APHASIOLOGY, 2001, 15 (5), 419–441
Comprehension of reversible active and passive sentences in agrammatism Claudio Luzzatti University of Milan-Bicocca, Italy
Alessio Toraldo International School for Advanced Studies, Trieste, Italy
Maria Teresa Guasti University of Milan-Bicocca, Italy
Graziella Ghirardi G. Salvini General Hospital, Rho, Italy
Lorena Lorenzi S. Maugeri Foundation, Veruno, Italy
Caterina Guarnaschelli S. Maugeri Foundation, Montescano, Italy Agrammatism is a language disorder characterised by a morphological and/or syntactic deficit in spontaneous speech. Such deficits are usually associated with comprehension disorders—though it is said that this is not always the case—which result in a certain degree of variability in syntactic, lexical, and morpholexical performance. The purpose of this study is to reconsider the nature of comprehension disorders in agrammatism, to test whether Grodzinsky’s Trace Deletion Hypothesis (TDH) can be generalised to all agrammatic patients, and to ascertain whether the pattern of impairment observed in agrammatism differs from that present in fluent aphasic patients. Eleven agrammatic patients were tested by means of a sentence comprehension task comprising simple active and passive reversible sentences. The performance of the agrammatic patients was compared to that of 16 fluent aphasic (10 Wernicke’s and 6 conduction) and 10 control subjects. The deficits observed in the agrammatic subjects were compatible with the TDH, but there was also impaired processing of pronouns (elements that are also subject to movement) and a mild deficit on the processing of simple active sentences. The fluent aphasic patients showed Address correspondence to: Claudio Luzzatti, Department of Psychology, Bld U6, University of MilanBicocca, Piazza dell’Ateneo Nuovo 1, I-20126 Milan, Italy, Email:
[email protected] This paper was supported by a grant from the Italian Consiglio Nazionale delle Ricerche (CNR), from the Italian Minstero dell’UniversitaÁ e della Ricerca Scientifica (MURST) and from the Alexander von-HumboldtStiftung. We are indebted to Ria De Bleser who collaborated in preparing the experimental material, to Roelien Bastiaanse for her suggestions and helpful comments on a previous version of the paper, and to Sara Mondini for supplying the data from patient M.B. Preliminary results were presented at the Meeting of the Socie te de Neuropsichologie de Langue FrancË aise, Journe es Franco-Italiennes, Carqueiranne, November 4–5, 1994, and at the 10th Annual Meeting of Theoretical and Experimental Neuropsychology (TENNET), Montreal, June 17–19, 1999. # 2001 Psychology Press Ltd http://www.tandf.co.uk/journals/pp/02687038.html
DOI:10.1080/02687040143000005
420
LUZZATTI ET AL.
a similar pattern of impairment. A logistic regression analysis was then applied to each single case separately, in order to study the homogeneity of the patients’ performance within each aphasic subgroup. Of the 11 agrammatic patients, 3 did not show comprehension disorders, 5 had a specific deficit for passive movement, 1 a lexical deficit for pronouns only, and 1 a pattern of impairment compatible with Linebarger et al.’s trade-off theory. The last patient showed a deficit for simple active reversible sentences compatible with damage to the mapping of grammatical functions to thematic roles. Similar patterns of impairment were also found in the fluent aphasic sample. Overall, the results lead to the conclusion that the TDH cannot be generalised to all agrammatic patients, that the mechanism it invokes is not the only source responsible for agrammatic comprehension disorders and also contributes to comprehension disorders in fluent aphasic patients.
INTRODUCTION The term agrammatism was introduced at the beginning of the 20th century to designate an aphasic disorder of verbal production with non-fluent spontaneous speech characterised by the omission of function words, i.e., articles, pronouns, auxiliaries, and prepositions. Verbs are often in the non-finite form and the remaining inflections are substituted by less marked endings and, in some languages, omitted altogether. In general, the production disorder co-occurs with an analogous comprehension deficit. Different hypotheses have been formulated to explain the symptomatology of agrammatism. The most important are: 1. Non-linguistic accounts of agrammatism. Some authors considered agrammatism as a non-linguistic deficit, i.e., a disorder in which the syntactic, morpho-syntactic, and lexical-semantic competence is supposed to remain intact. Isserlin (1922) considered that agrammatic behaviour is an adaptation to an output disorder following economy of effort. This hypothesis was repeatedly reconsidered until the end of the 1960s (for instance, Lenneberg, 1967) and more recently by Kolk & Heeschen, (1992). Notwithstanding its simplicity, this explanation would appear to be plausible for subjects whose agrammatic spontaneous speech is not associated with other morphological, syntactic, or lexical deficits. Linebarger, Schwartz, and Saffran (1983) suggested that agrammatism could be traced back to a general reduction of processing capacity (the trade-off theory) as a result of which subjects are not able to perform syntactic and semantic processing simultaneously. Syntax, in itself unimpaired, is deactivated because of the processing overload. A more recent explanation suggests that comprehension disorders are a result of the limitation of a specific short-term memory buffer which is selectively involved in syntactic processing (e.g., Caplan & Hildebrandt, 1988; Caplan & Waters, 1995; Just & Carpenter, 1992). 2. Linguistic but non-syntactic hypotheses. The agrammatic comprehension and production deficit has also been explained as a processing disorder of closed-class elements resulting from primary phonological (Kean, 1977) or lexical (Bradley, Garrett, & Zurif, 1980) damage. According to another hypothesis, the comprehension disorder could occur at the lexical level involving the thematic grid associated to verbs (see Grimshaw, 1990; Jackendoff, 1990 for a theoretical description of thematic notions), or at a level where thematic roles are assigned to arguments (Saffran, Schwartz, & Marin, 1980; Schwartz, Saffran, & Marin, 1980). 3. Agrammatism as a ‘‘global’’ syntactic disorder (asyntaxia). This hypothesis asserts that a single central grammatical processor is involved in both production and
SENTENCE COMPREHENSION IN AGRAMMATISM
421
comprehension; if this processor was impaired, an asyntactic analysis of constituents would be induced (Berndt & Caramazza, 1980; Caplan & Futter, 1986; Caramazza & Zurif, 1976). More recently a similar explanation has been proposed in terms of a general loss of functional categories (Ouhalla, 1993). However, a global asyntactic account is in contrast with the preserved ability of agrammatic patients to perform grammaticality judgements (Linebarger et al., 1983) and to perform tasks that require a preservation of functional categories (and therefore a hierarchical representation of the sentences) (e.g., Lonzi & Luzzatti, 1993). 4. Agrammatism as a selective damage within syntax. More recent studies from linguistic aphasiology have explained agrammatism within the framework of the Government and Binding Theory (Chomsky, 1981), according to which syntax contains several representational levels and subsystems (linguistic modules). Damage to any of these syntactic units could determine the outcome of specific and heterogeneous syntactic disorders. Grodzinsky’s Trace Deletion Hypothesis (TDH) states that agrammatism results from damage to a specific mechanism connecting the antecedent to its trace (Grodzinsky, 1986, 1995). Verbs assign thematic roles to elements in structurally defined positions. In some cases, however, due to the operation of other grammatical principles, a noun phrase (NP) normally occupying a thematic position moves to a non-thematic position, leaving an abstract marker (a trace) in the original position. This movement results in a chain, connecting the moved NP with its trace. For instance, the object of the active sentence [1] moves to the subject position in the corresponding passive sentence [2], leaving a trace (t), whereas the subject of the original active sentence is expressed through a by-phrase. Thus, in sentence [2] the NP ‘‘the girl’’ and the t form a chain. [1] the boy kisses the girl [2] the girli is kissed ti by the boy
Several studies on agrammatism have shown that comprehension of reversible active sentences may be spared, whereas comprehension of passive sentences may be poor (Caplan & Futter, 1986; Goodglass, 1968; Grodzinsky, Finkelstein, Nicol, & Zurif, 1988; Schwartz, Saffran, Marin, 1980). Grodzinsky (1986, 1995) suggested that this pattern of impairment may be explained by the fact that the relation between a trace and its antecedent has become inaccessible as the traces have been deleted. According to this view, thematic role assignment is unimpaired in active sentences because NPs remain in the thematic position. On the other hand, NPs that have moved—e.g., the subject of a passive sentence—are assigned a thematic role through the application of a default interpretative strategy that attributes the Agent role to an NP in the subject position. However, the NP argument of the by-phrase (the boy) is also assigned the thematic role of Agent by the preposition. As a consequence, an agrammatic patient who is asked to interpret the sentence is faced with two possible Agents and is forced to guess which NP should be assigned the Agent role and which the Patient role. Hickok, Zurif, and Conseco-Gonzales (1993) have revised Grodzinsky’s theory by incorporating the Verb-Phrase-Internal-Subject Hypothesis (see Koopman & Sportiche, 1991; McNally, 1992, among others, for a discussion of this view). According to this hypothesis, the subject, as the other arguments of the verb, is generated inside the Verb Phrase (VP) and moves to its surface position in the Inflectional Phrase (IP). Thus, an active sentence such as [1] or [3] is similar to a passive sentence in that it also includes the presence of a trace connecting the subject to its thematic position inside the VP [4].
