The Generalised Minimality (GM hereafter) hypothesis was formulated by Grillo (2009) within the linguistic framework of Relativized Minimality (RM; Rizzi 1990; 2004). RM operates as a principle that restricts the possible syntactic chains in natural languages to local configurations. For example in (1), X and Y are able to stand in a local configuration if Z does not belong to the same structural type as X and cannot, therefore, act as a potential intervener.

Rizzi (2004) postulated four different classes of syntactic features –see (2a)–(2d)– related to specific morphosyntactic features associated with syntactic heads.

For example, the displaced constituent which boy in (3a) can enter into a local configuration with its copy because they both belong to the same class (while the girl does not); whereas in (3b), an element of the same class between the antecedent and the copy intervenes, therefore minimality effects arise blocking the formation of the syntactic chain.

According to Grillo (2009), the aphasics’ representations are impoverished with respect to discourse-related morphosyntactic features due to a processing deficit.1 As a result, moved elements are indistinguishable from intervening elements if they lack the relevant features that make them belong to a specific feature class. This is represented in example (4) (from Grillo 2008): the feature wh is underspecified (marked with an asterisk) and, as a consequence, the displaced object belongs to the same class as the subject, which stands between the object and its copy; hence, the subject intervenes and blocks the formation of the syntactic chain (marked with a question mark), disrupting the transmission of the thematic role. Therefore, by hypothesis the comprehension deficit observed in agrammatism is the result of generalised minimality effects.

The hypothesis is motivated by psycholinguistic studies that show that different linguistic levels are processed at different times, more specifically, that discourse information is processed at a later stage (Frazier 1998). Grillo (2009) interprets this in line with the movement hierarchy proposed by Abels (2007), according to which morphosyntactic features are merged earlier than scope-discourse features: that is, arguments are first merged in the corresponding thematic position, and move up to other syntactic positions later.

The claim is therefore that aphasic individuals are able to represent morphosyntactic features merged at an earlier stage, namely case, gender and number. Therefore, the presence of discrepant ϕ-features should suffice to link moved constituents to their copies (5). Yet, Grillo was aware of the fact that a mismatch of ϕ-features does not improve agrammatic comprehension (see Burchert et al. 2003; 2005a for case in German; Friedmann & Shapiro 2003 in Hebrew; Friedmann et al. 2010 in Russian), and explicitly stated that a gender or number mismatch alone “is not enough to avoid a minimality effect unless that feature introduces a change of class” (2009:1433). The relevant features are those that can act as attractors of syntactic movement, which is the case for number in Catalan as it is overtly expressed in the inflected verb and is part of the ϕ-complex that can function as an attractor and trigger the movement of the subject. The presence of mismatched number features between the subject and the object is therefore predicted to block minimality effects in sentences like the one in (5) in agrammatism. Hence, the GM hypothesis predicts that these structures will be well-comprehended by agrammatic aphasics:

Further support for the relevance of the presence of certain features to avoid generalised minimality effects comes from some experimental work in acquisition, in which the poor comprehension of object relatives is also analysed as an instance of generalised minimality effects (see, among others, Friedmann et al. 2009; Adani 2010; Belletti et al. 2012), and also from studies on other types of aphasia (Martini et al. 2020).

Clitic left dislocation (hereafter CLLD; Rivero 1980; Cinque 1990; Rizzi 1990; Ott 2014) is a quite productive construction in Catalan. It involves a topicalised element –el vi ‘the wine’ in (6)– that undergoes movement to the left periphery (Rizzi 2004; Villalba 2009; López 2009; Rubio 2014), and a resumptive clitic that agrees with the topicalised element in number, gender and case –el in (6)–.

Focalised constructions can also present the focalised element in the left periphery –EL VI ‘the wine’ in (7)–, but they do not involve a resumptive element (Rizzi 2004; Cinque & Rizzi 2008):

Both structures have pragmatic import. Topicalisation is used to set apart the old information in an utterance, in order to emphasise that some new information –comment– about the topicalised constituent is going to be introduced (Bianchi & Frascarelli 2010). By contrast, focalisation is used to stress certain information, and especially to contrast new information with the presupposed belief of the listener (within parentheses in (7)).

