comprehension could have an effect on any of the discourse variables studied if syntactic complexity was carefully manipulated and contrasted in discourse. The research reported here addresses this question. METHODS Subjects Subjects were 16 aphasics and 16 normal subjects of similar age, proportion of females and males, and education, as shown in Table IA. Criteria for selection of the aphasic subjects were (1) medical diagnosis of brain damage as a result of cerebral vascular accident, (2) no audiologically or medically documented hearing impairment, (3) diagnosis of aphasia by a speech pathologist using standardized tests of aphasia, (4) good oral comprehension of single words, as judged by performance on the BDAE and/or PPVT, (5) time postonset of aphasia no less than 3 months, (6) ability to attempt all items on the Syntactic Com-

SYNTACTIC

STRUCTURE AND DISCOURSE COMPREHENSION

215

TABLE 2 SENTENCE TYPES IN THE SYNTACTIC COMPREHENSIONTEST

Sentences with one verb Two-place verb sentences: Active (A): The elephant hit the monkey. Passive (P): The elephant was hit by the monkey. Three-place verb sentences: Dative (D): The elephant gave the monkey to the rabbit. Dative Passive (DP): The elephant was given to the monkey by the rabbit. Sentences with two verbs Conjoined (C): The elephant hit the monkey and hugged the rabbit. Subject Object Relative (SO): The elephant that the monkey hit hugged the rabbit. Object Subject Relative (OS): The elephant hit the monkey that hugged the rabbit.

prehension Test. Aphasic subjects ranged in age from 36 to 83 years (mean = 60.1). There were 10 men and 6 women. Their education ranged from elementary to postsecondary level with the majority having had a secondary or postsecondary education. These characteristics and other relevant features including lesion site, aphasia type, and severity of aphasia are summarized in Table 18. Two of the subjects had suffered cerebral vascular accidents to the right hemisphere. Since they met the selection criteria for aphasia they were included in the study. The nonaphasic subjects were spouses of aphasic subjects, and volunteers from the community. Criteria for selection of nonaphasic subjects.were (1) no documented history of brain damage and (2) no audiologically or medically documented hearing impairment. Nonaphasic subjects ranged in age from 33 to 71 years (mean = 55.1 years). There were 6 men and 10 women in this group. Their formal education ranged from elementary to postsecondary with the majority of subjects having had a secondary school or postsecondary education.

Stimulus Materials Syntactic Comprehension Test. A shortened version of the Comprehension Test reported by Caplan et al. (1985) was administered to the aphasic subjects. The test consisted of 35 sentences, five examples each of the seven sentence types listed in Table 2. The Syntactic Comprehension Test requires subjects to enact the thematic roles of agent and theme (and goal) for the nouns associated with each verb in each sentence. The test provides a reliable procedure for determining the syntactic comprehension abilities of aphasic individuals, because the stimulus sentences differ systematically in difficulty level according to their syntactic features. Several aspects of syntactic structure contribute to the difficulty of any particular sentence type. These include noncanonical word orders, number of arguments per verb, and number of verbs per sentence. These features are additive, and the relative complexity of a sentence type can be predicted from the number of features that occur in a sentence, as documented by Caplan et al. (1985). On the basis of these data a hierarchy of sentence difficulty was constructed for the sentence types studied. This hierarchy of sentence difficulty was used as a basis for choosing the syntactic structures to manipulate to establish syntactic complexity of the story stimuli, as described below. The availability of the data of Caplan et al. (1985) also made it possible to ensure

216

CAPLAN AND EVANS

that the subjects chosen for the present study did not markedly differ from the larger sample of aphasics previously studied using in these materials. Sfory Comprehension Test. Four stories were composed: two narratives and two folk tales. These stories are presented in the Appendix. A syntactically simple and a syntactically complex form of each story was prepared. All eight stories were approximately equal in length. The number of words ranged from 184 to 191. The two narratives (N) concerned events expected to be within the subjects’ general world knowledge. The other two stories were based on tales from classical mythology revised and adapted for the experiment. The subjects were not expected to have knowledge of these stories. These are referred to as folk tales (FT). Event and problem frame analyses were completed for all stories as developed by Frederiksen (1986). There were no major differences between the syntactically simple and syntactically complex forms of the stories with regard to the event frames. Between all of the eight stories there was a range of 27 to 32 events and 2 to 6 episodes. The narratives stories were similar in the number of events (27 and 32). Narrative 1 had 2 episodes and Narrative 2 had 5 episodes. The folk tales were also similar with 27 events vs. 29 events and 6 episodes each. There were no major differences between the syntactically simple and syntactically complex forms of the stories with regard to the problem frames (defined as containing a problem statement, a set of complications, a set of procedures, and a resolution/conclusion). Between all of the eight stories there was a range of 2 to 6 problems. The narrative stories had 2 and 5 problems. The folk tales had 6 problems each. For each of the four stories, syntactic complexity was varied without affecting semantic content. Each sentence in each story was semantically irreversible and/or comprehensible on the basis of context, and thus could be understood by applying a pragmatic knowledgebased strategy. For each sentence or set of sentences, the syntactically simple form of a story utilized sentence structures that were syntactically less complex than those utilized in the syntactically complex form of the same story. Syntactic complexity was determined by the hierarchy of sentence difficulty established by Caplan et al. (1985). For example, a simple sentence would be “Three women looked after the station,” and a syntactically more complex sentence would be “The station was looked after by three women.” In this example, an active sentence is contrasted with a passive sentence. Where particular words, phrases, or sentences were necessary in order to maintain the general coherence of the stories, these were kept constant across the two forms of each story. A set of 18 true/false probes was designed for each of the stories to assess comprehension. There were three categories of probe: Event frame probes, problem frame probes, and verbatim probes: Event frame probes. These probes were designed to assess an individual’s ability to retain the sequential order of events in a story. Probes were statements involving event sequences in the stories. Problem frame probes. These probes were designed to assess an individual’s ability to make necessary inferences based on the problems included in each story. The inferences were considered necessary in order for an individual to have fully understood the story. Verbatim probes. These probes contained direct statements from the story texts. For each story there were six examples of each type of probe, three requiring a true response and three a false response. All probes were short statements utilizing simple syntax.’ The number of words ranged from 3 to 15 words (mean = 8.3 words). The sets of probes for each story are presented in the Appendix. * All questions were in syntactically simple form to eliminate the possibility that patients might fail because of effects of syntactic comprehension disturbances upon their ability to comprehend the questions themselves.

SYNTACTIC

STRUCTURE AND DISCOURSE COMPREHENSION

217

These materials test subjects’ abilities to understand discourse passages in which no syntactic-based comprehension was required. These materials closely approximate naturally occurring discourses, in which most sentences can be interpreted by nonsyntactic routes. They therefore present a strong but appropriate test of the hypothesis that a normal parser is required for discourse comprehension to be achieved.

