BRAIN AND LANGUAGE 48, 22[—237 (1995) Thematic Role Assignment in Two Severely Aphasic Patients: Associations and Dissociations MICHAEL WEINRICH, DENISE MCCALL, AND CAROL WEBER Department of Neurology and Rehabilitation, University of Maryland at Baltimore Two severely aphasic patients were compared in their abilities to comprehend and produce locative prepositional phrases and reversible S-V-O sentences using English and C-VIC, a computer-based iconographic communication system. One patient demonstrated a significant dissociation between his performance in inter- preting symbol order in C-VIC prepositional phrases vs. S-V-O sentences. Pa- tients were able to comprehend order in C-VIC S-V-O sentences significantly better than they were able to assign symbol order when they produced these sentences. These data suggest that the procedures for assigning thematic roles to nouns in sentences are at least partially distinct for comprehension and production and that the ordering of nouns around prepositions involves conceptual processes distinct from those involved in ordering nouns around verbs. © 1995 Academic Press, Inc. INTRODUCTION For language to have meaning, syntactic structures must be mapped onto objects and concepts in the real world, or at least onto mental projec- tions of them (Jackendoff, 1983). This problem is an issue for language production just as much as it is for language comprehension. There is fairly general agreement among contemporary theorists that this mapping must occur at an early stage in language production before phonological coding is specified (Levelt, 1989, 1992); however, much remains unspeci- fied regarding this process. For example, Garrett (1992) suggested that mapping of thematic roles onto the constituent nouns in a sentence occurs as a procedure carried out at the functional level of language representa- tion, while Bock (1987) held that the lexical representations of verbs carry with them the information necessary to assign these roles. Schwartz, Saffran, and Marin (1980) assumed that the mapping function for locative Supported by NIH Grant DCD 00856. We thank Dr. Rita Berndt and Charlotte Mitchum for the use of their testing materials and for their helpful comments. Address correspon- dence and reprint requests to Michael Weinrich, M.D., Department of Neurology and Rehabilitation, 22 S. Greene St., Baltimore, MD 21201. Fax: 410-554-5217. 221 0093-934X/95 $6.00 Copyright © 1995 by Academic Press, Inc. All rights of reproduction in any form reserved. 222 WEINRICH, MCCALL, AND WEBER prepositions is identical with that for verbs. Mitchum, Haendiges, and Berndt (in press) propose that while the information used to map thematic roles in comprehension and production may be shared, the actual proce- dures used in language comprehension may be distinct from those used in features production. ! Despite these uncertainties and differences of opinion, it is now clear that a failure to map thematic roles correctly can produce significant deficits in language comprehension and production. Indeed, the concept of a mapping deficit was invoked by Saffran and her colleagues (Saffran, 1982; Schwartz et al., 1980; Saffran & Schwartz, 1988) to account for the fact that some agrammatic patients who could comprehend non- reversible sentences (e.g., ‘‘the boy hits the building’) had grave difficul- ties in comprehending logically reversible sentences (e.g., ‘‘the boy hits the girl’’). The alternative hypothesis of a central syntactic deficit to explain such deficits (Berndt & Caramazza, 1980; Grodzinsky, Swinney, & Zurif, 1985) has foundered on the observation that these patients are quite sensitive to subtle grammatical errors in sentences spoken to them (Linebarger, Saffran, & Schwartz, 1983; Berndt, Salasoo, Mitchum, & Blumstein, 1988; Shankweiler, Crain, Gorrell, & Tuller, 1989). In addi- tion, Caramazza and Miceli (1991) have described a patient who appears to have difficulty in mapping between thematic and syntactic roles but who demonstrates no impairments in morphosyntactic processing. Perhaps most important, a variety of investigators using different tech- niques have demonstrated that mapping impairments in some aphasic patients may be remediable and that such remediation leads to demon- strable and generalizable language improvements. Byng (1988) reported the effects of training patient BRB, a chronic agrammatic aphasic who could not correctly interpret reversible sentences. After training on com- prehension of reversible locative prepositional phrases, he was able to comprehend reversible S-V-O and passive sentences, and his language production improved markedly. Jones (1986) used a different treatment paradigm with another chronic agrammatic patient, but reported very similar results. Saffran et al. (1992) described a comprehensive program to remediate mapping deficits, with dramatic improvements noted in sev- eral patients. These findings are encouraging, and the generalization from comprehension tasks to language production further suggests that at least part of the mechanism(s) used to map thematic roles during language comprehension is also used during language production. These authors differ, however, in their analysis of the mapping process. Byng and Jones suggested that mapping impairments are lexically based: i.e., loss of spe- cific rules associated with each lexical entry of sentence predicates, while 'In this report we follow the psycholinguistic approach of Jackendoff (1983) and Levelt (1989) and do not distinguish between semantic and thematic roles. ASSOCIATIONS AND DISSOCIATIONS 223 Saffran et al. argued that the generalization