- Email: [email protected]
Patients with psychosis struggle with scalar implicatures
SCHRES-07502; No of Pages 6 Schizophrenia Research xxx (2017) xxx–xxx
Contents lists available at ScienceDirect
Schizophrenia Research journal homepage: www.elsevier.com/locate/schres
Patients with psychosis struggle with scalar implicatures Martien Wampers a,b,⁎, Sofie Schrauwen c, Marc De Hert a,b, Leen Gielen a, Walter Schaeken c a b c
University Psychiatric Hospital KU Leuven, Leuvensesteenweg 517, 3070 Kortenberg, Belgium University of Leuven, Faculty of Medicines, Department of Neurosciences, Herestraat 49, Box 721, 3000 Leuven, Belgium University of Leuven, Faculty of Psychology and Educational Sciences, Laboratory of Experimental Psychology, Tiensestraat 101, Leuven, Belgium
a r t i c l e
i n f o
Article history: Received 27 March 2017 Received in revised form 6 July 2017 Accepted 28 August 2017 Available online xxxx Keywords: Psychosis Language Pragmatics Reasoning Scalar implicatures
a b s t r a c t Pragmatic language difficulties in people with psychosis have been demonstrated repeatedly but one of the most studied types of pragmatic language, i.e. scalar implicatures (SIs), has not yet been examined in this population. SIs are a special kind of pragmatic inferences, based on linguistic expressions like some, or, must. Such expressions are part of a scale of informativeness organized by informativity (e.g. some/many/all). Although semantically the less informative expressions imply the more informative ones, pragmatically people generally infer that the use of a less informative expression implies that the more informative option is not applicable. Based on the pragmatic language difficulties of people with psychosis we hypothesized that they may be less likely to derive these pragmatic SIs. We conducted two studies in which the ability of people with psychosis to derive SIs was compared to that of healthy controls matched for age and educational level. In the second study we additionally explored the possible link between the capacity to derive SIs and theory of mind (ToM) ability. In general, people with psychosis were less likely to derive SIs than controls. However, the patient group was not homogeneous: half had problems deriving SIs, the other half did not. This dichotomization seems linked to ToM ability because in the patient group, better ToM was associated with a higher ability to derive SIs. Based on the nature of the stimuli used in the SI-task we speculate that this link may not be a direct but an indirect one. © 2017 Published by Elsevier B.V.
1. Introduction Successful social interactions rely heavily on one's ability to go beyond the explicit, literal content of conversational statements and grasp the actual, intended meaning for in daily communication, the message that one wants to express is often not explicitly mentioned. For decades, researchers have illustrated the difficulties patients diagnosed with psychosis experience when they have to decode the non-literal content of conversational statements. These difficulties include trouble grasping the figurative meaning of proverbs and metaphors and problems with understanding humor and irony (e.g., Bambini et al., 2016; Brüne and Bodenstein, 2005; Sponheim et al., 2003). Langdon et al. (2002b) showed for instance that patients with psychosis made more errors than controls in a story comprehension task, when the speech was metaphorical or ironic. More broadly, Linscott (2005) demonstrated that such patients were less compliant with Gricean conversational rules. A broad assessment of the pragmatic skills of patients with psychosis, using the Assessment Battery of Communication, showed that the impairments in patients with psychosis is extended to many domains of communicative skills (Colle et al., 2013). It's important to note that even when syntax and semantics are more or less ⁎ Corresponding author at: Leuvensesteenweg 517, 3070 Kortenberg, Belgium. E-mail address: [email protected] (M. Wampers).
intact, people with psychosis sometimes still show difficulties at the discourse level (e.g., Andreasen et al., 1985; Frith and Allen, 1988). Although pragmatic language has been studied extensively in the context of psychosis, one of the most studied types of pragmatic inferences namely scalar implicatures (SIs) (Guasti et al., 2005; Papafragou and Musolino, 2003), to the best of our knowledge, have not yet been studied in people with psychosis. In the present study, we therefore aim to gain insight in the ability of people with psychosis to derive SIs. We believe that the study of scalar stimuli in the context of psychosis has the potential to advance our understanding of the mechanisms underlying pragmatic language deficit in patients with psychosis. Scalar stimuli allow a high level of control over both stimulus content and context, which may be problematic in more complex language stimuli like stories or in stimuli with fixed content like proverbs. This high level of manipulability of scalar stimuli creates new ways to study the influence of minor stimulus changes on pragmatic language comprehension. SIs are based on linguistic expressions like some, or, must etc. Such expressions are part of a scale of informativeness that is ordered from less informative to more informative. Examples of such scales are: b Some/many/allN, b May/mustN, b Sometimes/often/alwaysN. The statement (1) Some patients were attentive will be generally interpreted as (2) Some but not all patients were attentive
http://dx.doi.org/10.1016/j.schres.2017.08.053 0920-9964/© 2017 Published by Elsevier B.V.
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
2
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
And not as (3) All patients were attentive.