The sentences above show the subject-object inversion typically observed in Catalan left-peripheral structures. Null-subject languages like Catalan are relatively permissive regarding the position of the subject. In fact, the position of preverbal and postverbal subjects in Romance languages has been a much-discussed topic in the syntactic literature (among others, Rizzi 1982; Suñer 2003; Villa-García 2013; and Bonet 1990; Solà 1992; Vallduví 1993 for Catalan). It is argued that discourse properties motivate the movement of the subject (Belletti 2001) from the VP-internal position: postverbal subjects to the [Spec, FocP] position (Ordóñez 2007), and preverbal subjects to an A’-position higher than [Spec, TP] where they are interpreted as topics (Solà 1992; López 2009).

In the present study declaratives and subject topicalisations and focalisations displaying the canonical SVO word order were also included as control items. The same word order can therefore be found in declaratives, topicalisations and focalisations, and the only overt marker differentiating these structures in Central Catalan is the intonational contour (Prieto 2002). Therefore, prosody plays a crucial role in the interpretation of these structures. The intonational contour of neutral declaratives is a general falling F0 (L+>H*L*L%). The topic, on the other hand, constitutes an independent unit isolated from the rest of the utterance and is realized by a rising F0 (L+H*H%), which is the typical frequency to leave the utterance unfinished and introduce a new prosodic unit whose content depends on the first one. The second unit, the comment, is characterised by a falling contour (L*L%). In contrast, the contrastive focus is prosodically marked by a prominent accentuation of the focused item by a rising low F0: after the high peak, the F0 falls immediately to reach a low frequency (L+H*L%). The rest of the utterance is realized with a falling F0 (L*L%). There are some studies showing that individuals with aphasia are able to discriminate different intonational contours (among others, Gavarró & Salmons 2013 for Catalan, Burchert et al. 2005b and Raithel 2005 for German).

To our knowledge, the only study on the comprehension of Romance CLLD in agrammatism was carried out by Beretta et al. (2001). The authors conducted a sentence-picture matching task with two Spanish-speaking Broca’s aphasics including declaratives and clitic left dislocations with preverbal subjects. The declarative sentences were well-comprehended by the two patients (mean of 75% and 85% of correct responses, respectively). However, their performance on CLLD was not as consistent, given that one of the subjects reached a 75% of correct responses, whereas the other subject performed very poorly (35% of correct responses). Considering the overall performance on the CLLD condition, the authors concluded that the interpretation of this structure was compromised in agrammatism.

Other studies have investigated the agrammatic interpretation of syntactic constructions that involve movement of constituents and resumptive clitics, such as Jovanov (2011) for Serbian and Greek dislocations, and Friedmann (2008) for Hebrew relative clauses. In both cases, the aphasic participants failed to comprehend the sentences with the displaced direct object despite the presence of the resumptive clitic.

Studies on the agrammatic comprehension of clitics in Romance are less scarce. The comprehension of object clitics has been shown to be unimpaired in Catalan (Gavarró 2008), Galician (Martínez-Ferreiro 2010), Italian (Luzzatti et al. 2001) and Spanish (Baauw & Cuetos 2003), as is also the case for indirect object clitics (Luzzatti et al. 2001 for Italian and Salmons 2015 for Catalan). However, the focus of interest of most of these studies lay rather on whether aphasic individuals interpreted clitics as pronouns or as reflexives; for this reason, the experimental designs did not include role-reversals as distractors. The only exception is the study carried out by Luzzatti et al. (2001), who reported that 11 Italian-speaking agrammatic subjects reached a 75.5% of correct responses in a sentence-picture matching task that tested the comprehension of object clitics.

Several studies have investigated the comprehension of object focalisations in agrammatism, and have found robust results across languages which indicate that the aphasic subjects’ interpretation of such structures is impaired (Catalan: Gavarró (2005); German: Burchert et al. (2003); Greek and Serbian: Jovanov (2011); Hebrew: Friedmann & Shapiro (2003): Russian: Friedmann et al. (2010), among others). Most of these studies included languages that have morphologically overt case markers, which did not help the aphasic participants to interpret object focalisations.

On the other hand, the research on feature representation in agrammatism and how it may interact with the comprehension deficit focuses mostly on the role of case on other types of syntactic construction, namely relative clauses and wh-questions (Varlokosta et al. 2014; Sheppard et al. 2015; Adelt et al. 2017; Terzi & Nanousi 2018; Adelt et al. 2020).