422
LUZZATTI ET AL.
[3] Mary may invite Paul [4] [IP Maryi [I may [VP ti invite Paul]]]
In order to account for the pattern of impaired and of spared comprehension observed in agrammatic patients, Hickok et al. (1993) retain Grodzinsky’s idea that traces are deleted and assume that verbs are associated to a thematic assignment representation (TAR). Agrammatic patients can assign theta roles to NPs in their base position, but they cannot assign them through traces, as these are deleted. Therefore, the Patient thematic role is still assigned after deletion of traces, as the Object NP has not moved. Furthermore, the only NP left is assigned the remaining thematic role (i.e., the Agent) by default. By contrast, in a passive sentence neither the Subject nor the Object are in their base position; thus no theta role can be assigned, and the agrammatic patient is left to guess which NP has to receive the Agent role and which the Patient role. Both the TDH and the revised TDH hold that traces are deleted and that therefore agrammatic patients have difficulty in forming chains. Further investigation of the comprehension abilities of agrammatic patients (Hickok & Avrutin, 1996) has shown that not all the traces are deleted. It has been shown, for example, that agrammatic comprehension of which-questions can be unequally impaired. Subject which-questions (which boy chased the dog?) are understood better than Object which-questions (which boy did the dog chase?). On the other hand, Subject and Object who-questions (who chased the dog?/who did the dog chase?) are understood equally well. Hickok and Avrutin (1996) interpreted the different performance on which- versus who-questions as a sign that not all chains are disrupted in agrammatic comprehension. In fact, while whichquestions involve the formation of binding chains, who-questions require the formation of government chains (Cinque, 1990; Rizzi, 1990). Hickok and Avrutin (1996) argued that agrammatic comprehension affects the formation of binding chains, and not of government chains. This claim is consistent with Lonzi and Luzzatti’s (1993) findings that Italian agrammatic patients do not have problems with verb movement, a syntactic operation that also results in the formation of government chains.
AIM OF THE STUDY The aim of the present study is to reconsider the nature of sentence comprehension disorders in agrammatism and in aphasia in general. As the majority of the previous studies were conducted on single cases or on very small groups of patients, it was decided to form a larger sample of patients on which to address the following issues: 1. Are comprehension disorders in agrammatism specifically related to the deletion of traces, and to what extent can lexical impairments (either of the thematic grid of verbs, or of the processing of function words) parallel a specific syntactic disorder? 2. In addition to the classical passive reversible sentence structure, Italian also offers a further possibility to test the impairment of the relation between a moved element and its trace. While a full NP complement follows the verb, a non-tonic (clitic) pronoun is found in the preverbal position (see [5], [6]). It is assumed that the occurrence of a clitic pronoun in the preverbal position is the result of movement from the position usually occupied by a full NP (or PP) (Cecchetto, 2000; Kayne, 1989; Sportiche, 1995). [5a] [5b] [6a] [6b]
Mario cerca Flora, Mario seeks Flora; Mario lai cerca ti, Mario seeks her (lit. Mario her seeks) Mario daÁ un regalo a Flora, Mario gives a present to Flora; Mario lei daÁ un regalo ti, Mario gives her a present (lit. Mario her gives a present).
SENTENCE COMPREHENSION IN AGRAMMATISM
423
In a sentence with a clitic pronoun the canonical SVO order has been disrupted and both the S(ubject) and the O(bject) are in preverbal positions. It is thus possible that agrammatic patients have difficulty in assigning the Agent thematic role, as there is another candidate in the preverbal position, although the subject still precedes the direct object. However, the agrammatic patient could adopt a strategy that consists in assigning the Agent role to the first element of the sentence, and the Patient/Goal role to the second element, whatever its position is (pre- or postverbal). We should point out that in sentences with an indirect object clitic pronoun, the direct object (e.g., ‘‘a present’’ in example [6b]) is inanimate and thus there is no problem of reversibility between the Theme role (assigned to the direct object) and the Goal role (assigned to the indirect object). According to the TDH, patients may be unable to assign thematic roles because both the subject and the object (clitic) are in the preverbal position. Based on the revised version of the TDH proposed by Hickok et al. (1993) agrammatic patients would not be able to assign the theta roles because both the Subject and the Complement are displaced and theta assignment requires the connection of these elements to their traces. 3. Are the comprehension deficits displayed by agrammatic patients specific to this aphasic subgroup, or do non-agrammatic patients suffer from a similar pattern of impairment? According to Grodzinsky’s claim that all agrammatic—and only agrammatic—patients show a comprehension disorder resulting from damage to a specific syntactic ability and/or a specific unit of processing, the comprehension disorders shown by fluent aphasic patients with a non-agrammatic morpho-syntactic impairment should differ, at least in some instances, from that observed in agrammatic patients. The pattern of performance in agrammatic patients was therefore compared to that of fluent aphasic patients. 4. If there is a specific pattern of impairment, can this be generalised to all agrammatic patients? Single profile analyses were performed to test the degree of homogeneity within the agrammatic and fluent aphasic subgroups to investigate this question.
MATERIALS AND METHODS Subjects Eleven aphasic patients with prototypical agrammatic speech output were studied. Their performance was compared to that of 16 mild to moderate fluent aphasic patients (10 Wernicke’s and 6 conduction aphasics). The clinical classification was obtained on the basis of the Italian version of the Aachen Aphasia Test (AAT: Luzzatti, Willmes, & De Bleser, 1996). The two groups of patients were matched for level of education and aphasia severity based on the AAT profile. The study included 10 control subjects (7 females, 3 males) of comparable education. Table 1 summarises the characteristics of the subjects who participated in the experiment.
Comprehension task The study used the standard paradigm for testing syntactic comprehension of semantically reversible sentences, i.e., a sentence-to-picture matching task. The test involved different conditions where Agent and Patient could be either a masculine (and/ or a feminine) proper name or pronoun. In some of the previous studies, the sentences used were quite complex (subject- and object-relative clauses; subject- and object-cleft
424
LUZZATTI ET AL. TABLE 1 Experimental subjects
Subject
Sex
Age
Education (yy)
Aetiology
Aphasia type
AAT-L
A.D. C.M. D.R. E.Be. F.C. L.Z. M.B. M.G. P.Gh. P.Gi. R.O.
F M M M F F M M F M M
20 21 23 16 28 35 35 15 34 38 46
13 14 13 10 13 15 12 8 12 13 8
V,T T V,T V V V V T V V A+S
Agr Agr Agr Agr Agr Agr Agr Agr Agr Agr Agr
45.2 59.7 48.6 52 49.5 50.8 54.8 52.4 49 49.2 43
A.C. E.Bg. E.Bo. E.M. G.D. G.I. L.R. N.M. R.B. V.B.
M F M M M M M M M M
73 70 57 56 64 48 65 55 40 45
5 13 13 5 13 13 11 8 13 17
V V V V V V V V T V
W W W W W W W W W W
48.7 50.3 57.1 47.9 46.4 48.6 49.5 50.3 47.3 56.7
A.G. C.G. M.Ca. M.Cr. S.M. U.B.