from prepositional phrase to verbs and from comprehension to production argue for a loss of a more general mapping procedure. “Errors in thematic role assignment could result from impairment in syntactic analysis (‘parsing’), loss of verb-specific mapping rules, or fail- ure to properly integrate these two forms of information’’ (Saffran et al., 1992). In an alternative formulation, Miyake, Carpenter, and Just (1993) proposed that a reduction in processing resources may result in patterns of aphasic comprehension. Memory and lexical access impairments which may occur in aphasic patients (Berndt, 1991) could amplify these deficits. In this context we have recently described the performance of a group of aphasic patients on tasks involving locative prepositional phrases implemented on C-VIC, an iconographic, computer-based alter- native communication system (Weinrich et al., 1993a) for severely apha- sic patients. The computer system utilizes a very simplified set of syntac- tic rules, eliminates the need for morphological and phonological processing, and minimizes memory demands. Under these conditions, patient DP, a globally aphasic patient, demonstrated that he was able to comprehend and produce locative prepositions, but had difficulty decod- ing symbol order in reversible locative prepositional phrases. In contrast, patient MF, a Broca’s aphasic, who was initially unable to interpret sym- bol order in reversible locative prepositional phrases, recovered the abil- ity to interpret these phrases after training with C-VIC. In addition, his verbal production of these phrases improved markedly. The C-VIC system employs an iconographic language implemented on a Macintosh computer (Steele et al., 1989; Weinrich, 1991). The ‘‘lexi- con’’ of C-VIC is organized into grammatical markers, animate nouns (subjects), verbs, prepositions, modifiers and common nouns. Patients are taught to select symbols from these categories and to arrange them into linear strings delimited by grammatical markers. A simple, unambig- uous, active syntax is employed. There is no tense marking or passive voice option in C-VIC. Both preposition and verb icons are animated to aid in patient comprehension. A C-VIC prepositional phrase is illustrated in Fig. 1. Representational nouns are arranged around a preposition. The first noun carries the role of Reference Object and the second noun of Place (following the notation of Jackendoff, 1983). Fig. 2 illustrates a simple active sentence with conjoined subjects. Note that there are no articles used in C-VIC and there are no markings to indicate singular or plural use of verbs. While the task of assembling or comprehending a C-VIC construction can not be said to correspond exactly to a specific level of natural lan- guage processing, we have, nonetheless, argued (Weinrich et al., 1993a) that the dimensions of the processing problems encountered by patients using C-VIC are similar to those encountered in natural language, and in 224 WEINRICH, MCCALL, AND WEBER L Phrase 2 Fic. L. A locative prepositional phrase in C-VIC. Nouns are arranged around a preposi- tion. The preposition is animated so that the square moves into the surrounding shape. the case of thematic role assignment, are identical. C-VIC does not re- quire phonetic representations or articulation, but a C-VIC construction does at least contain the equivalent of a surface structure representation, i.e., lemmas which are semantically and syntactically specified. Thus, analagous to Garrett's (1992) analysis of potential loci of linguistic failures in speech production, failure of C-VIC production may result from failure to produce a lemma representation, or failure to manipulate the syntactic rules associated with these representations. The issues of accessing word- forms from lemma representations are obviated by the structure of the C-VIC interface. The parallels between C-VIC production and natural language production are underscored by the considerable improvements in natural language production observed in several chronically aphasic patients after C-VIC training (Weinrich et al., 1993a,b). It is simply not conceivable that such improvements could occur unless some aspects of processing in C-VIC and natural language are shared. Thus, mapping deficits comparable to those found in auditory—verbal comprehension and production of locative prepositional phrases may be demonstrated using C-VIC. These deficits are demonstrated in the pres- ence of intact lexical access, reduced memory load, greatly simplified syntactic structures and with problems of phonetic representation and articulation eliminated. In the case of patient DP, the mapping deficit Phrase 2 Fic. 2. A simple active sentence with conjoined subjects in C-VIC. The verb icon is animated so that the revolver fires. Note the absence of articles, There are no markings to indicate tense or subject—verb agreement in C-VIC. ASSOCIATIONS AND DISSOCIATIONS 225 appears intractable, while patient MF’s deficit improved with training. In the present study we seek to extend the analysis of these patients’ abili- ties to map thematic roles onto C-VIC and English constructions, and we present data from these two patients on their abilities to interpret and produce reversible S-V-O constructions in C-VIC and English. SUBJECTS DP is a 45-year-old right-handed man who suffered a left hemisphere stroke in October, 1988. A carotid arteriogram revealed high grade steno- sis in the internal carotid artery and he underwent carotid endarterectomy in December, 1988, only to suffer a postoperative stroke. A repeat left carotid endarterectomy was followed by