Procedure Each subject was tested in a single session of approximately 1 hr for the aphasic subjects and 30 min for the control subjects. The aphasic subjects received both the Syntactic Comprehension Test and the Story Comprehension Test. The control subjects received only the Story Comprehension Test because nonaphasic subjects were given the Syntactic Comprehension Test in a previous study (Caplan, 1987) and achieved near-perfect scores. Syntactic Comprehension Test. An object manipulation (play) task was used. This task requires the subject to manipulate toy animals in such a way as to demonstrate the thematic roles of nouns in verbally presented sentences. Five small toy animals were used. The animals occurring in each sentence were placed in front of the subject. The arrays were balanced such that each animal appeared in each thematic role an equal number of times and in each position in the array, left, right, or center, an equal number of times. The subject was instructed to listen to each sentence carefully and then to demonstrate who did what to whom by moving the animals. A practice and training session established that the subject made clear, interpretable responses. The sentences were presented at an average speaking rate with normal and appropriate intonation. The 35 sentences were presented in a pseudo-randomized order, such that there were not more than two consecutive examples of any sentence type. Responses were scored to reflect the patient’s interpretation of each sentence. Each noun in a sentence was labeled according to its sequential position in the sentence. Each noun was related to the arguments around each verb. For example, the formula 1,2 indicated that the elephant was Agent and the monkey Theme in: 2 1 The elephant hit the monkey. Using this method of scoring, the subject’s exact interpretations can be recorded to provide descriptive information concerning performance as well as total error score for each sentence type and for the entire test. For each aphasic subject the number of sentences correct was recorded for each of the 7 sentence types in addition to a total correct score out of a possible 35. For detailed description of this test, its method of presentation, and scoring, see Caplan et al. (1985). Story Comprehension Test. Each subject received all four stories. They received either the syntactically simple or the syntactically complex form of each narrative and each folk tale. If a subject received the syntactically simple version of one story, he automatically received the complex version of the other story, such that each subject heard a syntactically simple narrative and a syntactically complex narrative and a syntactically simple folk tale and a syntactically complex folk tale. The form of each story subjects received and the order in which the stories were presented were determined by assigning subjects to a story order condition. Eight story order conditions were established to ensure that each story type, narrative and folk tale, of each syntactic complexity level, simple and complex, was presented first, second, third, or fourth. There were two aphasics and two controls in each condition. Subjects were told that they would be told a short story and that following the story they would hear a number of statements. They would be required to decide if each statement was true, that it did happen in the story, or false, that it didn’t happen in the story. They were reassured that they would not be required to remember the stories verbatim because this was a test of how well they understood the stories, not a memory test. They were instructed that the story would only be read once but that they could have the probe statements repeated if they wished.

218

CAPLAN AND EVANS TABLE 3A

SENTENCE COMPREHENSION

TEST MEAN

SCORES AND STANDARD

TYPE FOR APHASIC

Sentence type Active Dative Passive Conjoined Object-Subject Dative Passive Subject-Object

DEVIATIONS

PER SENTENCE

SUBJECTS

(A) 0) m v-3 (OS) (DO (SO)

Mean

SD

4.81 3.88 3.19 3.00 2.81 I .69 1.25

0.40 1.15 1.91 1.59 1.68 1.82 1.39

TABLE 3B RESULTS OF TUKEY’S

Sentence types Active (A) Dative (D) Passive (P) Conjoined (C) Object-Subject (OS) Dative Passive (DP) Subject-Object (SO)

A

D

* * * *

* *

PROCEDURE

P

*

C

OS

DP

so

*

*

* *

* * * *

*

Note. Sentence types indicated by * have significantly different means @ < .05).

Each subject was first administered a practice story. This story was a short narrative followed by six true/false statements, two of each probe type, one requiring a true response and the other a false response. Subjects were encouraged to respond to all the probes and to use their knowledge of the story to decide on a response if they appeared unsure. Following the practice story each test story was read aloud, using normal intonation. After each story the probe statements were read aloud to the subjects. The subjects responded to each probe statement with a true/false judgment. When subjects were unwilling to respond they were encouraged to try, and if they did not respond within a reasonable time or insisted that they could not respond, then this was recorded and no credit was awarded for that statement. Probes were arranged for presentation by pseudorandomizing the order of probes such that no predictable pattern of question type or true/false response occurred. Each subject received all four sets of probes in the order dictated by the story order condition to which they had been assigned.

RESULTS Syntactic

Comprehension

Test

The sentence comprehension data consisted of a total Syntactic Comprehension Test score with a possible maximum of 35, and seven individual sentence scores with a possible maximum of 5 per sentence type. These data were analyzed using a one-way within-subjects analysis of variance. The within-subjects factor was sentence type, having seven

SYNTACTIC

STRUCTURE

AND DISCOURSE

COMPREHENSION

219

levels. The mean scores and standard deviations per sentence type for the aphasic subjects are presented in Table 3A. Analysis of variance showed a significant main effect for sentence type (F(6, 60) = 15.25, p < .OOl). A post hoc Tukey’s procedure was performed at a significance level of 0.05 to determine which sentence means were significantly different, as indicated in Table 3B. The significant differences between the sentence means used in the present study were similar to those reported by Caplan et al. (1985). Story Comprehension

Test

The story comprehension data consisted of a total Story Comprehension Test score with a maximum of 72, total narrative and folk tale story scores with a maximum of 36 each, total syntactically simple and complex story scores with a maximum of 36 each, and total story probe scores for each of three probe types for each story with a maximum of 6 each. The mean number of event, inference and verbatim questions answered correctly on each of the four stories was calculated separately for the aphasic and control subjects. These data were analyzed using a 2 groups (aphasics vs. controls) x 2 stories (narrative vs. folk-tale) x 2 levels of syntactic complexity (simple vs. complex) X 3 question types (event frame vs. problem frame inference vs. verbatim) analysis of variance.3 The between-group factor was subject type and the within-group factors were story type, syntactic complexity, and question type. Table 4 shows the mean number of correct responses per question type for each story type and for each level of syntactic complexity. The results of the analysis of variance are as follows. There were significant main effects of group (F(1, 30) = 14.74, p < .OOl); story type (F(1, 30) = 7.78; p < 001); and question type (F(2, 60) = 29.27; p < .OOl). There was a significant interaction between story and question type (F(2, 60) = 14.53, p < .OOl). The main effect of syntactic complexity and the remaining interactions were not significant. As seen in Table 4, the significant main effect for groups was due to the overall poorer performance of the aphasic subjects (mean = 4.36) as compared to the control subjects (mean = 5.13). The main effect for story type was due to the relative difficulty of the folk tales (mean = 4.56) as opposed to the narratives (mean = 4.93). In order to interpret the significant main effect of question type and the significant interaction of story type x question type, post hoc Tukey ’ Although there were two passages of each story type, passage was not included as a factor in the analysis. This was because the two passages of each story type were developed in matched sets such that they were approximately equal in number of words and semantic frames. In addition, initial inspection of the data showed that the means for the two passages of each story type were virtually identical.