Friedmann et al. (2016) investigated the role of case in the interpretation of Hebrew object focalisations by seven agrammatic subjects who performed at chance in a sentence-picture matching task. The authors interpreted the results as evidence that aphasic individuals were sensitive to the accusative marker2 but were unable to use it as a comprehension cue, given that this morphosyntactic feature does not attract movement in Hebrew and, hence, it is irrelevant in terms of blocking minimality effects (in line with what Grillo (2009) argued).

Terzi & Nanousi (2018) investigated whether gender had any effect on the comprehension of relative clauses in agrammatism. They conducted an experimental task with six aphasic participants on the comprehension of Greek relative clauses. The aphasic participants’ comprehension of subject relatives was good, whereas the comprehension of object relatives was poor. Terzi & Nanousi (2018) investigated whether gender had any effect on the comprehension of relative clauses in agrammatism. Despite the fact that gender is not a criterial feature in Greek, gender had an effect in intervention effects, since the aphasic subjects performed significantly better on object relatives where the subject and the object mismatched in gender (mean of 68.75% of correct responses) than on those sentences with matched gender (mean of 43.05% of correct responses).

To our knowledge, there is only one study on how a relevant criterial feature, number, may interact with the comprehension deficit: Adelt et al. (2017) conducted a task on the comprehension of German relative clauses and, interestingly, the aphasic participants only reached a mean of 60.6% and 59.4% of correct responses on subject and object relatives, respectively, when they presented mismatched number between the subject and the object. Number is a syntactically active feature in German and is predicted to block minimality effects by the GM, but that prediction does not seem to be fulfilled. The fact that the aphasic sample was very heterogeneous (5 Broca, 4 anomic and 1 unspecified aphasic participants) could have influenced the results.

In this connection, it has often been observed in the agrammatic literature that there is variability in performance among subjects through different types of structures. This led some researchers to claim that agrammatic aphasia is too broad to be associated with specific patterns of disruption (Beretta et al. 1996; Caramazza et al. 2001), and gave rise to a debate on the legitimacy of group studies and hypotheses that attempt to characterise Broca’s aphasia as a uniform syndrome. Variability is especially apparent when samples are small.

In fact, there is wide evidence that agrammatic aphasics tend to reproduce general patterns of performance across methodologies and languages (Grodzinsky et al. 1999; Drai 2006; Grodzinsky 2010) in spite of the variability among subjects. Drai and Grodzinsky (2006a; 2006b), for example, analysed data from sixty-nine aphasic patients with a new method that allowed quantification of individual variation and the analysis of group performance, and the results revealed that general patterns of response could still be observed across tasks and languages, and that they mainly concerned syntactic structures that involved movement. Neuroimaging studies with healthy subjects also revealed that minimal contrasts involving these type of sentences –but not other syntactic constructions (Santi & Grodzinsky 2007)– consistently activated the left inferior frontal gyrus (Santi and Grodzinsky 2010; 2012; Grodzinsky et al. 2021), which has been argued to be damaged in Broca’s aphasia (Geschwind 1970).

Hence, more research is needed given that the few studies on the morphosyntactic representation and intervention effects in agrammatism report diverging results. In what follows we present two experiments targeting the agrammatic comprehension of clitic left dislocation and contrastive focus in Catalan to evaluate the hypothesis that the deficit in agrammatic aphasia is an instance of generalised minimality effects (Grillo 2008), and to examine whether a relevant feature in Catalan like number can block intervention effects as hypothesised by the GM. On the basis of the theoretical framework outlined, the predicitions on the performance of subjects with agrammatic aphasia on the syntactic constructions evaluated appear in Table 1, where potential interveners are underlined. Under the GM hypothesis, no differences in performance are predicted between object topicalisations and focalisations with pre- and postverbal subjects,3 despite that OSV sentences involve one more movement operation: the subject acts as a potential intervener between the displaced object and the copy in both type of sentences when there is a match of ϕ-features between the subject and the object (8a) and, hence, generalised minimality effects arise. By contrast, mismatched OVS dislocations and focalisations are predicted to be well-comprehended.

Statistical analyses were performed using SAS software, version 9.2 (SAS Institute Inc., USA). Nonparametric tests were used given that the distribution of the data was not normal. A logistic regression model for binomial data was performed to model the probability of success, and to calculate the odds ratio and 95% confidence intervals considering group and sentence type. Also, a standard goodness of fit test was conducted to calculate and compare the individual estimated proportions of correct responses with a random pattern of response in order to learn whether the performance of the participants was at chance. Statistical significance was set at p<0.05.