another postoperative stroke in December 1989. Since his last stroke he has suffered from a moderate right hemiparesis, a seizure disorder well controlled on phenytoin and phenobarbitol, and a global aphasia. His word discrimination on the BDAE was at the 20th percentile, comprehension of body parts and com- mands was 0, and word picture matching below the 40th percentile. He was able to repeat 50% of single words, was unable to repeat any phrases, and his spontaneous speech was limited to a few stereotypies. DP was unable to write at all. He performed quite well on the Raven’s Progressive Matrices (Raven, 1976), obtaining 32/36 correct, while on the Wisconsin Card Sort (Heaton, 1981) he obtained only 3 categories (normal for age = 5) and his performance was characterized by marked perseveration. MF is a 44-year-old right-handed man who suffered a single left hemi- sphere stroke 5 years ago. Since his stroke he has been left with a severe right hemiparesis and a chronic Broca’s aphasia. MF has relatively good comprehension: auditory comprehension measures on the BDAE were at the 60th percentile and reading comprehension above the 90th percen- tile. His spontaneous speech was very sparse and telegraphic. His verbal agility on the BDAE was 0. He was able to repeat 90% of single words but only 25% of high frequency phrases. His writing ability was very poor. MF completed all of the Raven’s Matrices correctly, but like DP demonstrated marked perseveration and achieved only two categories on the Wisconsin Card Sort. METHODS Training Protocol Patients were trained according to a defined treatment protocol which has been described in detail (Weinrich et al., 1993). They were seen for I hr sessions three times per week. The stages of patient training included: (1) Mechanics of finding and selecting icons (2) Matching of common and proper nouns (3) Grammatical mood markers 226 WEINRICH, MCCALL, AND WEBER {4) S-V and S-V-O constructions with a limited verb set (5) Preposition training After the preposition stage of training was completed (18 sessions for DP and 6 sessions for MF), the patients received further C-VIC training on nine logically reversible verbs: “lift,” ‘‘wash,"’ “‘ride,”’ ‘‘give,’’ “‘pull,’’ *‘touch,"’ ‘‘shoot,’’ ‘‘push,”’ and ‘hit.”” They were trained to describe actions performed by the therapist as well as actions portrayed in pictures. Additionally, they were trained to follow simple commands employing these verbs. Several sessions focussed on explicitly training reversible constructions. After training, D.P. demonstrated good ability to use most of these verbs receptively and expressively. However, he continued to have difficulty differentiating among push, pull, and lift. MF was able to use all of the verbs without difficulty, Once their performance reached a plateau with no further improvements (54 sessions for DP and 57 sessions for MF), we assessed their abilities to describe a novel set of pictures. Performance after Training We have analyzed data from DP’s C-VIC training over the past year of therapy sessions after DP had completed the verb training and assessment. In this period he produced 544 C-VIC S-V-O sentences on which he made only 6 (1%) word order errors. Further, increas- ing the number of nouns in the sentence did not result in a deterioration of performance. Dative constructions did not cause DP to make substantially more word order errors (2/58 or 3%). In constructions with conjoined agents and themes he produced 2 order errors out of 42 constructions (§%). DP produced 60 C-VIC sentences which required a prepositional phrase (sentences employing the verb ‘‘put’’, e.g., ‘Denise put hat on table’’). He produced 14 word order errors in these sentences for an error rate of 23%. Note here that we have scored word order errors only on those sentences with the correct lexical elements (see Weinrich et al., 1993). Immediately after the second set of assessments below were completed, MF began inten- sive training on verbal production using the C-VIC system as a cueing system. For compari- son with patient DP we therefore report on his performance in his last month of verb training (7 sessions). During this period he produced 82 C-VIC sentences containing verbs. There were 3 word order errors in his constructions. All of these involved his attempts to elaborate simple sentences, e.g., to describe a picture of a man standing at a sink washing his face with a sponge (target sentence ‘‘man wash face’’) he produced ‘‘Man wash sponge using face sink."" MF made no order errors on simple S-V-O sentences nor on dative constructions. He tended to omit optional prepositional phrases and occasionally inserted prepositions inappropriately, ¢.g., ‘‘Woman drink in coffee using cup.”” Production of Reversible S-V-O Sentences Nine professionally drawn line drawings were constructed portraying an action between two animate subjects, e.g., “‘The man shoots the woman." The pictures constructed were designed to elicit the nine verbs previously trained. D.P. was presented with single drawings and asked to describe them using either C-VIC or English. The order of trials was counter- balanced for C-VIC and verbal production. In one-half of the trials he was provided with a structure template, ‘‘Who-Do-What-Who?"’ on the computer screen. For the remaining trials, task instructions were ‘“Te}i me what’s happening in the picture?’’ Each picture was tested four times in each domain. No feedback was given after responses. Two practice trials preceded the test items. Verbal responses were recorded on tape for off-line analysis. C-VIC responses were transcribed directly. Responses were first scored for presence of the correct verbs and nouns. Only those responses which contained the target verbs and nouns, or semantically acceptable alternatives, were scored for word order. Patients were not penalized for repairs or perseverations. ASSOCIATIONS AND DISSOCIATIONS 227 Comprehension of Reversible S-V-O Sentences Comprehension of reversible sentences was tested after the production assessment had been completed. The nine drawings used to test production were complemented by 27 additional drawings. A total of 18 pairs of reversible scenarios, e.g., “‘The man shoots the woman" and ‘‘The woman shoots the man’ were constructed. Each verb appeared in two sets of drawings which portrayed the verb in different contexts. Normative data were obtained for all of the test pictures. Five neurologically intact adults were asked to describe, ‘What's happening in the picture?”’ For 19/36 of the pictures presented, all five normal subjects spontaneously produced the target verb. Subjects produced the target verb or a synonym for the target (e.g., punch/hit) for 8/36 pictures. Nine of the pictures presented elicited verbs that were not considered synonyms for the target verb (e.g., carry/lift), How- ever, when presented with a choice of four verbs, including the target and three other verbs from the C-VIC verb category (e.g., lift, touch, put, give) at least three of the five normal subjects selected the correct target for these pictures. Based on these data, one set of pictures was eliminated from the testing materials. Drawings were presented in pairs to patients. They were simultaneously presented with a sentence which accurately described only one of the drawings and were required to point to the correct drawing. The order of presentation of sentences and the position of the correct picture were randomized. Sessions were counterbalanced between C-VIC and ver- bal presentations. In each session they were tested on all of the 17 pairs. Patients were tested in each modality twice on separate days. Comprehension of S-V-O Sentences with Conjoined Agents and Themes We wished to exclude the possibility that patients were using some heuristic rather than actually mapping thematic roles onto nouns in C-VIC and English constructions. To do this we constructed a task which required patients to discriminate between conjoined agents and patients. Use of this type of construction enabled assessment of patients’ ability to order multiple noun phrases in the absence of complex grammatical structures unavailable in C-VIC. Ten sets of three line drawings were developed. Each set portrayed three actors performing an action with three different permutations of agents and patients. In each set the two test sentences contained the same ordering of nouns but the roles played by them was different. Thus, patients could not choose the correct picture merely by remembering the order of nouns presented to them in a sentence. A third distractor drawing contained nouns in a different order: “The woman and the boy shoot the man.”’ “The woman shoots the boy and the man.”’ “The boy shoots the woman and the man." Patients were presented with a sentence in C-VIC or English and were required to select a drawing in response to the sentence. They were tested in each modality in separate sessions. In each session the patients were tested on the ten sets of drawings and received no feedback. Patients received two trials on each set of drawings in each modality. As above the order of presentation and position of the correct picture were randomized. English Sentence Comprehension Prior to initiating C-VIC training and after C-VIC training was completed, patients’ abili- ties to comprehend syntax in English sentences was evaluated with a battery of 48 sentences developed by Berndt and colleagues (Berndt et al., in preparation). The test included 12 non-reversible active sentences, 12 reversible active sentences, 12 non-reversible passive 228 WEINRICH, MCCALL, AND WEBER sentences and 12 reversible passive sentences. Each sentence was read aloud to patients and they were required to perform a forced choice between a picture which corresponded with the sentence and a distractor (a ‘‘reversal’’ of the sentence for reversible sentences and a semantically related noun or verb substitution for the non-reversible sentences). RESULTS Patient DP Comprehension of S-V-O sentences. DP’s auditory comprehension of S-V-O sentences was significantly different from chance (z = 3.03, p< .005, binomial), but still quite poor. He interpreted order correctly in 26/ 34 trials (chance performance = 17). However, his comprehension of symbol order in S-V-O sentences using C-VIC was nearly perfect (33/34) and significantly better than his auditory comprehension of these sen- tences (z = 3.16, p < .002, test for two proportions). Figure 3 illustrates his comparative performance in comprehending symbol order in locative prepositional phrases and S-V-O sentences for stimuli presented in English and C-VIC. DP’s comprehension of symbol order in C-VIC S-V-O sentences was significantly better (z = 3.16, p < .002) than his comprehension of symbol order in C-VIC locative preposi- tional phrases. (Data on locative prepositions is from Weinrich, et al., 1993a. An abbreviated description of the methods used in this experiment are available in Appendix.) Comprehension of S-V-O sentences with conjoined agents and themes. Comprehension of S-V-O sentences with conjoined agents and themes was very difficult for DP when these sentences were presented auditorily. 