4.76 4.94 4.84

4.28 4.34 4.31

Narratives Folk tales Combined

Narratives Folk tales Combined

TABLE 4

5.69 4.63 5.16 5.38 4.09 4.73

4.88 4.63 4.75

Aphasics and controls 4.91 4.31 4.57 4.25 4.74 4.28

5.38 4.47 4.92

5.06 4.91 4.98

5.31 5.31 5.31

Controls 5.21 5.02 5.11

5.56 4.81 5.19

4.81 4.50 4.66

5.19 4.13 4.66

I 5.38 4.28 4.83

E 4.30 4.29 4.30

V 5.11 5.11 5.11

Aphasics and controls Simple and complex (combined)

5.06 3.56 4.31

3.75 3.88 3.81

Aphasics 4.60 4.13 4.36

V

I

I

E

Total simple

Complex

OF CORRECT RESPONSES OF SUBJECTS ON THE STORY COMPREHENSION

Nore. E, Event Frame Probe; I, Problem Frame Inference Probe; V, Verbatim Probe.

Narratives Folk tales Combined

3.81 3.75 3.78

Narratives Folk tales Combined

E

NUMBER

Simple

MEAN

5.16 5.31 5.23

5.44 5.56 5.50

4.88 5.06 4.97

V

TEST

4.95 4.55 4.75

5.33 4.94 5.14

4.56 4.17 4.36

Total complex

4.93 4.56 4.75

5.27 4.98 5.13

4.58 4.15 4.36

Total S+ C

SYNTACTIC

221

STRUCTURE AND DISCOURSE COMPREHENSION TABLE 5

SPEARMAN’S

RANK ORDER CORRELATIONS BETWEEN SYNTACTIC COMPREHENSION SCORES

COMPREHENSION

Total syntactic comprehension score Rho Total S C T score Simple stories Complex stories Narratives Folk tales Simple narrative Complex narrative Simple folk tale Complex folk tale

AND STORY

P + DP + SO Rho

.52* .51* .32 .45 .4l .30

.53* .43 .47

.39 .53* .I8

Note. S C T, Story Comprehension Test; P + DP + SO, combined sentence scores for passive, dative passive, and subject-object sentences.

* p < .05.

tests were performed using a significance level of .05. The Tukey procedure performed for the significant main effect of question type showed that each question type was significantly different from the others. The highest mean score was on the verbatim probes (mean = 5.11) and the lowest on the event frame probes (mean = 4.30). The mean for the problem frame inference probes was intermediate (4.83). The significant interaction between story x question type resulted from a difference between the story types on problem frame inference questions but not on event frame or verbatim questions. The folk tale inferences (mean = 4.28) were significantly more difficult than the narrative story inferences (mean = 5.38). Correlations Between Syntactic Comprehension

Comprehension

and Story

To examine the relationship between aphasic subjects’ Syntactic Comprehension Test scores and their Story Comprehension Test scores, a series of Spearman’s rank order correlations were performed, as shown in Table 5. The total Syntactic Comprehension Test score was correlated with the total Story Comprehension Test score, and with eight scores derived from the Story Comprehension Test. In addition, the combined sentence score from the passive, dative passive, and subject object sentences was correlated with the total Story Comprehension Test score and with both the syntactically simple and complex story scores. This particular combination of sentences has been found to be highly discrim-

222

CAPLAN AND EVANS TABLE 6 RANK ORDERING OF APHASIC SUBJECTS ACCORDING TO THEIR F’ERF~RMANCE ON THE SENTENCE AND THE STORY COMPREHENSION TESTS

Subject 6 12 2 4 14 13 8 1 5 3 10 9 16 11 7 15

Sentence test score

Sentence rank

Story test score

story rank

9 13 14 14 16 16 16 19 24 24 26 26 26 28 29 30

1 2 3.5 3.5 6.0 6.0 6.0 8.0 9.5 9.5 12.0 12.0 12.0 14.0 15.0 16.0

52 58 45 46 45 51 57 52 51 61 47 49 63 61 60 55

8 12 1.5 3 1.5 6.5 11.0 9.0 6.5 14.5 4.0 5.0 16.0 14.5 13.0 10.0

mating between normal and aphasic subjects in previous studies (Caplan, 1987) and to predict how aphasic subjects perform on the total Syntactic Comprehension Test. Dative Passive and Subject Object sentences were also found to be significantly more difficult than other sentence types for aphasic subjects in this study and in previous studies (Caplan et al., 1985). Thus it was hypothesized that this score may be particularly related to aphasic performance on the syntactically complex stories. All correlations were positive. The aphasic subjects’ total score on the Syntactic Comprehension Test was significantly correlated with the total Story Comprehension Test score (rho = .52), the total simple story score (rho = Sl), and the total simple folk tale score (rho = 53). The passive + dative passive + subject-object sentence score was significantly correlated with the total Story Comprehension Test score (rho = .53) but not with either the syntactically simple or complex story scores. For several reasons, these correlational analyses may be misleading. The syntactic comprehension and story comprehension tests may not be equivalent in difficulty. The story comprehension test requires only yes/no responses, while the syntactic comprehension test allows for a number of responses; therefore, if patients were guessing on both tests, they would be expected to do considerably better by chance on the story comprehension test than on the syntax comprehension test. To rule out the possibility that these correlational analyses were significantly influenced by a small set of atypical subject scores, the individual subjects’

SYNTACTIC

STRUCTURE AND DISCOURSE COMPREHENSION

223

data on the two tests were compared. Table 6 shows the score and rank ordering of each subject on each of the two tests. Visual inspection of the table shows that the correlation between performance on the sentence comprehension test and the story comprehension test is due to a general relationship between performance across all subjects, not to isolated performance of a few individual patients. Despite any differences in complexity of the two tasks, and different susceptibilities of the two tasks to correct responses being obtained through guessing, these data indicate that performance on these two tasks is highly correlated. DISCUSSION