The control participants performed at ceiling on all sentence types, whereas the aphasic performance varied across conditions (Table 3). The aphasic group reached high percentages of correct responses on control items, that is, declaratives and subject topicalisations. However, the logistics regression analysis determined that the performance of control and aphasic participants on these two conditions differed significantly (OR = 6.56, 95% CI [1.89, 22.77]) for declaratives and OR = 46.32, 95% CI [4.92, 54.13] for subject topicalisations). Regarding the comprehension of direct object clitics, aphasic participants reached a 80.6% of correct responses. Hence, object clitics were well interpreted, though their comprehension was significantly poorer than the control’s performance (OR = 21.48, 95% CI [5.06, 91.28]) for controls vs aphasics). On the other hand, agrammatic subjects reached lower percentages of correct responses in the two dislocation conditions: patients showed a mean percentage of 48.9% of correct responses on CLLDs with postverbal subjects, whereas their performance on CLLDs with preverbal subjects was much better (mean of 67.8% of correct responses). The performance between control and aphasic participants also differed in these two conditions, the odds ratio of the control group being 46 and 21 times higher than the aphasics’ on dislocations with postverbal (95% CI [7.37, 59.35]) and preverbal subjects (95% CI [16.31, 129.73]), respectively.

Even though the aphasic performance was on average slightly worse than the controls’ performance through all conditions, the patients showed a good comprehension of declaratives, subject topicalisations and object clitics (estimated means of correct responses: 0.9, 95% CI [0.83, 0.94]; 0.78, 95% CI [0.69, 0.85] and 0.81, 95% CI [0.72, 0.87], respectively).5 Regarding the comprehension of dislocations, the aphasic performance on dislocations with preverbal subjects differed significantly from their performance on CLLDs with postverbal subjects (estimated means of 0.68, 95% CI [0.58, 0.76] and 0.49, 95% CI [0.39, 0.59], respectively), as the former were better comprehended than the latter. This was also confirmed by the odds ratio: OR = 2.2, 95% CI [1.43, 3.38]. Nevertheless, the comprehension of CLLDs with preverbal subjects was significantly worse than the comprehension of declaratives (OR = 0.23, 95% CI [0.13, 0.42]), object clitics (OR = 0.46, 95% CI [0.25, 0.85]), and subject topicalisations (OR for SVO vs OSclV topicalisations = 1.72, 95% CI [1.07, 2.76]).

Since variability among subjects was observed (Table 4), the individual estimated proportions of correct responses were compared with a random pattern of response in order to learn whether aphasics’ performance could be at chance. This allowed us to determine if the performance was good –above chance– or poor –at or below chance– in order to evaluate the predictions of the GM hypothesis. The p-values from the GOF test for all the conditions are represented in Figure 2: p-values lower than 0.05 –marked with a line in the graph– indicate that the null hypothesis that the aphasics’ performance presents a random distribution is to be ruled out. The test revealed that the guessing performance can be ruled out for all patients in the declarative condition. Regarding the comprehension of subject topicalisations, seven out of nine patients showed an above-chance performance, while two patients, A8 and A9, performed at chance. The aphasic subjects’ performance on object clitics was also above chance, except for three patients whose performance was at chance. Guessing performance cannot be ruled out for most aphasic subjects’ performance on both dislocation conditions: on the one hand, eight out of nine patients performed at chance in the OclVS condition, and the other one below chance; and on the other hand, six patients showed an at-chance performance in OSclV dislocations while the other three performed above chance.

Concerning the impact of mismatch of the relevant ϕ-features –number– between the subject and the object in comparison with the comprehension of matched sentences, the overall results showed that there were no statistically significant differences between matched and mismatched sentences (estimated means of correct responses: 0.72, 95% CI [0.64, 0.79] and 0.79, 95% CI [0.72, 0.85], respectively). The correct percentages by condition and sentence type are represented in Figure 3; the odds ratio analysis revealed that significant differences between matched and mismatched sentences are only found in the object clitic condition (OR = 0.45, 95% CI [0.21, 0.97]), despite the fact that the aphasic performance on this condition was above chance (estimated means of 0.75, 95% CI [0.62, 0.84] for matched sentences and 0.87, 95% CI [0.77, 0.93] for mismatched sentences). No statistically significant differences were found between matched and mismatched sentences in dislocation conditions (OR = 0.74, 95% CI [0.39, 1.38] for OSclV and OR = 0.59, 95% CI [0.32, 1.06] for OclVS); thus, the presence of the clitic, and the number mismatch did not prevent aphasic participants from performing poorly in the task.