0.95 0.9 0.85 0.87 0.75 3 English Mil c-vic 0.7 + Proportion Correct 0.65 7 0.6 7 0.55 7 0.5 Locatives $-V-0 Conjoined Fic. 3. Patient DP’s comprehension of symbol order in C-VIC and English constructions. Note that use of C-VIC raises his comprehension of symbol order in S-V-O and conjoined sentences to nearly perfect levels. DP’s comprehension of symbol order in locative preposi- tional phrases in English is nearly at chance (.5) and is not significantly improved by use of C-VIC. ASSOCIATIONS AND DISSOCIATIONS 229 English 2 8 °7 Ml c-vic 0.5 -— — Locatives S-V-0 Fic. 4. Patient DP’s production of symbol order in S-V-O sentences and locative phrases. He was unable to produce these constructions verbally; indeed, he was unable to produce even single words for most trials. DP chose the correct drawing in response to only 12/20 such sentences and required two repetitions for the majority of trials. As illustrated in Fig. 3, the same sentences in C-VIC posed no great challenge for him—he was able to interpret sentence order correctly in 19/20 trials. Thus, he was able to correctly map thematic roles onto S-V-O structures con- taining more arguments than locative prepositional phrases in a task which eliminated a simple left-to-right strategy for mapping roles. Production of reversible S-V-O sentences. DP was unable to produce S-V-O sentences verbally; however, his performance in producing revers- ible S-V-O sentences using C-VIC was quite good. There was no signifi- cant difference between his performance with and without a template cue indicating the sentence structure. He produced 32/36 verbs correctly (3 errors in targets containing ‘‘lift’’ and one error in target containing ‘*give’’), and he made four errors in the 32 trials in which he produced the correct verbs. DP was also unable to produce any locative prepositional phrases verbally. His C-VIC production of order in locatives and S-V-O sentences is illustrated in Fig. 4. There was no significant difference in his error rates in producing prepositions vs. verbs in C-VIC, nor in his error rates in producing symbol order in C-VIC locative prepositional phrases vs. S-V-O constructions. Patient MF Comprehension of S-V-O sentences. MF comprehended symbol order in 33/34 of C-VIC S-V-O sentences, which is significantly better than he comprehended word order in verbally presented ones (26/34, z = 3.16, p < .002). As illustrated in Fig. 5, there was no significant difference 230 WEINRICH, MCCALL, AND WEBER 1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50 Bi english Ml c-vic Proportion Correct Locatives S-V-0 Conjoined Fic. 5. Patient MF’s comprehension of symbol order in C-VIC and English constructions. Note that while his performance in interpreting symbol order is uniformly better in C-VIC than in English for all construction types, his performance across these types in C-VIC is quite uniform and nearly perfect. between his comprehension of symbol order in verbally presented prepo- sitional phrases and verbally presented S-V-O sentences, nor between his comprehension of symbol order in C-VIC S-V-O sentences and C-VIC locative prepositional phrases. In other words, MF performed compara- bly on these two tasks within a communication domain (verbal or visual/ iconographic), although his performance on the same task across the two domains was different. Comprehension of S-V-O sentences with conjoined agents and themes. MF exhibited only mild difficulties in comprehension of verbally pre- sented conjoined sentences. He chose the correct picture in 18/20 trials. His C-VIC performance was flawless—20/20 trials (Fig. 5). Production of reversible S-V-O sentences. After C-VIC training, MF’s verbal production improved remarkably (Weinrich et al., 1993b), so that he can now produce sufficient English S-V-O constructions for analysis. MF could retrieve the correct English verb on 2/3 of trials. In the 24/36 sentences for which he retrieved the correct verb, he made two order errors. He retrieved the correct C-VIC verb in 34/36 trials, but made four order errors on these sentences. His performance on these tasks is illustrated in Fig. 6. Comparison of patients DP & MF across tasks in C-VIC. Data from the C-VIC comprehension tasks on S-V-O sentences and locatives are compared in Fig. 7. The difference between DP’s performance on these tasks is significant at the p < .002 level, while MF performs identically (and nearly perfectly) in interpreting symbol order in these tasks. Although the difference between comprehension and production of symbol order in C-VIC S-V-O sentences was not significant for the pa- ASSOCIATIONS AND DISSOCIATIONS 231 38 Enotish i cvic Fic. 6. Patient MF's production of symbol order in locatives and S-V-O sentences. There is no Statistically significant difference in his production of symbol order in C-VIC or English. e 2 °e Y o& 68 + Proportion Correct S o 0.5 Locatives S-V-0 tients individually, this difference was significant for their pooled data (z = 2.24, p < .05) and is illustrated in Fig. 8. English sentence comprehension, Both DP and MF interpreted revers- ible sentences at chance while performing significantly better on non- reversible sentences (Weinrich, et al., 1993a). There was no significant difference between the performance of either patient on passive vs. active sentences. Neither patient demonstrated any significant change in com- prehension of any sentence type after C-VIC training. 3S 2 8 e 38 Locatives g 5 Bsv.o a 2 a Fic. 7, Comparison of patients DP and MF in comprehension of order in C-VIC locatives and S-V-O sentences. Note that MF performs identically in both tasks while there is a significant (p < .002) difference between DP’s comprehension or order in prepositional phrases and S-V-O sentences. 