M M M F M M
52 45 35 54 53 48
8 18 10 6 17 8
V V T V V V
C C C C C C
44.6 46.2 54.3 54.3 52.9 56.7
Aetiology: V = vascular, T = traumatic, A = abscess, S = surgery; aphasia type: Agr = agrammatic, W = Wernicke’s, C = conduction, AAT-L = overall T-score profile level for the Italian version of the AAT.
clauses) (e.g., Grodzinsky, 1989) and lengthy, thus causing excessive loading of the phonological short-term memory, a function that is damaged in most agrammatic patients. To reduce the effects of sentence length, the present study used active and passive sentences with a maximum of four constituents, and relative and cleft clauses were not included. Active sentences contained an Agent (e.g., Mary), a verb (seeks), and a Patient (John); passive sentences contained the moved Patient, a passive verb, and an Agent (by-phrase). In another group of stimuli, Agent/Patient reversibility resulted between the subject and an indirect object: this set of sentences contained an Agent (e.g., Mary), a verb + an invariant direct object (gives a present) and an indirect object (to John). Each sentence contained a masculine or feminine agent ‘‘acting’’ on a masculine or feminine Patient/Goal. Thus, there were four possible conditions: two ‘‘same gender’’ (SG) (male on male; female on female), and two ‘‘different gender’’ (DG) conditions (male on female; female on male). Each multiple choice set included four pictures representing a male acting on a female, a female acting on a male, a male on a male, and a female on a female (i.e., the target picture and three distractors). When the stimulus was an SG sentence, the task
SENTENCE COMPREHENSION IN AGRAMMATISM
425
could also be solved through the use of purely lexical knowledge, as only one of the four pictures represented two males or two females. DG sentences, on the contrary, required a choice to be made between two pictures, which could only be made through syntactic analysis. The task thus developed followed a factorial design with four orthogonal variables: (i) same vs different gender of Agent and Patient/Goal (SG vs DG sentences); (ii) sentence type (simple active, passive, active with an indirect object); (iii) pronominalisation of the grammatical subject (full-NP vs pronoun); and (iv) pronominalisation of the direct or indirect object (full-NP vs pronoun). In order to be complete, the factorial design should have been composed of 2 6 3 6 2 6 2 = 24 types of sentences; however one of the a priori sets of sentences [passive SG sentences with the same pronoun in both the subject and the object position, e.g., lui eÁ cercato da lui, he is sought by him (lit. ‘‘he is sought by he’’)] was excluded because of the ambiguous solution of the pronominal reference (‘‘he is sought by him’’ vs ‘‘he is sought by himself’’). The corresponding active sentence, on the other hand, was appropriate because the accusative pronoun is different from the subject pronoun and would be different from the reflexive one. The task was therefore composed of 23 6 10 = 230 items (80 transitive active sentences with a direct object, 70 passive sentences, 80 active sentences with an indirect object). Stimuli were randomised for types of sentences. Examples from each subset of the experimental set are given in Table 2. The examiner read aloud a semantically reversible sentence after which the subjects were asked to point to the target picture, making a choice between four alternatives. If the patient so wished, the examiner could repeat the sentence a second time. There was no time limit and self-corrections were allowed.
Statistical methods (1) Agrammatic group sample. Data from the agrammatic group were analysed by means of paired t-tests in order to check the a priori hypotheses (as the factorial design was not complete it was not possible to use a general ANOVA ). (2) Comparison between agrammatic and fluent aphasic patients. The group performance of the 11 agrammatic patients was compared to that of the 10 Wernicke’s and 6 conduction aphasic patients (t-tests for independent samples) for each of the effects previously studied on the agrammatic sample. (3) Analysis of the single case profiles. To study the homogeneity of the patterns of performance within the aphasic subgroups, the profile of each single agrammatic and fluent aphasic patient was analysed by means of a logistic regression procedure (McCullagh & Nelder, 1983), by which the potential effects of different variables on the patients’ performance can be independently estimated. This method permits the introduction of continuous variables, the effects of which could not be estimated in the group analyses (e.g., the length of sentences and the level of difficulty for normal subjects), and the disentanglement of the effects of variables that are usually associated— for instance, passive sentences are also longer, clitics are also pronouns. The analysis was run on DG sentences only in order to exclude sentences where the correct picture could be chosen through the use of a purely lexical strategy. The single items of the comprehension task constituted the units of the analysis; for each item the dependent variable was dichotomous (correct = 1, incorrect = 0). Categorical independent variables were (i) presence vs absence of a passive structure and (ii) presence vs absence of clitic movement. Continuous independent variables were (iii) degree of pronominalisation (full-NP sentence = 0; one constituent is pronominalised= 1; both NPs are pronomina-
426
LUZZATTI ET AL. TABLE 2 Examples of sentences with Agent and Theme of different gender (DG)
Item type
Example
Literal translation
CI
SVO sVO SoV soV
Active reversible sentences Mario cerca Flora Lui cerca Flora Mario la cerca Lui la cerca
Mario seeks Flora He seeks Flora Mario her (clitic accusative) seeks He her (clitic accusative) seeks
0.79 0.71 1.67 2.00
PVA pVA PVa pVa*
Passive reversible sentences Flora eÁ cercata da Mario Lei eÁ cercata da Mario Flora eÁ cercata da lui Lei eÁ cercata da lui
Flora is sought by Mario She is sought by Mario Flora is sought by him She is sought by him
2.24 2.56 2.12 2.78
SVDI sVDI SiVD siVD
Sentences with indirect object Mario daÁ un regalo a Flora Lui daÁ un regalo a Flora Mario le daÁ un regalo Lui le daÁ un regalo
Mario gives a present to Flora He gives a present to Flora Mario her (clitic dative) gives a present He her (clitic dative) gives a present
1.00 1.33 3.06 3.38
Active sentences: S, s = subject; V = verb; O, o = direct object. Passive sentences: P, p = patient; A, a = agent. Sentences with indirect object: I, i = indirect object; D = non-reversible direct object. Upper-case letters designate full-NPs, lower-case letters designate pronouns. CI: Complexity Index (see Statistical methods). * This set of sentences does not have its counterpart of sentences with Agent and Theme of same gender.
lised = 2) and (iv) length of sentences (number of syllables). A further variable (v) difficulty for control subjects (number of errors made by the 10 control subjects in each subset of the task) was introduced to disentangle the effects of the previous four main variables from a possible non-specific effect arising from the general difficulty of the items. The logistic regression analysis (SPSS 6.1.2) was applied to each subject’s data set. As the single case diagnostic procedure was used, the significance level was set at .05 for each effect under study, without adopting an overall protection for the whole sample (Willmes, 1985). A further variable (general complexity index) was introduced in the regression model to test if impaired comprehension may result from a general reduction of processing resources. This reduction would emerge when patients have to accomplish several operations simultaneously, but not when these are executed separately (e.g., Just & Carpenter, 1992; Linebarger et al., 1983). The complexity index (CI) was computed as the sum of the levels of difficulty of each of the five previous variables. All the variables were attributed the same weight, transforming each range into a 0–1 interval (presence of passive movement: active = 0, passive = 1; presence of clitic movement: non-clitic = 0, clitic = 1; degree of pronominalisation: no pronouns = 0, single pronoun = 1/2, two pronouns = 1; sentence length: the number of syllables was rescaled to the 0–1 range assigning 0 to the minimum and 1 to the maximum number of syllables; difficulty level: no errors in the control sample = 0, one error = 1/3, three errors = 1). The CI value resulting for each item type is reported in Table 2.
RESULTS AND DISCUSSION First, the mean performance of the agrammatic patients relative to the different variables of the experimental design was considered, after which a comparison between
SENTENCE COMPREHENSION IN AGRAMMATISM
427
agrammatic and fluent aphasic patients was made (see Table 3). Finally, the homogeneity of the comprehension disorder within the two groups of patients was studied through the single case analysis.