The main purpose of this study was to investigate the relationship between aphasic subjects’ ability to comprehend syntactic structures and their ability to comprehend discourse. This experiment differed from previous studies in that syntactic comprehension was measured with a battery that tested specific syntactic constructions and in that the syntactic complexity of the discourse was varied without affecting the semantic content or structure of the discourse to be comprehended. Given that the patients were shown to have syntactic comprehension impairments, the question of greatest interest was whether the syntactically complex versions of the stories were more difficult to comprehend than the syntactically simple versions. It was hypothesized that patients’ syntactic comprehension deficits would result in their obtaining lower story comprehension test scores for the syntactically complex versions of the stories than the syntactically simple versions. However, neither aphasics nor nonaphasics experienced greater difficulty with the syntactically complex stories than with the syntactically simple stories. This was true regardless of story type or question type. Each group had almost identical scores for simple versus complex narratives and for simple versus complex folk tales. Subjects did not appear to be affected by the syntactic complexity of the stories whether the story was a narrative or a folk tale, even though the folk tales had a more complex discourse structure. They responded to questions about the content of syntactically complex and syntactically simple discourses equally well for both types of discourse. These results hold for patients with all levels of syntactic comprehension capacities. Although these comprehension results are unequivocal, several additional observations merit consideration. Subjects frequently reported that they perceived themselves to be experiencing greater difficulty with the syntactically complex stories. They judged these stories to be longer and harder to retain. This was particularly evident in the subjects’ assessment of the complex folk tales. This suggests the possibility that the complex stories did in fact place greater processing demands on the subjects, but that the story comprehension probes provided cues which facilitated

224

CAPLAN AND EVANS

comprehension. Objective support for this interpretation is provided by a recent study by Waters & Caplan (1986). They presented the stories constructed for this study to a group of normal young adults as a reading task. They measured reading times for each story. Their subjects then answered the same series of true-false questions as the subjects in the present study. Though Waters & Caplan also did not find a difference in their subjects’ success on the story probes, there was a significant difference in reading time for the syntactically complex stories as opposed to the syntactically simple stories. This may indicate that additional processing demands were made by the syntactically complex stories at the initial decoding stage. It is also possible that the verification task used to probe story comprehension in both these studies facilitates comprehension more than a story recall task would. If subjects were obligated to organize and retrieve the relevant story information without the cues necessarily provided in a verification task, the increased normal processing documented by Waters and Caplan for syntactically complex stories might result in a difference between stories differing in syntactic complexity. In addition, the verification task lends itself to guessing, which may have improved performance on the complex versions of the stories; a recall task would eliminate guessing effects. However, this type of assessment would necessarily exclude many aphasics due to the production demands of a story recall procedure and the complicating factor of increased memory requirements. Other methods for assessing discourse comprehension in aphasia are needed for further determination of the possible effects of syntactic complexity. The present study, however, does indicate that, in common discourse structures containing semantically and discourseconstrained sentences, the syntactic complexity of the sentences in the discourse does not have an independent effect upon aphasic patients’ abilities to answer questions about the content of a passage, regardless of a patient’s ability to comprehend sentences by a syntactic route. The second question addressed by this research was whether the aphasic subjects’ performance on the Syntactic Comprehension Test predicted their performance on the Story Comprehension Test. There was a relatively low but significant positive correlation between these two tests. In general, if an aphasic had difficulty comprehending sentences, he had a similar disturbance in discourse comprehension. However, two aphasic subjects displayed severely impaired sentence comprehension but relatively unimpaired discourse comprehension. Interestingly, both these subjects suffered right hemisphere CVAs resulting in nonfluent aphasia of mild-moderate severity. Both were agrammatic and experienced some difficulty comprehending complex ideational material. One may speculate that these subjects had language representation in both hemispheres and that their discourse comprehension was left relatively

SYNTACTIC

STRUCTURE AND DISCOURSE COMPREHENSION

225

intact by the event. Inspection of the results for two subjects who displayed the opposite trends in their performance on the two tests revealed no striking similarities between these two subjects. The existence of patients who did poorly on syntactic comprehension but relatively well at discourse comprehension indicates that discourse in which all sentences are semantically irreversible can be understood without recourse to a syntactically based sentence comprehension mechanism. The existence of patients who show the opposite pattern-relatively good syntactic comprehension but poor discourse comprehension-indicates that the processes involved in structuring and representing discourse may be impaired without an equally significant level of impairment of those involved in structuring and comprehending semantically reversible sentences.4 The total picture that emerges regarding the relationship between syntactic comprehension abilities and discourse comprehension is complex. Several conclusions can be drawn. First, the double dissociation between 4 Several other issues can also be addressed by this research. One is whether narratives are easier than folk tales for aphasics to understand. It was hypothesized that the folk tales would be more difficult than the narratives due to their relatively unfamiliar content and their more complex problem frames. The results confirmed this expectation: the folk tales were more difficult than the narratives for both aphasics and nonaphasics. This occurred because the problem frame inferences for the narratives were easier than those based on the folk tales. There were no differences between narrative and folk tales for event frame probes and verbatim recall probes. The familiar content and simpler structure of narratives appeared to result in making it easier for both aphasics and nonaphasics to respond to problem inference probes, but folk tales were no more difficult for either group with regard to comprehension of event frames or verbatim recall. Another issue is whether event frame information and directly stated information is significantly easier for aphasic subjects to retain than the indirectly stated inferences based on the problem frame. The problem frame inferences were actually less difficult than event frame recall for both aphasics and nonaphasics when results were combined for narratives and folk tales. Tukey’s test shows that this difference was attributable to better performance for combined groups on problem frame inferences for narratives, with no difference between event and problem frame probes for folk tales. Thus, the higher level comprehension skills of aphasics were not selectively impaired in complex discourse, and were actually superior to lower level (event frame) skills in simpler discourse. Aphasics showed the same trends as nonaphasics in both instances. Previous studies by Brookshire & Nicholas (1984) and Wegner, Brookshire and Nicholas (1984) found no difference between these different types of information, but all of their stories were short narratives (similar to those used in the present study) that had simple problem frames. Ulatowska and her colleagues (1981; 1983a,b), using a modified fable as one of their story stimuli, found that their aphasic subjects, when asked to recall and summarize the presented discourse, produced few propositions which evaluated story events and formed conclusions and judgments. All these results suggest that aphasic patients are better able to draw inferences in simple discourses than in complex ones. However, type and complexity of discourse are confounded in all these studies. It would be necessary to present aphasic and nonaphasic subjects with both narratives and folk tales with both simple and complex problem frames to separate the effects of these two factors.