To summarize, we conducted a sentence-picture matching task to test the comprehension of left clitic dislocation and object clitics in Catalan. The nine individuals diagnosed with Broca’s aphasia that participated in the study showed a good performance on declaratives (9a), subjects topicalisations (9b) and sentences with object clitics (9e). In contrast, their performance on clitic left dislocations with preverbal (9c) and postverbal (9d) subjects was impaired. Also, intersubject variability was observed, specially on dislocations and object clitics. The study aimed at evaluating the GM hypothesis, according to which the comprehension of matched CLLD is predicted to be poor (8a), whereas the comprehension of mismatched dislocations is predicted to be good (8b). The statistical analysis, however, revealed that there were no significant differences between matcheded and mismatched dislocations, against the hypothesis’ predictions.

The statistical analysis conducted is described in section 3.2. The control subjects performed at ceiling on all conditions, whereas the aphasic individuals only reached a high percentage of correct responses in the declarative and subject focalisation conditions (see Table 6). In fact, there were no statistically significant differences between groups in these two conditions, as the odds ratio revealed (OR = 5.22, 95% CI [0.54, 50.33] for declaratives and OR = 2.38, 95% CI [0.59, 9.47] for subject focalisations). The aphasics’ performance on object focalisations was much lower, and the statistical analysis indicated that the control and aphasic performances differed significantly in both conditions (OR = 101.04, 95% CI [12.16, 839.95] for object focalisations with preverbal subjects and OR = 22.56, 95% CI [6.38, 79.80] for object focalisations with postverbal subjects).

Concerning the aphasics’ responses by sentence type, the statistical analysis revealed that the difference in performance between the OVS and OSV focus conditions was not significant (OR = 1.09, 95% CI [0.52, 2.27]), and neither was the difference between declaratives and subject focalisations (OR = 1.88, 95% CI [0.43, 8.31]). On the other hand, the aphasics’ performance on the two object focalisations differed significantly from the declarative and subject focalisation conditions, as the odds ratio indicated: OR = 17.86, CI (4.91, 64.96) for declaratives vs OVS focus, OR = 19.38, 95% CI [5.33, 79.48] for declaratives vs OSV focus, OR = 9.49, 95% CI [3.36, 26.86] for subject focus vs OVS focus, and OR = 10.30, 95% CI [3.64, 29.15] for subject focus vs OSV focus.

The pattern of response described above is reproduced individually (Table 7), except for patient A1, who showed an almost normal performance on all conditions.

A standard goodness of fit test was carried out to learn whether the distribution of the aphasics’ responses was at chance or not. As the p-values represented in Figure 5 indicate, the null hypothesis that their performance on object focalisations is at chance cannot be ruled out for most patients. The only exceptions are, on the one hand, patient A1 who, as already pointed out, performed above chance on all conditions and, on the other hand, patient A4, whose performance on OSV focalisations was below chance, and patients A2 and A6, who performed below and above chance on OVS focalisations, respectively. On the other hand, all participants performed above chance on declaratives, as did 6 of the 7 participants on subject focalisations.

We carried out a sentence-picture matching task to test the comprehension of contrastive focus in Catalan. The seven subjects with agrammatic aphasia that participated in the study showed a good performance of declaratives (12a) and subject topicalisations (12b), which is consistent with GM hypotheses. Their performance on object focalisations with preverbal (12c) and postverbal (12d) subjects, on the other hand, was poor. Like in the previous experiment, the object focalisation conditions involved greater intersubject variability, though most participants performed at or below chance on these sentences. Our findings call into question the GM hypothesis, as all the sentences included in the task involved mismatching of ϕ-feature between the subject and the object and, therefore, are predicted to be well-comprehended in agrammatic aphasia (11).

The results showed that the agrammatic subjects were able to comprehend declaratives, subject topicalisations and subject focalisations (Tables 3 and 6); that is, those sentences that display the superficial word order SVO to which thematic roles are assigned canonically. According to the logistic regression model, the results were verified at the individual level: on the one hand, all the participants performed above chance on declaratives in both experiments and, on the other hand, 7/9 and 6/7 subjects with aphasia showed an above-chance performance on subject topicalisations and focalisations, respectively. These findings are in line with previous research on the comprehension of active sentences with canonical word orders (among others, Burchert et al. 2003; Gavarró 2005; Friedmann et al. 2010; Jovanov 2011; Maviş et al. 2019). The results are also consistent with the GM hypothesis since these structures do not involve any potential intervener between the displaced constituent and its copy (14) and, therefore, minimality effects are avoided.