1. Mean performance of the agrammatic sample (1) Comprehension of full-NP active sentences. The comprehension of full-NP active sentences was tested by comparing the performance of the agrammatic patients on DG sentences (e.g., ‘‘John seeks Lucy’’ and ‘‘John gives a present to Lucy’’) where syntactic processing is mandatory for correct interpretation, with that on SG sentences (e.g., ‘‘John seeks Mark’’ and ‘‘John gives a present to Mark’’) where the task can be solved using a purely lexical strategy, as only one of the four pictures represents two males or two females. Agrammatic patients as a group showed a better performance on SG sentences (91.82%) than on the DG counterparts (82.73%) [paired t (10) = 2.82, onetailed p < .01]. (2) Comprehension of active vs passive sentences. The effect of passive movement was isolated by considering only the DG sentences, as these are the only sentence type where syntactic processing is mandatory for a correct interpretation. In particular, active sentences with full-NP or pronominalised subjects and a full-NP object (e.g., ‘‘John seeks Lucy’’ and ‘‘he seeks Lucy’’) were compared to the corresponding passive sentences (e.g., ‘‘Lucy is sought by John’’ and ‘‘Lucy is sought by him’’). Active sentences with two pronouns were excluded because the object moves to the preverbal position, and they are therefore not comparable to their passive counterpart. The patients’ mean performance was significantly better on active sentences [active = 79.55%; passive = 62.73%; paired t (10) = 2.69, one-tailed p < .01; see Figure 1]. A further proof of specific impairment on reversible passive sentences emerges from the fact that the weakness observed in passive sentences disappears in SG sentences (‘‘John seeks Mark’’ and ‘‘John seeks him’’ vs ‘‘Mark is sought by John’’ and ‘‘Mark is sought by him’’), i.e., when the task can be solved by a purely lexical strategy (active = 86.82%; passive = 89.55%; paired t (10) = .8, NS). The significant difference between SG and DG passive sentences [SG = 89.55%; DG = 62.73%; paired t (10) = 3.95, one-tailed p < .01; see Figure 1] indicates that agrammatic patients do in fact use this lexical cue. (3) Agrammatism as an impairment of the processing of function words: Full-NPs vs pronouns in subject position. The objective of this section is to determine whether the comprehension disorder could be, even partially, the consequence of a lexical disorder of function words (i.e., of pronouns). SG and DG sentences with full-NPs in the subject position were compared to their pronominalised counterparts [the full-NP/pronoun comparison in the object position was not carried out, as object pronouns underlie preverbal clitic movement (Mario la bacia, Mario kisses her; lit. ‘‘Mario her kisses’’), which would have introduced a further source of variability]. Pronominalisation was significant in agrammatic patients as a group (full-NP subject = 76.86%; pronominalised subject = 71.57%; paired t (10) = 2.69, one-tailed p = .01). (4) Clitic movement of the pronominalised object. The effect of the clitic movement of object pronouns was investigated in this section. The comparison between DG sentences with a pronoun either in the subject or in the object position, which isolates the effect of the clitic movement, is shown in Figure 2. The statistical analysis showed a
428
mean SD mean SD mean SD mean SD
mean SD mean SD mean SD mean SD
99 3.2 81.8 16.6 69 17.3 75 21.7
100 – 90 14.1 86 13.5 86.7 8.2
100 – 77.3 22.8 62 24.9 71.7 22.3
100 – 83.6 25.4 76 19 71.7 27.1
sVO
100 – 75.5 24.6 60 25.4 40 26.8
95 8.5 80.9 13.8 67 25.8 76.7 13.7
SoV
100 – 70 26.5 50 27.5 51.7 24
98 4.2 60.9 21.1 59 22.8 65 28.8
soV
99 3.2 64.5 28.1 58 21.5 66.7 18.6
99 3.2 91.8 14.7 78 22.5 86.7 10.3
PVA
99 3.2 60 31 55 25.1 53.3 33.9
100 – 82.7 24.5 86 14.3 88.3 14.7
pVA
100 – 60.9 25.1 53 28.7 45 32.1
100 – 87.3 14.9 79 24.7 86.7 24.2
PVa
99 3.2 60 29.3 57 23.1 56.7 20.7
pVa*
100 – 83.6 15 70 30.2 71.7 29.3
100 – 93.6 10.3 79 26.9 93.3 12.1
SVDI
100 – 78.2 22.3 62 38.5 71.7 27.1
100 – 90.9 17.6 78 29 76.7 23.4
sVDI
97 6.7 52.7 32 37 20.6 36.7 29.4
87 17.7 70 29 40 33 55 16.4
SiVD
97 6.7 54.5 27.7 50 27.1 46.7 8.2
90 17.6 69.1 21.7 57 23.1 61.7 11.7
siVD
See Table 2 for abbreviations. SD = standard deviation. * pVa sentences do not have a counterpart set of sentences with Agent and Theme of same gender (SG) sentences.
Agrammatic patients Wernicke’s aphasic patients Conduction aphasic patients
DG sentences Control subjects
Agrammatic patients Wernicke’s aphasic patients Conduction aphasic patients
SG sentences Control subjects
SVO
TABLE 3 Mean performance (% correct) from control subjects, agrammatic patients, Wernicke’s and conduction aphasic patients on SG and DG sentences
SENTENCE COMPREHENSION IN AGRAMMATISM
429
Figure 1. Mean performance of the 11 agrammatic patients on active and passive reversible sentences (SG = same gender, DG = different gender of Agent and Patient/Goal).
significant disadvantage for pronouns in the object vs the subject position for sentences with an indirect clitic object [pronouns in the subject position = 78.18%; clitic pronouns in preverbal position= 52.73%; paired t (10) = 3.48, one-tailed p < .01], but this does not hold true for sentences with a direct clitic object [pronouns in the subject position = 77.27%; object clitic pronouns in preverbal position= 75.45%; paired t (10) = 0.3, NS]. The discrepancy of the effect between sentences with direct and indirect objects was significant [paired t (10) = 3.49, p < .01]. The difference in the performance on sentences with a pronoun in subject vs object position can be interpreted as the result of an impairment in analysing the movement of object pronouns to a preverbal position. There may be two explanations for the fact that sentences with an indirect clitic pronoun showed a significant effect, whereas sentences with direct clitic pronouns showed no effect at all. The first explanation holds that the difference is at the lexical level and results from a different word frequency of clitic direct object pronouns with respect to clitic indirect object pronouns (direct object >> subject >> indirect object) (De Mauro, Mancini, Vedovelli, & Voghera, 1993) [direct object pronouns: masculine = 3219, feminine = 934; subject pronouns: masculine = 635, feminine = 474; indirect object pronouns: masculine = 638, feminine = 172 out of a corpus of 500,000 words normalised to 100,000 occurrences]. The second explanation, which agrees with the TDH, maintains that in sentences with direct object clitic pronouns, agrammatic subjects adopt a strategy consisting in assigning the Agent role to the first argument (the subject) and the Patient role to the second argument (the clitic pronoun). However, this explanation does not address the problem of why agrammatic subjects cannot adopt the same strategy in the case of the indirect object. Using this strategy, in fact, agrammatic patients could assign the Agent role to the first element of the sentence, the Goal role to the second one (the clitic), and the Theme role to the direct object, which is inanimate.
430
LUZZATTI ET AL.
Figure 2. Mean performance of the 11 agrammatic patients on DG reversible active sentences with only one pronoun (S-DO = subject and direct object are reversible; S-IO = subject and indirect object are reversible). Sentences with a pronoun in the subject position (e.g., lui cerca Flora, he seeks Flora: lui daÁ un regalo a Flora, he gives a present to Flora) are compared to sentences with an object clitic pronoun in the preverbal position (e.g., Mario la cerca, Mario seeks her [lit. Mario her seeks]; Mario le daÁ un regalo, Mario gives her a present [lit. Mario her gives a present]).