226

CAPLAN

AND

EVANS

performance on these two functions seen in several individual patients in this study clearly indicates that each of these functions has operations and/or representations not contained in the other. This is an unsurprising finding. Second, the fact that syntactically more complex passages were, on average, no more difficult than syntactically simple passages for these aphasic patients to understand indicates that discourse comprehension can proceed without utilization of a syntactically based sentence comprehension mechanism-at least for passages which contain semantically constrained sentences. Third, the overall correlation between performance on the syntactic comprehension test and performance on the discourse comprehension test needs to be explained (assuming, as argued above, that it is not due to a statistical artifact related to the differences in scoring performances on the two tests). There are two possible explanations of this relationship: it can be due to some neuroanatomical relationship between the two tasks or it can be due to some functional relationship between them. We cannot be certain which of these factors is responsible for this correlation of deficits. If the correlation is due to the lesions in our patients affecting functionally independent processing systems involved in syntactic and discourse comprehension, the correlation between the impairments on these tasks should be diluted by the addition of patients with different lesion sites and types to a subject population; future research will be needed to see if this occurs. If the two impairments are due to an impairment in one or more cognitive functions that are involved in both syntactic and discourse comprehension, then we should be able to specify what these shared components of processing are. We can speculate regarding the functional mechanisms that may be shared between parsing and discourse comprehension. The purpose of this study was to investigate the possibility that syntactic comprehension itself was needed for discourse comprehension. Since patients do equally well on the syntactically complex and syntactically simple versions of the story comprehension test, this cannot be the case: whatever is shared between the two tasks must be separate from the ability to accomplish syntactic comprehension itself. A second possibility is that disturbances affecting lexical comprehension affect patients’ performances on both tasks. The fact that patients were screened to have a single word receptive vocabulary that was adequate to accomplish both tasks provides evidence that a deficit in lexical comprehension processing is not the basis for the related performances on the two tasks.5 A third possibility ’ Lexical processing has not been examined in detail in these patients, either for the full range of vocabulary items used in the passages or with respect to the time course of activation of lexical semantic features. It does remain possible that the relationship between these two tasks may be partially due to patients’ disturbances in lexical comprehension.

SYNTACTIC

STRUCTURE AND DISCOURSE COMPREHENSION

227

is that the sentence and discourse comprehension share some processing resources, perhaps to be conceived of as part of the “central executive” of a Working Memory system (Baddeley, 1986), and that patients’ reductions in this resource system partially determine the magnitude of their failure on both tasks.‘j In summary, this study indicates that patients with syntactic comprehension failures have a reliably associated disturbance in discourse comprehension. However, this related problem does not appear to be directly due to their inability to understand syntactically complex sentences, since their performance is not worse in syntactically complex passages than in syntactically simple passages. This conclusion must be qualified in two ways: (1) the measure of discourse comprehension used a recognition memory task, and a more demanding task-such as story recall-may show differences in performance as a function of the differences in syntactic structure of sentences in a discourse; and (2) all the sentences in the discourses used were either semantically irreversible or pragmatically constrained in the context of the previous discourse and it may be that syntactic comprehension abilities are required when a larger number of sentences are semantically reversible and not easily predicted from the previous discourse. These possibilities will require additional research to be resolved. APPENDIX Discourse Passages and Probe Questions

Narrative I: Simple The robber approached the quiet gas station. He looked in the window. Three women looked after the station. So, he decided it would be easy to rob. The robber grabbed his gun. He kicked open the door. Lorraine, Kathy and Sue were talking inside. The gunman surprised them. But, he did not frighten them. He saw the safe in the office. He demanded money. So, the robber ordered the women to the back of the office. Then, he pushed Lorraine to the safe. He gave Sue an empty bag. Lorraine opened the safe. Sue put the money into the bag. The robber watched Sue greedily. He didn’t watch Kathy. Kathy waited for the right opportunity. She picked up an oil can. She threw it. It hit the robber’s arm. He dropped the gun. Lorraine pulled the gunman to the floor. Sue hit him. Lorraine held the robber who now yelled for help. Kathy called the police. Soon they arrived. A policeman handcuffed the robber who looked at the women in disbelief. The policeman led the failed robber to the car. He had made a serious mistake. However, the fact that patients clearly understood the vocabulary used in the sentence comprehension task makes a lexical basis for a related failure on that task and the story comprehension task quite unlikely. 6 It is possible that the correlation between syntactic comprehension and story comprehension occurred because test performance on both tasks was correlated with other variables such as age, severity of aphasia, and education. However, inspection of these characteristics for individual subjects in relation to the ranked scores showed no obvious relationships of this sort.

228

CAPLAN AND EVANS

Narrative

1: Complex

It was a quiet gas station that the robber approached. The station was looked after by three women. So, he decided that it would be easy to rob. The robber grabbed his gun and kicked open the door. Lorraine, Kathy and Sue were talking inside. They were surprised but not frightened. The gunman saw the safe in the office and demanded money. The women were ordered to the back of the office. Lorraine was pushed towards the safe. Sue was given an empty bag. The safe was opened by Lorraine. Then the money was put into the bag by Sue. The robber, who watched Sue greedily, didn’t watch Kathy. Kathy waited for the right opportunity. She picked up an oil can and threw it. It hit the gunman’s arm and he dropped the gun. Sue hit the gunman who Lorraine pulled to the floor. The robber who Lorraine held now yelled for help. The police were called by Kathy. The robber who the policeman handcuffed looked at the women in disbelief. Then, the failed robber was led to the police car. A serious mistake had been made by that robber. E I I E V E V I V I E V V I E V I E

I. 2. 3. 4. 5. 6. 7. 8. 9. IO. I I. 12. 13. 14. IS. 16. 17. 18.

The robber watched Sue then Kathy picked up the oil can. The robbery was successful. The women attacked the robber. The robber grabbed his gun then kicked open the door. The robber approached the quiet gas station. The robber yelled for help then the oil can hit him. Three men looked after the store. The oil can hit the robber accidentally. Kathy waited for the right opportunity. The robber thought that women could not stop a robbery. Lorraine pulled the robber to the floor then he dropped his gun. He saw the cash register in the office. The robber watched Sue greedily. The robber escaped with the money. Lorraine opened the safe then Sue put the money in the bag. The policeman led the failed murderer to the car. The robber was arrested. The robber pushed the women to the back of the office then he demanded money.

Narrative

T F T T T F F F T T F F T F T F T F

2: Simple

One Saturday the New York police department received a call. Mrs. Brown reported an alligator in her backyard. The sergeant sent officers Smith and Davidson to the scene. They thought it was a joke. On arrival a worried Mrs. Brown directed them to the backyard. They had a big surprise. A large alligator was staring at them. It was wearing a red collar. Smith coaxed the alligator into a comer. It hissed and snapped. Davidson got a rope. He wanted to put the rope through the collar. He approached the alligator. But the angry animal headed towards him. Smith drew his gun. Suddenly, the policemen heard a shout. A man rushed into the yard. He walked up to the alligator. He snapped a leash onto the red collar. Then, he calmly led his pet out of the yard. The astonished policemen followed. Apparently the man kept the alligator in his swimming pool. That day he left the gate open. The alligator had escaped. Smith gave the man a ticket. Davidson filled in the report. The boys at the station would never believe this.