The agrammatic participants also showed a good comprehension of object clitics (Table 3), as has been observed in previous literature (among others, Luzzatti et al. 2001; Baauw & Cuetos 2003; Martínez-Ferreiro 2010) and despite the superficial noncanonical word order of these structures (15). However, as already pointed out, most of these studies focused on the pronominal versus reflexive interpretation of clitics. In contrast, the focus of the present study was on the thematic interpretation by agrammatic aphasics, who succesfully pointed to the picture that displays the correct thematic interpretation both at the group and individual level, as most participants performed above chance on this condition (6/9 patients). The inclusion of tokens with clitics also allows us to rule out the possibility that the presence of the clitic is responsible for the poor comprehension on CLLDs. Both movement-derived and base-generated accounts (see, among others, Suñer 1988) have been proposed in order to explain why clitic constructions present an atypical word order configuration. According to derivational analyses, clitics ultimately move as heads and adjoin the T head (Kayne 1989; Belletti 1999; and Solà 2002 for Catalan). Head movement (mainly verb movement) has been shown not to be problematic for agrammatic subjects (among others, Lonzi & Luzzatti 1993; Grodzinsky & Finkel 1998; Friedmann & Shapiro 2003; Bastiaanse & Edwards 2004), and it is predicted not be impaired within the GM framework (Grillo 2009). Further support for this conclusion comes from a study by Garraffa (2009), who carried out a grammaticality judgement task to evaluate the subject-verb agreement in Italian sentences with object clitics. One subject with nonfluent aphasia performed above chance on grammatical sentences with pre- and postverbal subjects, suggesting that the clitic did not intervene in these cases. Hence, our findings do not challenge the GM hypothesis under the analysis of clitics as an instance of head movement, since individuals with agrammatism are not hypothesised to have problems in comprehending these structures (15). The fact that significant differences between matched and mismatched sentences were only observed in the object clitic condition is orthogonal to the hypothesis, since no intervention configuration holds here (Table 1).

An anonymous reviewer pointed out the possibility that subjects with aphasia may have been able to interpret sentences with object clitics by assigning an SV structure (or simply identifying the first DP as subject). Yet, if that was the case, we would have expected them to apply the same strategy with dislocations with preverbal subjects. Also, in a study conducted by Peristeri & Tsimpli (2013), Broca’s aphasics correctly interpreted the gender and number of object clitics in Greek. These findings suggest that they interpret object clitics.

The agrammatic interpretation of object dislocations and focalisations was compromised (Tables 3 and 6). This finding was verified at an individual level: 6 and 9 out of 9 subjects with aphasia showed a poor performance on CLLDs with preverbal and postverbal subjects, respectively, and 6 and 5 out of 7 participants performed poorly on object focalisations with preverbal and postverbal subjects (recall that, in our sample, one patient who had been classified as mild performed well on all item types). The presence of a resumptive clitic did not improve the comprehension of CLLD. The results on the comprehension of object focalisations are consistent with crosslinguistic previous research (among others, Burchert et al. 2003; Friedmann & Shapiro 2003; Gavarró 2005; Friedmann et al. 2010; Jovanov 2011).

Concerning the comprehension of CLLDs, the results reported in the present study differ from the only antecedent in Romance, Beretta et al. (2001), given that the nine Catalan-speaking aphasic individuals showed a higher percentage of correct responses in OSclV dislocations (67.8%) than the two Spanish-speaking (55%) who participated in Beretta et al. (2001). However, the sample of the latter study was smaller and very heterogeneneous: the two aphasic subjects’ performances contrasted sharply as one participant reached a mean of 75% of correct responses, while the other one only produced 35% of correct responses.

In the present study, CLLDs with preverbal subjects were slightly better understood than their OVS counterparts, while object focalisations were comprehended at chance regardless of the position of the subject. In fact, the position of the subject was proven not to be relevant in Spanish actives and adjectival passives with preverbal and postverbal subjects in Beretta et al. (1996). It is possible that the three aphasic participants who understood OSclV above chance in the present study (Figure 2) assimilated them to clitic constructions. Several factors might have favoured this interpretation: on the one hand, the items tested contained an adverbial phrase to stress the contrastive feature of CLLD that stood between the dislocate and the matrix clause (en canvi ‘in contrast’), and, on the other hand, the topic and the comment belonged to two different units prosodically differentiated to which aphasics are sensitive (Gavarró & Salmons 2013), hence highlighting the SclV configuration of the comment. If so, the participants may have parsed OSclV simply as SclV, and therefore reached the target interpretation.