Discussion. In summary, the comprehension of agrammatic patients was already impaired in simple full-NP active sentences. This is consistent with the hypothesis that comprehension is affected by a disorder of the lexical analysis of the thematic grid, or by damaged mapping of theta roles to arguments (Schwartz et al., 1980). Furthermore, the agrammatic patients showed a mild impairment of the comprehension of pronouns, consistently with the hypotheses formulated by Kean (1977) and by Bradley et al. (1980) regarding a processing disorder of closed-class elements. Alternatively, it is possible that this low performance might be ascribed to the fact that in Italian pronominal subjects are rarely used and only in particular circumstances. With regards to syntactic processing, an impairment both of the passive movement and of the movement of the clitic pronouns to a preverbal position was observed, which is consistent with Grodzinsky’s TDH (1986, 1995). The dissociated impairment of indirect clitic pronouns with respect to the less impaired direct clitic pronouns is evident and can be explained by the word frequency of
SENTENCE COMPREHENSION IN AGRAMMATISM
431
the direct clitic pronouns, which is five times higher than the word frequency of the indirect clitic pronouns, or by the possibility of adopting a strategy for the assignment of theta roles in sentences with direct object clitic pronouns (a strategy that does not seem to be adopted in sentences with indirect object clitic pronouns, however).
2. Comparison of agrammatic and fluent aphasic patients As usual (Basso, Capitani, Laiacona, & Luzzatti, 1980; Obler, Albert, Goodglass, & Benson, 1978), fluent aphasic patients were on average older than agrammatic patients (agrammatic patients = 28.3 ± 10.0; Wernicke’s patients = 57.3 ± 10.8; conduction aphasic patients = 47.8 ± 7.1). Mean educational level and aphasia severity were identical in the two groups, as agrammatic and fluent aphasic patients were matched for those variables (see Table 1). The subjects’ mean performances on the comprehension of DG and SG sentences are shown in Table 3, while Figure 3 shows the mean score profiles on DG sentences for the control subjects, the agrammatic patients and the fluent aphasic patients. Control subjects performed almost flawlessly on all subsets of items. The mean profile of the agrammatic patients shows the effects described in section 1: each value of the passive subprofile is lower than all the values of the active profile; furthermore, the active profile shows a progressive decline starting from full-NP sentences to those in which both subject and object are pronominalised. The decline is even more pronounced for sentences with an indirect object. The profile of the Wernicke’s patients is parallel to but lower than that of the agrammatic patients. The conduction patients’ profile shows a main discrepancy with respect to that of the agrammatic patients, i.e., a poorer performance on sentences with direct clitic pronouns (SoV and soV) with respect to the non-clitic counterparts (SVO and sVO) Table 4 summarises the results of the comparison between agrammatic and fluent aphasic patients for the main effects under study. The advantage in performance on SG with respect to DG active full-NP sentences was not significantly different between the agrammatic and the two fluent aphasic groups. The same was also true for the advantage on DG active with respect to DG passive sentences and for the disadvantage caused by subject pronominalisation. However, it should be noticed that the effect of subject pronominalisation was virtually nil in the fluent aphasic sample, whereas in the agrammatic sample it was certainly present albeit very small. Overall, only the agrammatic patients would appear to be affected by lexical impairment. The comprehension of direct clitic pronouns is impaired in conduction aphasic patients, whereas no significant effect was found in the agrammatic and the Wernicke’s groups. On the contrary, the movement of the indirect clitic object influences the performance of all three groups of patients to a similar extent. Discussion. Overall, although the fluent aphasic patients performed the task at a quantitatively lower level, and the performance pattern was qualitatively similar to that of the agrammatic patients, the only exceptions being a more impaired comprehension of subject pronouns in the agrammatic sample, which seems to be lexical in nature (i.e., impaired analysis of function words), and an impaired comprehension of direct clitic pronouns in the conduction aphasic group, which may be explained by a lower effect of word frequency in this patient sample. The poorer performance of the fluent aphasics compared to the agrammatic patients could be ascribed to a more severe ‘‘mapping’’ deficit, as there was no difference in the overall severity of aphasia between the two groups.
432 p = .39 p < .01
(see above)
Agrammatic patients (n = 11)
t = 2.82 t = 2.69 t = 2.69 t = 0.30 t = 3.48
6.29 6.41 2.57 5.12 11.18
t = 2.07 t = 1.56 t = 0.18 t = 0.39 t = 2.24
p = .08 p = .43 p = .35 p = .03
16.66 17.50 1.67 31.66 35.00
7.38 11.60 3.82 15.58 15.00
t = 2.26 t = 1.51 t = 0.44 t = 2.03 t = 2.33
p = .10 p = .34 p = .05 p = .03
p = .04
Conduction aphasic patients (n = 6) MD SE Paired t (df = 5), p (one-tailed)
13.00 10.00 0.45 2.00 25.00
p = .03
3.22 6.26 1.97 6.15 7.31
p < .01 p = .01 p = .01
9.09 16.82 5.29 1.82 25.45
Wernicke’s aphasic patients (n = 10) MD SE Paired t (df = 9), p (one-tailed)
Agrammatic patients (n = 11) MD SE Paired t (df = 10), p (one-tailed)
p = .58 p = .46 p = .15 p = .98 p = .97
t = 1.10 t = 0.06 t = 0.94 t = 2.13 t = 0.65
p = .29 p = .96 p = .36 p = .05 p = .53
Comparison t (df = 15), p (two-tailed)
t = 0.57 t = 0.76 t = 1.51 t = 0.02 t = 0.03
Comparison t (df = 19), p (two-tailed)
Significant results are reported in bold. The size of each effect is quantified as the difference between the mean performances (% of correct responses) in the two conditions. MD = mean difference; SE = standard errors.
Same gender vs different gender active full-NP sentences Active vs passive sentences Full-NP vs pronoun in subject position Pronoun in subject vs (clitic) direct object position Pronoun in subject vs (clitic) indirect object position
Same gender vs different gender active full-NP sentences Active vs passive sentences Full-NP vs pronoun in subject position Pronoun in subject vs (clitic) direct object position Pronoun in subject vs (clitic) indirect object position
Comparisons between types of sentences
TABLE 4 Mean differences of performance (%) between major types of sentences for agrammatic, Wernicke’s, and conduction aphasic patients, and comparisons between agrammatic and fluent aphasic patients
SENTENCE COMPREHENSION IN AGRAMMATISM
433
Figure 3. Mean profiles of agrammatic patients, fluent aphasic patients, and control subjects on DG sentences. See Table 2 for abbreviations.
3. Single case profiles One of the principal objectives of the study was to analyse the variability in the performances of the individual patients within each aphasic subgroup. Table 5 summarises the results obtained by each single patient and the median performance of the control subjects on the DG sentences. The performances of three agrammatic patients (C.M., E.Be., F.C.) were clearly within the normal range, and they were therefore excluded from the single-case analysis. In addition, the comprehension performance of several agrammatic and fluent aphasic patients was already impaired for full-NP active sentences (see Table 5, column SVO). As already described (see statistical methods, point 3), the single-case analysis was performed using a logistic regression procedure whereby the influence of each single variable on the comprehension performance of a single patient can be studied in isolation from the effects of other variables. For example, the effect of a passive structure can be detected independently of the effect caused by the additional length of the passive sentence. Table 6 summarises the results of the analysis of the agrammatic and fluent aphasic patients with comprehension disorders. The table also shows the results of the linear regression analysis that estimates the effect of the general complexity index CI (i.e., the sum of the difficulty levels of each of the five variables) on the comprehension performance. Agrammatic patients with comprehension disorders. The passive structure significantly influenced the performance of five out of the eight agrammatic patients with a comprehension deficit (L.Z., M.B., P.Gh., P.Gi., R.O.); a clitic pronoun moving to the preverbal position affected the performance of two patients (M.B., R.O.); the degree of pronominalisation and sentence length individually influenced the performance of M.G.
434
LUZZATTI ET AL.