Narrative

2: Complex

One Saturday a call was received by the New York police. An alligator was reported by a Mrs. Brown in her backyard. Officers Smith and Davidson were sent to the scene

SYNTACTIC

STRUCTURE AND DISCOURSE COMPREHENSION

229

by the sergeant. They thought it was a joke. On arrival the policemen were directed to the backyard. They had a surprise. The large alligator wearing a red collar was staring at them. The alligator, who hissed and snapped, was coaxed into a comer by Smith. Davidson got a rope and wanted to put it through the collar. He approached the angry alligator that headed towards him. Smith drew his gun. Suddenly a shout was heard. A man rushed into the yard. The man, who walked up to the alligator, snapped a leash onto the collar. The astonished policemen followed the man who calmly led his pet away. Apparently the alligator was kept in his swimming pool. That day, the gate was left open by the man. So, the alligator had escaped. The man was given a ticket by Smith and the report was filled out by Davidson. The boys at the station would never believe this. Answer Questions for Narrative 2 I V E V E I E V E V I E E V I I E V

1. The alligator was easy for the policemen to catch. 2. The boys at the station would never believe this. 3. The man left the gate open then the alligator escaped. 4. Davidson got a cage. 5. When Mrs. Brown called the police they believed her story. 6. The policeman intended to shoot the alligator. 7. Davidson filled in the report then Smith gave the man a ticket. 8. A Mrs. Brown reported an alligator in her back yard. 9. The man led the alligator out of the yard then put it on a leash. 10. The man kept the alligator in his room. 11. The owner was not afraid of the alligator. 12. Davidson approached the alligator then it headed towards him. 13. The man walked right up to the alligator then the policemen followed. 14. A large tiger was staring at them. IS. The man was not punished for letting his alligator escape. 16. The policemen were surprised to see the alligator. 17. Smith drew his gun then the alligator was coaxed into a corner. 18. The policemen heard a shout.

F T T F T T F T F F T T T F F T F T

Folk Tale 1: Simple According to myth, dwarves own all buried treasure. A dwarf, Andvari, owned one such hidden treasure. Many men desired Andvari’s treasure. But, no one could catch him. Andvari would turn himself into a fish. Then, he jumped into the sea. Thus, he escaped. A great king particularly desired Andvari’s treasure. This king ordered his wizard, Loki, to catch Andvari. Loki tried many tricks. He could not catch the dwarf. One night, Loki made a magic fishing net. That night he captured the dwarf. Andvari was forced to give his treasure to the wizard. Andvari placed a curse on the treasure. It would only bring misfortune. As a result, the treasure obsessed Loki. The king demanded the treasure. But, Loki would not give it to him. Loki killed the king. He kept the treasure for himself. So, he turned himself into a dragon. He guarded the treasure day and night. Meanwhile, Andvari hunted Loki tirelessly. Eventually he found the dragon. He killed it. Then he took his treasure to a new hiding place. The treasure never tempted man again.

Folk Tale 1: Complex According to myth, all buried treasure is owned by dwarves. It was one such hidden treasure that Andvari the dwarf owned. Andvari’s treasure was desired by many men. But he could be caught by no one. Andvari, who would turn himself into a fish, jumped into the sea and escaped. Andvari’s treasure was particularly desired by a great king. The king’s wizard, Loki, was ordered to catch him. Loki, who tried many tricks, couldn’t

230

CAPLAN AND EVANS

catch the dwarf. One night, he made a magic fishing net that caught Andvari. The dwarf, who the wizard had trapped, was forced to give his treasure to Loki. Quickly a curse was placed on the treasure by Andvari. As a result, this treasure that would only bring misfortune obsessed Loki. He killed the king who demanded the treasure, and kept it for himself. But he worried about his prize and could never enjoy it. So, Loki, who turned himself into a dragon, guarded the treasure day and night. Meanwhile, Andvari hunted Loki tirelessly and eventually found and killed the dragon. The treasure was taken to a new hiding place. Man was never tempted by it again. Questions for Folk Tale 1 Answer E 1. Loki made a magic fishing net then forced Andvari to give him the treasure. I 2. Loki was loyal to the king. I 3. Andvari’s treasure helped him get his treasure back. E 4. The king desired Andvari’s treasure then Loki tried to catch Andvari. V 5. The treasure obsessed Loki. E 6. Loki gave the king the treasure then he killed him. V 7. Giants own all buried treasure. I 8. Andvari wanted to give his treasure to Loki. V 9. Andvari hunted Loki tirelessly. I 10. The treasure was safe forever in Andvari’s new hiding place. E 11. Andvari placed a curse on his treasure then Loki captured him. V 12. Andvari would turn himself into a bird. V 13. The king ordered his wizard, Loki, to catch Andvati. I 14. Loki caught Andvari the first time he tried. E 15. Andvari killed the dragon then the treasure never tempted man again. V 16. Loki turned himself into a giant. I 17. Andvari was in the shape of a fish when Loki caught him. E 18. Loki guarded the treasure then he turned himself into a dragon.

T F T T T F F F T T F F T F T F T F

Folk Tale 2: Simple According to myth, Odin was the god of poetry. He made a magic potion. Odin gave it to many men. It made all these men wonderful poets. An evil giant called Suttang demanded some potion. But Odin refused to give it to him. So, Suttang stole all of it. He hid it in an underground chamber. His daughter, Brunhilde, guarded it. Odin searched for his potion. Finally he found the underground chamber. But he could not get in. A kindly giant showed Odin the hidden door. Odin quickly changed his appearance. He fooled Suttang. Odin gave Suttang many compliments. He won the giant’s trust. Brunhilde fell in love with Odin. Each night, she gave Odin some of the potion. Secretly, Odin stole all of it. When it was gone, Odin turned himself into an eagle. He flew away. Once home Odin put the potion into two vases. He would give some only to a few poets he favored. But some drops of potion fell to earth. Odin forbid men to drink these drops. But some men did. These men also became poets. Thus the world had enough poets.

Folk Tale 2: Complex According to myth, the god of poetry was Odin. A magic potion that Odin made turned many men into wonderful poets. Some potion was demanded by an evil giant called Suttang. Suttang, who was refused by Odin, stole all of the potion. It was hidden in an underground chamber. There it was guarded by Suttang’s daughter, Brunhilde. Odin searched for his potion. He finally found the chamber and couldn’t get in. The hidden door was shown to Odin by a kindly giant. Odin, who quickly changed his appearance, fooled Suttang. Odin

SYNTACTIC

STRUCTURE AND DISCOURSE COMPREHENSION

231

gave Suttang many compliments. Thus he won Suttang’s trust and Brunhilde fell in love with him. Each night some potion was given to Odin by Brunhilde. Secretly, all of it was stolen by Odin. When it was gone, Odin turned himself into an eagle and flew away. Once home, the potion that Odin put into two vases was given only to a few poets he favored. But a little potion fell to earth. Some men drank these drops that Odin forbid them to drink. These men also became poets and so the world had enough poets. Answer Questions for Folk Tale 2 I V E V I I E V E V I E E V I I E V

1. Odin wanted Suttang to have some potion. 2. Odin gave Suttang many compliments. 3. Some men drank the spilled drops of potion then they became poets. 4. Odin was the god of thunder. 5. Odin returned all of his stolen potion to his home. 6. Odin wanted to choose which men became poets himself. 7. Odin flew away then Brunhilde fell in love with him. 8. Some drops of potion fell to earth. 9. Odin fooled Suttang then he found the underground chamber. IO. Odin turned himself into a mouse. 11. Odin tricked Brunhilde. 12. Suttang stole the potion then Brunhilde guarded it. 13. Odin searched for his potion then he changed his appearance. 14. A kindly elf showed Odin the hidden staircase. 15. Odin found the hidden door himself. 16. Some men disobeyed Odin. 17. Men became poets then Odin gave them some potion. 18. Odin searched for his potion.