Regarding the GM’s predictions, the example in (16) reproduces the agrammatic representation of the matched CLLD tested in the present study: since the discourse-feature topic is underspecified, the topicalised element al cocodril ‘the cocodrile’ belongs to the argumental class and, such as, it is undistinguishable from the intervening subject el mico ‘the monkey’. Therefore, our findings do not contradict the prediction that the aphasic comprehension of these sentences will be impaired:

The present study crucially contributes to testing the GM hypothesis by examining the representation of relevant morphosyntactic features that can prevent minimality effects, number in particular. Under the GM, the presence of intact mismatched number features is predicted to suffice to link moved constituent to their copies, as they will block intervention effects in both object dislocations and focalisations with preverbal (17a) and postverbal (17b) subjects:

However, the results on object dislocations and focalisations in the present study (Figure 3 and Table 3) did not improve when the subject and the object were mismatched in number. These findings bring the GM hypothesis into question. Still, Grillo (2008; 2009) could argue that the number feature does not suffice to avoid minimality effects, as the featural class that it introduces –argumental– is no different from the intervener’s class. However, in a similar situation, the presence of the distinctive feature NP –which also introduces argumental class– is argued to account for the poor performance of aphasic individuals on object which-questions in contrast to object who-questions, which are well comprehended (Hickok & Avrutin 1996). This contrast in performance was explained in Grillo (2009:1436) because ‘a perfect match between the impoverished moved object and the intervening subject obtains only with the which-NP type of question’, and was later developed by Friedmann et al. (2009), who pointed out that the relevant property that sets which-questions (18a) apart from who-questions (18b) is the fact that the former contain a lexical NP, the same as the intervening subject.

The presence of mismatched features is therefore predicted to block minimality effects in object focalisations (17a) and dislocations (17b) in Catalan. Our findings challenge this hypothesis, as agrammatics perform at chance in these conditions, in line with previous research (Adelt et al. 2017). These findings may suggest that the aphasics’ representations are much more impoverished than argued by Grillo. The results in the present study do not suffice to rule out the hypothesis that the deficit observed in agrammatic aphasia is an instance of generalised minimality effects. The notion of intervention is relevant in linguistic theory in order to explain the formation and blocking of syntactic chains, and it has also been shown to be useful to explain the aphasics’ poor performance in specific movement-derived sentences (Grodzinsky 2006; Grillo 2008; Friedmann & Gvion 2012), given that canonicity or movement alone do not suffice to account for the comprehension deficit (Table 8). What is called into question is the specific division of features proposed by Grillo.

Finally, the presence of the colloquial accusative marker in both dislocations and focalisations did not suffice to interpret them, in line with what has been observed in the literature (see 2.2). These results are consistent with the GM hypothesis, since case is not predicted to block or give rise to minimality effects. But as already discussed in the introduction, few studies have claimed that case markers improve the comprehension of subjects with Broca’s aphasia (Kljajevic & Murasugi 2010; Terzi & Nanousi 2018; Adelt et al. 2017; 2020). Case markers cannot be used as cues to comprehension because the assignment of thematic roles is not mediated by case markers. In fact, specific cases do not correspond to specific thematic roles (e.g. the subject in passives or unaccusatives receives the theme role) as they express the syntactic function and not the semantic interpretation of the constituent. Therefore, the putative sensitivity of Broca’s aphasics to case marking would not entail a correct thematic assignment.

In the present study, variability among subjects was also observed, probably due to the heterogeneity of the aphasic group with regard to factors that have been shown to be relevant such as etiology, brain anatomy or lesion site (e.g. Geschwind 1970; Tomaiuolo et al. 1999; Amunts & Zilles 2006). Despite the variability, our findings reflected general patterns of response that have been previously observed in the aphasiology literature and that are replicated at an individual level. Interestingly, our results showed that problematic structures involved more heterogenous responses than well-comprehended sentences: between-subject variability increased on poorly understood constructions, which probably reflected the aphasics’ problems in comprehension.