TABLE 5 DG sentences: Results obtained by each single patient and by the control group (median and 5th percentile)
SVO Control subjects Median 10 5th perc. 9 Agrammatism C.M. 10 E.Be. 10 F.C. 10 A.D. 8 D.R. 5 L.Z. 7 M.B. 10 M.G. 7 P.Gh. 8 P.Gi. 7 R.O. 8 Wernicke’s aphasia A.C. 6 E.Bg. 9 E.Bo. 5 E.M. 7 G.D. 7 G.I. 6 L.R. 4 N.M. 7 R.B. 9 V.B. 9 Conduction aphasia A.G. 10 C.G. 5 M.Ca. 9 M.Cr. 5 S.M. 7 U.B. 9
sVO
SoV
soV
PVA
pVA
PVa
pVa
SVDI sVDI SiVD siVD Total
10 10
10 10
10 10
10 9
10 9
10 10
10 9
10 10
10 10
10 8
10 8
120 117
10 10 10 6 3 10 9 6 7 7 7
10 10 10 6 5 8 8 8 9 7 2
10 10 10 7 2 9 8 5 6 5 5
10 10 10 3 7 7 5 6 7 4 2
10 10 10 4 4 6 8 3 5 1 5
10 10 9 5 4 5 5 7 5 4 3
10 9 10 6 6 4 8 2 2 5 4
10 9 10 7 8 8 8 8 10 9 5
10 9 10 5 3 8 7 7 10 9 8
10 10 8 1 2 4 6 6 5 5 1
9 9 8 4 2 7 5 3 7 1 5
119 116 115 62 51 83 87 68 81 64 55
7 8 3 5 4 5 3 9 9 9
7 9 4 6 2 6 3 9 5 9
2 7 1 6 3 7 5 10 3 6
6 6 5 7 2 5 3 8 7 9
4 3 8 5 4 4 2 8 9 8
4 3 9 4 1 6 3 5 9 9
9 3 8 7 4 4 2 6 7 7
9 10 4 7 2 7 3 8 10 10
7 9 4 9 0 3 1 10 10 9
4 4 2 2 5 2 1 5 4 8
7 6 2 7 6 2 3 8 1 8
72 77 55 72 40 57 33 93 83 101
5 6 10 6 10 6
2 2 1 7 5 7
3 3 5 4 9 7
7 5 5 6 10 7
4 5 0 5 9 9
4 4 0 4 10 5
6 3 4 6 9 6
6 5 10 3 10 9
7 5 10 3 9 9
8 3 0 1 5 5
6 4 4 4 5 5
68 50 58 54 98 84
Scores are the number of correct responses out of 10 items. See Table 2 for abbreviations.
and M.B. respectively. Three patients (M.G., P.Gh., P.Gi.) showed a significant effect of the difficulty level. Two patients (A.D., D.R.) did not show any specific effect for any of the variables. The linear regression analysis showed a significant effect of the complexity index (CI) in seven out of the eight agrammatic patients with comprehension disorders. However, the two patients who showed no specific effect from the five previous variables did differ on the outcome of the linear regression analysis: A.D. showed a general complexity effect, while D.R. did not. Fluent aphasic patients. The passive structure significantly influenced the performance of five fluent aphasic patients (E.Bg., E.Bo., A.G., M.Ca., A.C.), one of whom, however, showed a paradoxical advantage for passive sentences (E.Bo.). Clitic
435
.03
3.33 .78 12.68 2.23 .10 1.51
< .01
.04 < .01 < .01*
.04 < .01 < .01
.03 .04
3.01 10.95 8.05 2.12 .01 .61 .15 2.23 1.57 .68
1.48 1.24 3.33 2.96 .90 3.21 9.24 10.94
5.49 2.48 15.84 .18 2.21 .20
2.55 .07 1.21 .77 3.16 .45 .05 .23 6.68 .07
1.75 .03 .06 3.45 1.68 .41 .47 11.82
< .01
.01
< .01
.04
< .01
.03
1.05 .01 1.59 1.01 1.69 .01
.83 .42 .01 1.60 1.46 2.24 .80 .75 .32 .71
.01 2.10 .07 1.07 4.17 .36 .01 2.29 .02
Degree of pronominalisation w2 p
.21 .01 .40 3.69 1.00 1.02
.68 .47 .03 1.25 6.36 .81 1.54 .42 1.09 .42
1.51 .25 .24 3.50 .27 2.11 .92 .35
w2
.03
.01
.03
p
Number of syllables
9.99 .35 .52 .22 3.64 1.42
.06 3.21 .13 .90 13.52 5.13 .25 4.39 2.01 .13
1.94 .89 2.47 .33 3.73 3.03 5.29 .14
.03
< .01
.02
< .01 .01
.04
.03 .04 .01
Difficulty level for normal controls w2 p
72.35 76.66 733.68 74.01 79.52 710.62
75.82 722.48 1.87 76.40 5.14 712.43 73.91 77.97 722.42 76.64
715.00 76.99 714.05 712.23 715.50 716.49 723.42 715.37
B
.03
< .01
.01 .03
.01
< .01
.01 .01 .01 .01 < .01 .02
.01
p
Regression of CI on performance
CI = Complexity Index. B = angular coefficient of the linear regression. As effects are expected in one direction only, one-tailed (significant) p-values are given. An asterisk indicates a significant result in the unexpected direction (the two-tailed p-value is given in this latter case).
Agrammatism AD DR LZ MB MG PGh PGi RO Wernicke’s aphasia AC EBg EBo EM GD GI LR NM RB VB Conduction aphasia AG CG MCa MCr SM UB
p
w2
w2
p
Clitic movement
Active-Passive
Logistic regression
TABLE 6 DG sentences: Results of the logistic regression and linear regression analyses on single cases
436
LUZZATTI ET AL.
pronouns moving to the preverbal position affected the performance of four patients (G.D., R.B., A.G., M.Ca.); none of the patients showed any effect of the degree of pronominalisation, while the performance of two patients (G.D., M.Cr.) was influenced by sentence length and six patients showed a significant effect of the level of difficulty (E.Bg., G.D., G.I., A.G., S.M., N.M.). Five subjects showed no specific effect of any of the five variables (V.B., C.G., U.B., E.M., L.R.). The linear regression analysis showed a significant effect of the complexity index in six out of sixteen fluent aphasic patients (E.Bg., G.I., R.B., V.B., M.Ca., U.B.). Of the five patients who showed no effect on the logistic regression analysis, two showed a complexity effect on the linear regression analysis (V.B., U.B.), whereas three subjects did not show any effect in either analyses (C.G., E.M., L.R.).
Discussion. The results obtained from the logistic regression analysis show that specific deficits associate (and dissociate) in different patients forming various patterns of impairment, from which inferences can be made regarding the damage to the cognitive system responsible for the comprehension deficit of each single subject. For instance, the comprehension disorder observed in patients L.Z., P.Gh., P.Gi., and R.O. is clearly compatible with Grodzinsky’s account of a selective syntactic impairment of a mechanism connecting the antecedent to its trace. A parallel impairment in the comprehension of passive structures and of clitic pronouns was anticipated, as in both cases movement is involved; however, only two out of five patients with impaired passives also had difficulty with clitic pronouns. It could be argued that clitic pronouns undergo the same kind of movement that the NP subject undergoes in passive sentences, but that unimpaired comprehension of sentences with clitic pronouns results from the adoption of a strategy that assigns the Agent thematic role to the first element of the sentence and the Patient role to the second element, although both elements are in the preverbal position. However, this does not account for the fact that comprehension of both sentences with clitic pronouns and passive sentences was impaired in two patients. Why did they not adopt the strategy? Alternatively, in the spirit of Hickok and Avrutin (1996), it could be proposed that the chain connecting the clitic pronoun to its trace is different from the chain connecting the NP subject in passive sentences to its trace, as some agrammatic patients are unable to form only the latter kind of chain. M.B.’s impairment of the processing of passive sentences and of clitic pronouns is associated to an effect of the sentence length. This association has two possible interpretations: (i) simultaneous damage to the syntactic processing of traces and to phonological short-term memory, or (ii) a unique short-term memory impairment causing in turn a defective syntactic processing of traces. The latter hypothesis, however, contrasts with results from other studies showing normal syntactic processing in the presence of severe short-term memory damage (see Caplan & Waters, 1999, for a review). The significant effect of pronominalisation on M.G.’s comprehension performance is to be attributed mainly to a lexical impairment of pronouns. Two different explanations can be given for the two agrammatic patients (A.D. and D.R.) with a comprehension disorder but no specific impairment to the five variables under study, in relation to the presence or absence of a ‘‘summation’’ effect, i.e., of a significant effect of the complexity index (CI) on the comprehension performance. A.D.’s pattern of performance is compatible with Linebarger et al.’s (1983) trade-off theory, as the patient showed no specific effect for any single variable while there was a general
SENTENCE COMPREHENSION IN AGRAMMATISM
437
summation effect across variables1. D.R., on the other hand, showed a severe (correct analyses for only 51 out of 120 DG sentences) but unspecific comprehension disorder, as no single variable, nor the sum of the variables (CI) can account for his performance. Nevertheless, he showed severe comprehension impairment for simple active sentences, which would appear to be caused either by impaired access to the basic thematic representation of sentences, or by damage to mapping arguments to NP in the sentences. All these impairments are mingled in the fluent aphasic patients with no clear differences with respect to the pattern of distribution observed in agrammatic patients. There are, however, two conditions that only appeared among the fluent aphasic patients: a paradoxical advantage for passive with respect to active sentences was evident in one Wernicke’s patient (E.Bo.) (see Druks & Marshall, 1995, for a possible account of this phenomenon) whereas a length effect with no disadvantage for passive sentences was observed in two cases (G.D., M.Cr.). The performance of these two patients (who had been classified as suffering from a Wernicke’s aphasia and from a conduction aphasia of the repetition type, respectively) is compatible with previous results in favour of an independent processing of syntax from phonological short-term memory (see Caplan & Waters, 1999, for a review).