F T T F F T F T F F T T T F F T F T

REFERENCES Altmann, G., & Steedman, M. J. 1988. Interaction with context in human syntactic processing. Cognition, 30, 191-238. Ansell, B. J., & Flowers, C. R. 1982. Aphasic adults’ use of heuristic and structural linguistic cues for sentence analysis. Brain and Language 16, 61-72. Baddeley, A. 1986. Working memory. Oxford: Oxford Univ. Press. Berko-Gleason, J., Goodglass, H., Obler, L., Green, E., Hyde, M. R., & Weintraub, S. 1980. Narrative strategies of aphasic and normal-speaking subjects. Journal of Speech and Hearing

Research,

23, 370-382.

Bever, T. 1970.The cognitive basis for linguistic structures. In J. R. Hayes (Ed.), Cognition and the development of language. New York: Wiley. Boiler, F., Cole, M., Vrtunski, P. B., Patterson, M., & Kim, Y. 1979.Parahnguistic aspects of auditory comprehension in aphasia. Bruin and Language, 7, 164-174. Brookshire, R. H., & Nicholas, L. E. 1980. Verification of active and passive sentences by aphasic and non-aphasic subjects. Journal of Speech and Hearing Research, 23, 878-893. Brookshire, R. H., & Nicholas, L. E. 1984. Comprehension of directly and indirectly stated main ideas and details by brain-damaged and non-brain damaged listeners. Brain and Language, 21, 21-36. Caplan, D. 1986. Patterns of sentence comprehension in aphasia, unpublished manuscript. Caplan, D. 1987. Discrimination of normal and aphasic subjects on a test of syntactic comprehension. Neuropsychologia, U, 173-184. Caplan, D., Baker, C., & Dehaut, F. 1985. Syntactic determinants of sentence comprehension in aphasia. Cognition, 21, 117-175.

CAPLAN AND EVANS

232

Caplan, D., & Futter, C. 1986. Assignment of thematic roles to nouns in sentence comprehension by an agrammatic patient. Bruin and Language, 27, 117-134. Caplan, D., & Hildebrandt, N. 1988. Disorders of syntactic comprehension. Cambridge, MA: Bradford Books. Caramazza, A. 1988. When is enough enough? A Comment on Grodzinsky and Marek’s “Algorithmic and Heuristic Processes Revisited.” Brain and Language, 33, 3X1-399. Caramazza, A., & Zurit, E. B. 1976. Dissociation of algorithmic and heuristic processes in language comprehension: Evidence from aphasia. Brain and Language, 3, 572-582. Clifton, C., & Ferreira, F. 1987. Modularity in sentence comprehension. In J. Garfield (Ed.), Modularity in knowledge representation and natural language processing. Cambridge: Cambridge University Press. Crain, S., & Steedman, M. 1985. On not being led up the garden-path: The use of context by the psychological parser. In D. Dowty, L. Karttunen, & H. Zwicky (Eds.), Natural language parsing. Cambridge: Cambridge Univ. Press. DeRenzi, E., & Fagiolini, P. 1978. Normative data and screening power of shortened version of the Token Test. Cortex, 14, 41-49. DeRenzi, E., & Vignolo, L. 1962. The Token Test: A sensitive test to detect receptive disturbances in aphasics. Brain 85, 665-678. Donin-Frederiksen, J. 1985, April. Discourse processing of oral and written language in hearing-impaired children. Paper presented at the America1 Educational Research Association, Chicago. Ferreira, F., & Clifton, C. 1986. The independence of syntactic processing. Journal of Memory

and Language,

25, 348-368.

Fodor, J. 1983. The modularity of mind. Cambridge, MA: MIT Press. Ford, M., Bresnan, J., & Kaplan, R. 1982:A competence-based theory of syntactic closure. In J. Bresnan (Ed.), The mental representation of grammatical relations. Cambridge, MA: MIT Press. Frazier, L. 1987. Sentence processing: A tutorial review. In M. Colheart (Ed.), The psychology of reading, attention and performance XII. London: Erlbaum. Frazier, L. 1989. Against lexical generation of syntax. In W. Marslen-Wilson (Ed.), Lexical representation and process. Cambridge, MA: MIT Press. Pp. 505-528. Frazier, L., Clifton, C., & Randall, J., 1983. Filling gaps: Decision principles and structure in sentence comprehension. Cognition, 13, 187-222. Frederiksen, C. H. 1975.Representing logical and semantic structure of knowledge acquired from discourse. Cognitive Psychology, 7, 371-458. Frederiksen, C. H. 1977. Structure and process in discourse production and comprehension. In M. A. Just & P. A. Carpenter (Eds.), Cognitive processes in comprehension. Hillsdale, NJ: Erlbaum. Pp. 313-322. Frederiksen, C. H. 1985, June. Comprehension of different types of text structure. Paper presented at the Canadian Psychological Association meeting, Halifax, Nova Scotia. Frederiksen, C. H. 1986. Cognitive models and discourse analysis. In C. R. Cooper & S. Greenbaum (Eds.), Studying writing; Linguistic approaches. Beverly Hills, CA: Sage. Pp. 144-161. Freedman-Stem, R., Ulatowska, H. K., Baker, T., & Delacoste, C. 1984. Disruption of written language in aphasia: A case study. Brain and Language, 22, 181-205. Gamsey, S. M., Tanenhaus, M. K., & Chapman, R. M. Evoked potentials and the study of sentence comprehension. Journal of Psycholinguistic Research, in press. Goodglass, H., Berko-Gleason, J., & Hyde, M. 1970. Some dimensions of auditory language comprehension in aphasia. Journal of Speech and Hearing Research, 13, 595-606. Grodzinsky, Y. 1986. Language deficits in a theory of syntax. Bruin and Language, 27, 135-159.