232 WEINRICH, MCCALL, AND WEBER 0.9 0.8 a7 + Proportion Correct 0.6 0.5 - Comprehension Production Fic. 8. Pooled data from MF and DP comparing production vs. comprehension of symbol order in C-VIC S-V-O sentences. There is a significantly (p < .05) better appreciation of symbol order in comprehension than in production. DISCUSSION These data suggest that patient MF had a mapping deficit prior to C-VIC training which has now been at least partially ameliorated. On the other hand, patient DP continues to have difficulty in mapping thematic roles onto locative prepositional phrases but can map thematic roles onto S-V-O sentences. The data from C-VIC comprehension and production tasks are crucial to this claim. Without these data, we might have con- cluded from the auditory comprehension data that both patients have persistent and global mapping deficits. Both patients have excellent comprehension of symbol order in revers- ible S-V-O C-VIC constructions. This level of comprehension was re- tained in the conjoined agents and patients task, in which the simple heuristic of assigning the first noun as agent will not assure successful completion of the task. Both patients demonstrated poorer performance on the same tasks when they had to process verbally presented sentences. Thus, performance in auditory comprehension of sentences may not re- flect patients’ abilities to map thematic roles onto constituent nouns in these sentences. Deficits in auditory processing and limitations of verbal memory stores may be limiting these patients’ performance. (See Wein- rich et al., 1993a, for a discussion of comparative processes in sentence production in C-VIC and English.) Both patients demonstrated even greater difficulty interpreting reversible sentences in the context of the sentence assessment developed by Berndt and colleagues, in which active and passive voice sentences are presented verbally. Rather than implying a mapping deficit, we interpret these data as indicating that the additional ASSOCIATIONS AND DISSOCIATIONS 233 load of interpreting different sentence structures signaled by morpho- syntactic elements strained systems already burdened by auditory pro- cessing and memory limitations. Prior to C-VIC training on prepositions, MF did indeed appear to have profound difficulties in mapping nouns onto locative prepositional phrases in either English or C-VIC. He was unable to interpret symbol order in locative prepositional phrases in either domain, and his produc- tion showed similar confusion. After training, he no longer exhibited any such difficulties in C-VIC, although he continued to make errors in audi- tory comprehension and verbal production of locative prepositional phrases, albeit at a considerably lower rate than that prior to C-VIC training. In contrast, DP continued to have difficulty in comprehending and producing correct symbol order in C-VIC locative prepositional phrases. DP thus demonstrates a clear dissociation between comprehen- sion of order in C-VIC locative prepositional phrases and reversible S-V-O sentences on the same tasks in which patient MF demonstrates no such dissociation. Note that both patients received comparable train- ing on verbs, while DP received more preposition training than MF. The tasks used to test comprehension of S-V-O sentences and preposi- tional phrases were different. Prepositions were tested using the manipu- lation of real objects, while S-V-O sentences were tested with sets of pictures and foils. DP has no difficulty manipulating objects, so that the task used to test comprehension of prepositions did not place a significant motor load on him and did not require the abstraction of objects and actions from static pictures, as did the S-V-O comprehension task. If anything, therefore, the preposition task might have been less demanding. We have also tested DP in a preposition task requiring a forced choice between two pictures and found that he continues to make errors inter- preting symbol order. DP’s behavior during training further suggests that the above results are not an artifact of different task demands. He was able to produce order in locative prepositional C-VIC phrases without error when he was allowed to observe the therapist placing the objects into position. Indeed, when trying to produce the correct phrase, he would often simulate plac- ing the objects into position. After training was completed DP continued to produce order errors in more complex sentences containing preposi- tions, e.g., ‘‘The man touches the knife in the bowl.’ However, his errors involve confusion of ordering nouns around the preposition. He does not make errors confusing agent and patient in S-V-O sentences. Moreover, DP rarely confuses noun order in dative sentences. How can we explain DP’s difficulties ordering nouns correctly around prepositions while he has such good performance in ordering nouns around verbs? Certainly, his excellent performance on conjoined sen- tences and on datives vitiates any hypotheses based on complexity of 234 WEINRICH, MCCALL, AND WEBER sentence or argument structure. DP’s flawless comprehension and pro- duction of the preposition symbols also eliminates the possibility that he simply did not understand their meaning. Appreciation of the meaning of prepositional symbols appears to require the appreciation of similarities between different situations (Talmy, 1983; Barwise & Perry, 1983). Schwartz et al. (1980) argued that locative prepositions are in the cate- gory of verbs and that the processes of assigning thematic roles to nouns in locative prepositional phrases