Past participle agreement in Italian as a cue for an asyntactic analysis of passive sentences On the basis of the TDH and of Hickok et al.’s revised version of the TDH, agrammatic patients would have been expected to perform at chance level in a task testing comprehension of passive sentences. On single-case analysis, however, three patients with severely agrammatic speech output turned out to have normal comprehension. These results are in agreement with previous descriptions of agrammatic patients with spared comprehension (e.g., Miceli, Mazzucchi, Menn, & Goodglass, 1983), but are in contrast with Grodzinsky’s explanation of agrammatism as the consequence of damage to a specific syntactic processor. A defender of the TDH or the extended TDH theory could object that the integrity of the comprehension of these three patients is due to a strategy exploiting the gender and number agreement between the past participle and the grammatical subject in Italian passive sentences. Such agreement would offer a morphophonological cue for an asyntactic interpretation of the sentence when the Agent and the Patient are of different grammatical gender. To rule out this possibility, one of the three patients with normal comprehension of passive sentences but severe agrammatic speech output (F.C.) was tested on a supplementary task in which no gender cue was available and animals were used instead of human stimuli. The other two patients had been studied 10 years previously and were no longer available for testing. As in Italian all nouns are either masculine or feminine, sentences were used in which the Agent and Patient were either both masculine (e.g. il riccio, the porcupine and il coniglio, the rabbit) or feminine (e.g. la lumaca, the snail and la tartaruga, the turtle). A total of 96 sentences (50%active, 50% passive) were used. Stimuli were balanced for the verbs used, the gender of Agent and Patient, and the spatial location of the target pictures in the multiple-choice sets. Items were presented as in Experiment 1, the only difference being that the choice was to
1
It must be pointed out that, as CI corresponds to the sum of the five variables studied in the logistic regression analysis, a significant effect of any of them will easily produce a significant effect of CI as well. Therefore, when a significant effect of CI co-occurs with a significant effect of any of the five single variables, the former cannot be unambiguously interpreted in terms of Linebarger et al.’s trade-off theory.
438
LUZZATTI ET AL.
be made between two instead of four pictures. F.C. performed almost flawlessly on both active (46/48) and passive (47/48) sentences: a result which cannot be explained as the consequence of a non syntactic cueing, and confirms once again that severely agrammatic speech output may co-exist with spared syntactic comprehension.
GENERAL DISCUSSION The results obtained from this study provide information about the mechanisms underlying the syntactic comprehension of reversible active and passive sentences and the pattern of impairment in agrammatism. In addition, it has been shown that sentence comprehension may be damaged along a variety of processing loci, including specific damage to the mechanism connecting the antecedent to its trace (as proposed by Grodzinsky’s TDH, or the modified TDH), a primary processing deficit of function words, and non-linguistic processing disorders caused by a more general reduction in the processing capacity (i.e. Linebarger et al.’s trade-off theory). Different types of impairment emerge in various combinations, thus substantiating a heterogeneous pattern of damage. Furthermore, the flawless performance of three of the agrammatic patients on both active and passive sentences seems to exclude the feasibility of generalising Grodzinsky’s TDH to all agrammatic patients. These results are compatible with those emerging from a meta-analysis of the studies on agrammatic comprehension disorders of reversible active and passive sentences published from 1980 to 1995 (Berndt, Mitchum, & Haendiges, 1996). Excluding the cases in which the set of items used was too small (e.g., Jarema, Kadzielawa, & Waite, 1987), the 41 remaining cases distribute almost equally across three major patterns of performance. The type of impairment described by Grodzinsky—i.e., active sentences better than chance and passive sentences at chance— appears in only 36.6% of the sample, while the remaining patients performed either at chance (26.8%), or better than chance on both active and passive sentences (36.6%). Grodzinsky, Pinango, Zurif, and Drai (1999) attributed the results of Berndt et al.’s metaanalysis to a statistical artefact. According to Grodzinsky and colleagues, the proportion of agrammatic patients showing an above-chance performance on passive sentences is indistinguishable from that predictable on the basis of mere random fluctuations, the true probability of correct identification of the target picture (out of two available) being .5 in every case. The inadequacy of this conclusion has already been convincingly pointed out by Berndt and Caramazza (1999), who noticed that the selection criteria for the inclusion of patients differed considerably in the two studies. However, Grodzinsky et al.’s (1999) objection can also be ruled out by applying the Bonferroni technique to bring the risk of false positives (i.e., agrammatic patients classified as performing above-chance on passive sentences, when in reality they did not) back to .05. In the present study, 11 agrammatic patients were tested: therefore, the null hypothesis that none of them can interpret passive sentences had to be rejected, if at least one of the 11 patients obtained a p-value lower than .05/11, i.e., < .0045. The performance of F.C. and E.Be. (39/40 correctly interpreted passive DG sentences, see Table 5) corresponds to a p-value of .00000000004, and that of C.M. (40/40) to a p-value of .0000000000009. The patients performed above-chance to the extent that the null hypothesis, i.e., the prediction of the TDH, would have been rejected even if a billion patients had been tested. The comprehension deficits observed in the agrammatic patients also emerge with a similar distribution among the fluent aphasic patients, thus confirming the findings of Caplan, Hildebrandt, and Makris (1996), who failed to reveal differences between the patterns of comprehension disorders in patients with anterior and posterior left-hemisphere lesions.