SYNTACTIC

STRUCTURE AND DISCOURSE COMPREHENSION

233

Grodzinsky, Y., & Marek 1988. Algorithmic and heuristic processes revisited. Brain and Language, 33, 216-225. Hildebrandt, N. 1988. Non-syntactic interpretation in aphasic sentence comprehension. Paper presented at the Johns Hopkins Cognitive Neuropsychology Workshop, Baltimore. Hildebrandt, N., Caplan, D., and Evans, K. 1987. The man left ti without a trace: A case 4, 257study of aphasic processing of empty categories. Cognitive Neuropsychology, 302. Holmes, V. M. 1987. Syntactic parsing: In search of the garden path. In M. Coltheart (Ed.), Attention and Performance XII: The Psychology of Reading. London: Erlbaum. Pp. 587-600. Huber, W., & Gleber, J. 1982. Linguistic and nonlinguistic processing of narratives in aphasia. Brain and Language 16, l-18. Khosrowshahi-Pouratzai, F. 1984. Comprehension and question answering: A comparative study. Unpublished master’s thesis, McGill University, Montreal. Kintsch, W. 1974. The representation of meaning in memory. Hillsdale, NJ: Erlbaum. Kintsch, W. 1977. On comprehending stories. In M. A. Just and P. A. Carpenter (Eds.), Cognitive processes in comprehension. Hillsdale, NJ: Erlbaum. Pp. 33-62. Kintsch, W., and van Dijk, D. A. 1978. Toward a model of text comprehension and production. Psychology Review, 85, 363-394. Kormos, L. 1984. Question understanding: Effects on children’s comprehension of stories. Unpublished doctoral thesis, McGill University, Montreal. Lasky, E. Z., Weidner, W. E., & Johnson, J. P. 1976. Influence of linguistic complexity, rate of presentation and interphase pause time on auditory-verbal comprehension of adult aphasic patients. Bruin and Language, 3, 386-395. Leslie, C. M. 1981. The interactive effects of syntax, pragmatics and task difficulty in aphasic language comprehension. Doctoral dissertation, Columbia University. clniversity Microfilms International, 8113528. Linebarger, M. C., Schwartz, M. F., & Saffran, E. M. 1983. Sensitivity to grammatical structure in so-called agrammatic subjects. Cognition, 13, 361-392. Mandler, J. M., & Johnson, N. S. 1977. Remembrance of things parsed: Story structure and recall. Cognitive Psychology 9, 11l-151. McClelland, J. L., St. John, M., & Taraban, R. 1990. Sentence comprehension: A parallel distributed processing approach, Language and Cognitive Processes, 4, 287-336. Paris, S. G. 1975. Integration and inference in children’s comprehension and memory. In F. Restle, R. Shiffrin, J. Castehan, H. Lindman, & D. Pisoni (Eds.), Cognitive Theory. Hillsdale, NJ: Erlbaum. Vol. 1, pp. 223-245. Paris, S. G., & Upton, L. R. 1976.Children’s memory for inferential relationships in prose. Child Development,

47, 660-668.

Pastore, R. E., & Scheirer, C. J. 1974. Signal detection theory: Considerations for general application. Psychological Bulletin, 81, 945-958. Rayner, K., Carlson, M., & Frazier, L. 1983. The interaction of syntax and semantics during sentence processing: Eye movements in the analysis of semantically biased sentences. Journal of Verbal Learning and Verbal Behaviour, 22, 358-374. Schwartz, M., Saffran, E., and Marin, 0. 1980. The word order problem in agrammatism I: Comprehension. Brain and Language, 10, 249-262. Shewan, C. M., & Canter, G. J. 1971. Effects of vocabulary, syntax and sentence length on auditory comprehension in aphasic adults. Cortex, 7, 209-226. Slobin, D. 1966. Grammatical transformations and sentence comprehension in childhood and adulthood. Journal of Verbal Learning and Verbal Behavior, 2, 219-227. Stachowiak, F. J., Huber, W., Poek, K., & Kerschensteiner, M. 1977.Text comprehension in aphasia. Brain and Language, 4, 177-195.

234

CAPLAN AND EVANS

Stowe, L., Tanenhaus, M. K., & Carlson, G. 1985. Parsing Filler-Gap Sentences. Paper presented at 26th Annual Meeting of the Psychonomic Society, Amherst, MA. Tanenhaus, M. K., & Carlson, G. N. 1989. Lexical structure and language comprehension, in W. Marslen-Wilson (Ed.), Lexical representation and process. Cambridge, MA: MIT Press. Pp. 529-561. Tanenhaus, M. K., Stowe, L. A., & Carlson, G. 1985.The interaction of lexical expectation and pragmatics in parsing filler-gap constructions. Proceedings of the Seventh Annual Cognitive Science Society Conference. Irvine, CA. Thomdyke, P. 1977. Cognitive structures in comprehension and memory of narrative discourse. Cognitive Psychology, 9, 77-l 10. Tyler, L. 1989. The role of lexical representations in language comprehension. In W. Marslen-Wilson (Ed.), Lexical Representation and Process. Cambridge, MA: MIT Press. Ulatowska, H. K. 1981. Neurolinguistic approaches to aphasia. Seminars in Speech, Language and Hearing, 2, 269-281. Ulatowska, H. K., Freedman-Stern, R., Weiss-Doyel, A., & Macahtso-Haynes, S. 1983a. Production of narrative discourse in aphasia. Brain and Language, 19, 317-334. Ulatowska, H. K., North, A. J., & Macaluso-Haynes, S. 1981. Production of narrative and procedural discourse in aphasia. Brain and Language, 13, 345-371. Ulatowska, H. K., Weiss-Doyel, A., Freedman-Stern, R., & Macahtso-Haynes, S. 1983b. Production of procedural discourse in aphasia. Brain and Language, 18, 315-341. Van Dijk, T. A. 1977. Semantic macro structures and knowledge frames in discourse comprehension. In M. A. Just & P. A. Carpenter (Eds.), Cognitive processes in comprehension. Hillsdale, NJ: Erlbaum. Pp. 3-32. Van Dijk, T. A. 1981. Episodes as units of discourse analysis. Georgetown University round table on languages and linguistics, analyzing discourse: Text and talk, 177195. Van Dijk, T. A., & Kintsch, W. 1978. Cognitive psychology and discourse: Recalling and summarizing stories. In W. U. Dressier (Ed.), Current trends in text linguistics. Berlin, NY: Walter de Gruyter. Van Dijk, T. A., & Kintsch, W. 1983. Strategies of discourse comprehension. New York: Academic Press. Waller, M. R., & Darley, F. L. 1978. The influence of context on the auditory comprehension of paragraphs by aphasic subjects. Journal of Speech and Hearing Research, 21, 732-745. Waters, G., & Caplan, D. 1986. Reading times for Story Comprehension Test, unpublished. Wegner, M. L., Brookshire, R. H., & Nicholas, L. E. 1984. Comprehension of main ideas and details in coherent and noncoherent discourse by aphasic and non-aphasic listeners. Brain and Language, 21, 37-51. Wilcox, M. J., Davis, G. A., & Leonard, L. B. 1979. Aphasics’ comprehension of contextually conveyed meaning. Brain and Language, 6, 362-377.