and S-V-O sentences are identical. In fact, review of their data on the comprehension of locative prepositional phrases and S-V-O sentences by five agrammatic patients reveals signifi- cant dissociations between patient performance on these tasks. (We note, in passing, that the agrammatic patients in our study, such as MF, to date do not demonstrate dissociations between their performance on these tasks using C-VIC). The present data lend further support to the data of Saffran et al. in that DP’s difficulties in comprehending order in locative prepositional phrases remain when he uses C-VIC even though C-VIC requires no morphosyntactic processing and minimizes memory load. These data, however, do not support their assumption that mapping of thematic roles around locative prepositions and around verbs are equivalent, but rather point in the opposite direction. Locative prepositions are closed class elements which play a crucial role in the psychological structure of space (Talmy, 1983). To assign word order in locative prepositional phrases is equivalent to discriminating fig- ure from ground. Thus, DP’s unerring performance when allowed to watch the therapist move the objects into position may have derived from his observation of which object was salient and which was reference. The task employed for testing comprehension of locative prepositions in this series of experiments provided no natural cues as to reference or saliency as might be encountered in normal circumstances (e.g., “‘glass in cupboard’’). In contrast, there is no comparable difficulty in deciding the agent in some reversible S-V-O sentences, e.g., ““The man shoots the woman.’’ It is the figure holding the gun. Saliency is determined by the depiction of the action. In reviewing the developmental literature on acquisition of preposi- tions, Miller and Johnson-Laird (1976) conclude that . .. children learn to use expressions describing paths and directions on the basis of relativistic perceptual information; the relatum is given by external features of the environment. This dependence on external context can be overcome by internalizing vertical and horizontal reference directions and learning to rely on them when external cues are missing or in conflict with internal cues. The domain-independent difficulties observed in aphasic patients, such as DP, in trying to assign thematic roles to nouns in locative prepositional phrases may derive from a deficit in using these internal cues. ASSOCIATIONS AND DISSOCIATIONS 235 This problem appears to be distinct from the problem of assigning the- matic roles around verbs in S-V-O sentences, as illustrated by the dissoci- ation in DP’s performance in these tasks. DP does have difficulty in distinguishing between directional verbs, such as ‘“‘push’’ and “‘pull.”’ Jones (1984) observed that aphasic patients appear to have particular difficulty with directional verbs and ascribed these difficulties to preposi- tional content inherent in the semantics of these verbs. Thus, she would propose that ‘‘pull’’ contains ‘‘towards’’ and ‘‘push’* contains‘‘away.”’ Jackendoff (1983) also classifies such verbs as path verbs. But, as noted above, DP had no difficulty with prepositions when the path was explicit and had no difficulty in comprehending symbol order in C-VIC S-V-O sentences, even those in the present study containing directional verbs. The present data therefore suggest that mapping procedures for different form classes appear to be distinct and are more likely to be lexically- based rather than global procedures. Moreover, the persistent difficulties exhibited by both patients in producing, as opposed to comprehending, reversible S-V-O sentences provide evidence that the mapping proce- dures for comprehension and production of such sentences may be at least partially distinct. The singular contribution of the data obtained from these patients using C-VIC is that some differences in performance remain despite the reduc- tions in processing load (memory, lexical access, morpho-syntactic pro- cessing) afforded by C-VIC. Thus, aphasic patients such as those de- scribed in this study may be able to attend to symbol order in interpreting C-VIC S-V-O constructions because the use of the computer interface supports deficient components of their language comprehension. In con- trast, the problems associated with production of the same constructions, i.e., assembling symbols into linear orders, are only partially ameliorated. DP, in particular, appears to suffer from a conceptual deficit in matching internal cues regarding spatial location with the external environment. Such domain-independent deficits may represent fundamental contraints for some aphasic patients’ language capacities, around which alternative communication systems need to be designed. APPENDIX Comprehension and Production of Locative Prepositional Phrases Methods for testing comprehension and production of reversible loca- tive prepositional phrases in English and C-VIC have been reported pre- viously (Weinrich et al., 1993). Briefly, 15 test phrases were constructed, 6 using ‘‘in,’’ 6 using ‘‘on,”’ and three using ‘‘by.’’ Phrases were con- structed by assigning nouns randomly into the first or second position in the phrases. Nouns did not appear twice in the same phrase, and test phrases were arranged into randomly ordered lists for presentation. To 236 WEINRICH, MCCALL, AND WEBER test comprehension, patients were given C-VIC or English phrases and required to arrange two objects presented in response. (Two differently sized versions of each object were provided to eliminate logical cues.) No time limits were imposed. If the patient failed to respond to the En- glish phrase, it was repeated once, but no cues were provided. To test production, the therapist presented the patients with two ob- jects arranged in space and presented the question ‘“‘where?”’ either in C-VIC or verbally. The therapist arranged the two objects out of view of the patients so that they could not observe them moving into position. Patients were required to produce either a C-VIC or verbal description of the location of the objects. Patients were tested in separate sessions on either C-VIC or English. Verbal responses were recorded on tape for off-line analysis. Production and comprehension were tested separately, with production tested prior to comprehension. Patient responses were scored first as to whether the preposition was correct and then as to whether the order of symbols or objects was cor- rect. If “‘by’’ was chosen either correctly or incorrectly, order was not scored, since symbol order is irrelevant for this preposition. Order was not scored for responses in which the patient chose the wrong preposition or wrong objects. Patients did not produce order errors of the form NNP or PNN. REFERENCES Barwise, J., and Perry, J. 1983. Situations and attitudes. Cambridge: MIT Press. Berndt, R. S. 1991. Sentence processing in aphasia. In M. T. Sarno (Ed.), Acquired aphasia, 2nd ed. New York: Academic Press. Berndt, R. S., & Caramazza, A. 1980. A redefinition of the syndrome of Broca's aphasia: Implications for a neuropsychological model of language. Applied Psycholinguistics, 1, 225-278. Berndt, R. S., Salasoo, A., Mitchum, C. C., & Blumstein, S. E, 1988. The role of intonation cues in aphasic patients’ performance of the grammaticality judgement task. Brain and Language, 34, 65-97. Bock, J. K. 1987. Co-ordinating words and syntax in speech plans. In A. W. Ellis (Ed.), Progress in the psychology of language. London: Erlbaum. Byng, S. 1988. Sentence processing deficits: Theory and therapy. Cognitive Neuropsychol- ogy, 5, 629-676. Caramazza, A., & Miceli, G. 1991. Selective impairment of thematic role assignment in sentence processing. Brain and Language, 41, 402-436. Garrett, M. 1992. Disorders of lexical selection. Cognition, 42, 143-180. Goodglass, H., & Kalpan, E. 1983. The Assessment of Aphasia and Related Disorders. Philadelphia: Lea and Febiger. Grodzinsky, Y., Swinney, D., & Zurif, E. 1985. Agrammatism: Structural deficits and antecedent disruptions. In M.-L. Kean (Ed.), Agrammatism. Orlando, FL: Academic Press. Heaton, R. K. 1981. Wisconsin card sorting test. Odessa, FL: Psychological Assessment Resources, Inc. Jackendoff, R. S. 1983. Semantics and cognition. Cambridge: MIT Press. ASSOCIATIONS AND DISSOCIATIONS 237 Jones, E. V. 1984. Word order processing in aphasia: Effect of verb semantics. In F. C. Rose (Ed.), Advances in neurology, Vol. 42: Progress in aphasiology. New York: Raven Press. Jones, E. V. 1986. Building the foundations for sentence production in a non-fluent aphasic. British Journal of Communications, 21, 63-82. Levelt, W. J. M. 1989. Speaking. Cambridge, MA: MIT Press. Levelt, W. J. M. 1992. Accessing words in speech production. Cognition, 42, 1-22. Linebarger, M., Schwartz, M. M., & Saffran, E. M. 1983. Sensitivity to grammatical struc- ture in so-called agrammatic aphasics. Cognition, 13, 361-392. Miller, G. A., & Johnson-Laird, P. N. 1976. Language and perception. Cambridge: Harvard University Press. P. 409. Mitchum, C. C., Haendiges, A. N., & Berndt, R. S. 1995. Treatment of thematic mapping in sentence comprehension: Implications for normal processing. Cognitive Neuropsy- chology, in press. Miyake, A., Carpenter, P. A., & Just, M. A. 1993. Normal adults’ syntactic comprehension under strong temporal constraints: Implications for theories of syntactic comprehension disorders. Academy of Aphasia Abstracts. Raven, J. C. 1976. Raven's colored progressive matrices A Ag B. London: H. K. Lewis. Saffran, E. M. 1982. Neuropsychological approaches to the study of language. British Jour- nal of Psychology, 73, 317-337. Saffran, E. M., & Schwartz, M. F. 1988. ‘Agrammatic comprehension it’s not: Alternatives and implications. Aphastology, 2, 389-394. Saffran, E. M., Schwartz, M. F., Fink, R., Myers, J., & Martin, N. 1992. Mapping therapy: An approach to remediating agrammatic sentence comprehension and production. In J. Cooper (Ed.), Aphasia treatment: Current approaches and research opportunities. NIDCD Monograph Volume 2. Bethesda, MD: NIH Publications. Schwartz, M. F., Saffran, E. M., & Marin, O. S. 1980. The word order problem in agramma- tism: Comprehension. Brain and Language, 10, 249-262. Shankweiler, D., Crain, S., Gorrell, P., & Tuller, B. 1989. Reception of language in Broca’s aphasia. Language and Cognitive Processes, 4, 1-33. Steele, R. D., Weinrich, M., Wertz, R. T.. Kleczewska, M. K., & Carlson, G. S. 1989. Computer-based visual communication in aphasia. Neuropsychologia, 27, 409-426. Talmy, L. 1983. How language structures space. In H. Pick and L. Acredolo (Eds.), Spatial orientation: Theory, research and application. New York: Plenum Press. Weinrich, M. 1991. Computerized visual communication as an alternative communication system and therapeutic tool. Journal of Neurolinguistics, 6, 159-176. Weinrich, M., McCall, D., Shoosmith, L., Thomas, K., Katzenberger, & Weber, C. 1993a. Locative prepositional phrases in severe aphasia. Brain and Language, 45, 21-45. Weinrich, M., McCall, D., Weber, C., Shoosmith, L., Thomas, K., & Katzenberger, K. 1993b. Training on alternative communication systems can improve aphasic patients’ verbal output. Annals of Neurology, 34, 280.