SENTENCE COMPREHENSION IN AGRAMMATISM
439
The present study involved a double analysis of the data obtained, thus integrating the advantages of a group study with those of the single case procedure. The group analysis detected small but consistent effects across the agrammatic patients that did not reach significance in the single cases (e.g., the impaired processing of pronouns). The single case analysis, on the contrary, provided the possibility of studying the inter-individual variability of comprehension disorders, a critical test for the generalisation of a unitary pattern of impairment to all agrammatic patients. Although trace deletion is often associated with agrammatism, the data from this study show that the impaired comprehension of passive sentences is neither necessary, nor in itself sufficient, for the emergence of an agrammatic language disorder. The comprehension performance of several agrammatic and fluent aphasic patients was damaged even on full-NP active sentences. A deficit in this set of items cannot be explained either by an impaired mechanism connecting the antecedent to its trace, or by impaired processing of closed-class elements, or by a trade-off of the syntactic parser, as none of these explanations could account for the better level of comprehension of SG sentences shown by agrammatic as well as by fluent aphasic patients. The only possible explanation for these results is the existence of lexical damage to the thematic grid and/or to the assignment of the thematic role to arguments. The comprehension of sentences with clitic pronouns was less impaired or not impaired at all compared to the comprehension of passive sentences. The fact that some patients have difficulty in comprehending passive sentences, but not sentences with clitic pronouns, supports the theory that clitic pronouns and NP subjects in passive sentences do not undergo the same kind of movement. Sportiche (1995) suggested that clitics are generated in a clitic Voice Projection and move (as head) to the position adjacent to the verb. This movement is an instance of head movement, a process that does not constitute a problem for agrammatic patients (see Lonzi & Luzzatti, 1993). Finally, three patients showed almost flawless comprehension of reversible passive sentences, in spite of the severe agrammatism of their speech output. The possibility that this may be due to a particular aspect of Italian grammar, in which past participles agree with the grammatical subject of passive sentences, was ruled out in one of the patients by using a different condition of the task. Overall, the results of this study confirm that the ability to handle traces may be impaired in agrammatism, but they confute the hypothesis of a unique description and of a unitary explanation for comprehension disorders in agrammatic patients. It is clear that the mechanisms invoked by either the TDH or the revised TDH are not the only mechanism responsible for comprehension disorders in agrammatic patients. Neither the TDH nor the revised TDH can be generalised to all agrammatic patients. Moreover, they also account for comprehension deficits in fluent aphasics. Manuscript received 22 June 2000 Manuscript accepted 17 January 2001
REFERENCES Basso, A., Capitani, E., Laiacona, M., & Luzzatti, C. (1980). Factors influencing type and severity of aphasia. Cortex, 16, 631–636. Berndt, R., & 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., & Caramazza, A. (1999). How ‘‘regular’’ is sentence comprehension in Broca’s aphasics? It depends on how you select the patients. Brain and Language, 67, 242–247.
440
LUZZATTI ET AL.
Berndt, R.S., Mitchum, C.C., & Haendiges, A.N. (1996). Comprehension of reversible sentences in ‘agrammatism’: A meta-analysis. Cognition, 58, 289–308. Bradley, D.C., Garrett, M.F., & Zurif, E.B. (1980). Syntactic deficits in Broca’s aphasia. In D. Caplan (Ed.), Biological studies of mental processes. Cambridge, MA: MIT Press. Caplan, D., & Futter, C. (1986). Assignment of thematic roles to nouns in sentence comprehension by an agrammatic patient. Brain and Language, 27, 117–134. Caplan, D., & Hildebrandt, N. (1988). Disorders of syntactic comprehension. Cambridge, MA: MIT Press. Caplan, D., Hildebrandt, N., & Makris, N. (1996). Location of lesions in stroke patients with deficits in syntactic processing in sentence comprehension. Brain, 119, 933–949. Caplan, D., & Waters, G.S. (1995). Aphasic disorders of syntactic comprehension and working memory capacity. Cognitive Neuropsychology, 12, 637–649. Caplan, D., & Waters, G.S. (1999). Verbal working memory and sentence comprehension. Behavioral and Brain Sciences, 22, 77–126. Caramazza, A., & Zurif, E.B. (1976). Dissociation of algorithmic and heuristic processes in sentence comprehension: Evidence from aphasia. Brain and Language, 3, 572–582. Cecchetto, C. (2000). Doubling structures and reconstruction. Probus, 12, 93–126. Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Floris. Cinque, G. (1990). Types of A’-dependencies. Cambridge, MA: MIT Press. De Mauro, T., Mancini, F., Vedovelli, M., & Voghera, M. (1993). Lessico di frequenza dell’italiano parlato (LIP). Milan: Etaslibri. Druks, J., & Marshall, J.C. (1995). When passives are easier than actives: Two case studies of aphasic comprehension. Cognition, 55, 311–331. Goodglass, H. (1968). Studies in the grammar of aphasics. In S. Rosenberg & J. Koplin, (Eds.), Developments in applied psycholinguistics research. New York: Macmillan. Grimshaw, J. (1990). Argument structure. Cambridge, MA: MIT Press. Grodzinsky, Y. (1986). Language deficits and the theory of syntax. Brain and Language, 27, 135–159. Grodzinsky, Y. (1989). Agrammatic comprehension of relative clauses. Brain and Language, 37, 480–499. Grodzinsky, Y. (1995). Trace deletion, y-roles and cognitive strategy. Brain and Language, 51, 469–497. Grodzinsky, Y., Finkelstein, D., Nicol, J., & Zurif, E.B. (1988). Agrammatic comprehension and the thematic structure of verbs. Paper presented at the 26th Annual Meeting of the Academy of Aphasia. Montre al, Que bec. Grodzinsky, Y., Pinango, M.M., Zurif, E., & Drai, D. (1999). The critical role of group studies in neuropsychology. Comprehension of regularities in Broca’s aphasia. Brain and Language, 67, 134–147. Hickok, G., & Avrutin, S. (1996). Comprehension of wh-questions in two Broca’s aphasics. Brain and Language, 52, 314–327. Hickok, G., Zurif, E., & Conseco-Gonzales, E. (1993). Structural description of agrammatic comprehension. Brain and Language, 45, 371–395. Isserlin, M. (1922). UÈ ber Agrammatismus. Zeitschrift fuÈ r die gesamte Neurologie und Psychiatrie, 75, 332–410. Jackendoff, R. (1990). Semantic structures. Cambridge, MA: MIT Press. Jarema, G., Kadzielawa, D., & Waite, Y. (1987). On comprehension of active/passive sentences and language processing in a Polish agrammatic aphasic. Brain and Language, 32, 215–232. Just, M.A., & Carpenter, P.A. (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review, 99, 122–149. Kayne, R. (1989). Facets of Romance past particle agreement. In P. BenincaÁ (Ed.), Dialects variation and the theory of grammar. Dordrecht: Foris. Kean, M.L. (1977). The linguistic interpretation of aphasic syndromes: Agrammatism in Broca’s aphasia. An example. Cognition, 5, 9–46. Kolk, H., & Heeschen, C. (1992). Agrammatism, paragrammatism and the management of language. Language and Cognitive Processes, 7, 89–129. Koopman, H., & Sportiche, D. (1991). The position of subjects. Lingua, 85, 211–258. Lenneberg, E.H. (1967). Biological foundations of language. New York: Wiley. Linebarger, M.C., Schwartz, M., & Saffran, E. (1983). Sensitivity to grammatical structure in so-called agrammatic aphasics. Cognition, 13, 361–393. Lonzi, L., & Luzzatti, C. (1993). Relevance of adverb distribution for the analysis of sentence representation in agrammatic patients. Brain and Language, 45, 306–317. Luzzatti, C., Willmes, K., & De Bleser, R. (1996). Aachener Aphasie Test (AAT). Florence: Organizzazioni Speciali. McNally, L. (1992). VP coordination and the VP internal subject hypothesis. Linguistic Inquiry, 23, 336–341.
SENTENCE COMPREHENSION IN AGRAMMATISM
441
McCullagh, P., & Nelder, J.A. (1983). Generalised linear models. London: Chapman & Hall. Miceli, G., Mazzucchi, A., Menn, L., & Goodglass, H. (1983). Contrasting cases of Italian agrammatic aphasia without comprehension disorder. Brain and Language, 19, 65–97. Obler, L.K., Albert, M.L., Goodglass, H., & Benton, D.F. (1978). Aphasia type and aging. Brain and Language, 6, 318–322. Ouhalla, J. (1993). Functional categories, agrammatism and language acquisition. Linguistische Berichte, 143, 3–36. Rizzi, L. (1990). Relativized minimality. Cambridge, MA: MIT Press. Saffran, E.M., Schwartz, M.F., & Marin, O.S. (1980). The word order problem in agrammatism. II. Production. Brain and Language, 10, 263–280. Schwartz, M.F., Saffran, E.M., & Marin, O.S. (1980). The word order problem in agrammatism. I. Comprehension. Brain and Language, 10, 249–262. Sportiche, D. (1995). Clitic constructions. In R. Johan & L. Zaring (Eds.), Phrase structure and the lexicon. Bloomington, IN: IUCL Press. Willmes, K. (1985). An approach to analysing a single subject’s scores obtained in a standardised test with application to the Aachen Aphasia Test (AAT). Journal of Clinical and Experimental Neuropsychology, 7, 331–52.