However, on a strictly semantic level “some” means “some and possibly all”. The (implicit) addition of “but not all” does not follow logically but is the result of a SI. These SIs occur because interlocutors regard human conversation as a cooperative process, guided by a number of communicative principles or maxims (Grice, 1975). One of these maxims, the maxim of quantity, requires utterances to provide no more and no less information than is necessary for the purpose and clarity of the conversation. This maxim underlies the SI above. A sentence like (1) might be considered underinformative: it's not clear whether or not all patients were attentive. The speaker thus seems to violate the maxim of quantity. But since the receiver/listener assumes the speaker to be cooperative and to obey the maxim of quantity, he infers that the speaker used the weaker “some” because the stronger “all” was not applicable. He therefore derives a SI and concludes that “some but not all” patients were attentive. According to Nieuwland et al. (2010), people with pragmatic language difficulties, would be less sensitive to violations of the conversational maxims and hence less likely to derive pragmatic inferences including SIs. Based on data about pragmatic language difficulties in people with psychosis, we hypothesize that people with psychosis will be less prone to derive SIs. However, we don't want to make too strong a claim, because not all work on scalar expressions in clinical populations points in this direction. For instance, Chevallier et al. (2010) reported no significant differences in the number of pragmatic responses on the disjunction (interpreting “or” as “A or B, but not both”, instead of the logical “A or B or both) between children with and without autism spectrum disorders (ASD). In the present study, we test how people with psychosis respond to underinformative statements containing “some”. We expect patients with psychosis to have problems deriving SIs: they will respond less pragmatically when confronted with the scalar expression “some” than controls.
2. Experiment 1 2.1. Method 2.1.1. Participants Two groups of participants took part in the study. The patient group consisted of 25 adults diagnosed with schizophrenia according to DSMIV by an experienced psychiatrist. All patients were outpatients. The second group, the control group, was matched to the patient group with respect to age and educational level (see Table 1). All participants were of Dutch literacy and provided written informed consent.
Table 1 Characteristics of the two groups in Experiment 1.
Age (in years) Education level Primary school Lower secondary school Higher secondary school Higher education University Gender male
Patient group (N = 25)
Control group (N = 25)
Mean (STD)
Mean (STD)
49.29 (8.97) % 8 12 48 24 8 52
50.93 (9.72) % 0 16 40 36 8 36
2.1.2. Procedure and study material We used a binary statement-evaluation-task. Each participant received one of two equivalent stimulus sets each of which contained 20 statements. Half of the statements were underinformative through the use of “some” and could lead to SIs. These statements always took the form “some bexemplar N are b category N”, e.g. “Some oaks are trees” (Bott and Noveck, 2004; De Neys and Schaeken, 2007). Participants were asked to judge them as either true (logical) or false (pragmatic). Participants also evaluated 10 filler items (containing “some” or “all”). They were either clearly true (4 items, e.g. “All parrots are birds”) or clearly false (6 items e.g. “Some poplars are fish”) and allowed us to verify whether participants were attentive and able to correctly perform the task. The 20 statements were bundled in random order in booklets that displayed one sentence per page. 2.1.3. Statistical analysis To test our hypothesis, we compared the number of pragmatic answers between both groups with a one-sided Mann-Whitney-U test. A two-sided Mann-Whitney-U test was used to compare both groups with respect to the number of correctly answered filler-items. Withingroup comparisons were performed with a Wilcoxon signed-rank-test. Additionally, we verified whether participants remained consistent in their response to the underinformative items, i.e. whether they stuck to either pragmatic or logical answers. Participants who chose the same response for at least 7 of the 10 target statements were considered consistent. They were then classified as either consistently pragmatic (at least 7 pragmatic answers) or consistently logical (at least 7 logical answers). A chi-square test of independence was used to test whether the prevalence of these categories differed between study groups. 2.2. Results In the control group, participants gave an average of 7.2 (STD = 4.1, Median = 10) pragmatic responses. This number was lower in the patient group in which an average of 4.6 (STD = 4.7, Median = 1) pragmatic responses was given. This difference was marginally significant (Z = − 1.58, p = 0.0568). Patients thus tend to derive fewer SIs than controls. Overall, participants were consistent in their response to underinformative statements (Table 2) for all of them chose the same response alternative for at least 7 of the 10 target items. However there was a clear difference between patient and control group (χ2(1) = 5.33, p = 0.021). The control group was rather homogeneous since the vast majority of control subjects (76%) were classified as consistently pragmatic. The patient group on the other hand could be divided into two groups of comparable size with approximately half of them (56%) classified as consistently logical and the other half (44%) as consistently pragmatic. The results of the within groups comparisons are in line with those of the consistency analysis. In the patient group, pragmatic (45.6%) and logical (54.4%) answers occurred with comparable frequency whereas pragmatic responses (72.4%) were significantly more frequent than logical ones (27.6%; Z = −2.12, p = 0.017) in the control group. Participants made few mistakes in responding to the filler items with the patients even making fewer errors (2%) than the controls (6%; Z = 2.48, p = 0.014). Table 2 Number (%) of patients and controls with a consistently pragmatic or logical response pattern. Group
Consistently pragmatic
Consistently logical
Patients Controls
11 (44%) 19 (76%)
14 (56%) 6 (24%)
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
2.3. Discussion On average, patients derived fewer SIs than controls, which is in line with our hypothesis and this difference was marginally significant. Moreover, the number of participants that consistently derives SIs, is significantly higher in the control group than in the patient group. These differences are not due to differences in the ability to perform the task since both patients and controls attain high levels of accuracy on the filler items. Rather, these findings suggest that patients with schizophrenia are less likely to derive SIs. However, a decreased pragmatic ability to derive SIs in patients with psychosis is not the only possible explanation. Alternatively, the lower number of SIs derived in the patient group may be a consequence of their higher tolerance for underinformativeness. According to Katsos and Bishop (2011), underinformative statements are true but suboptimal. When people are confronted with underinformative statements in the context of a binary judgement task, they cannot express their sensitivity to underinformativeness because they are forced to choose between “true” and “false”. They will then consider underinformative statements as either true or false depending on whether they focus respectively on the fact that these statements are true or on the fact that they are suboptimal. If patients with psychosis are more tolerant for underinformativeness than controls, they will, more often than controls, respond logically to an underinformative statement. Testing the validity of this alternative explanation is the purpose of our next experiment. We replaced the binary statement-evaluationtask of Experiment 1 by a ternary statement-evaluation-task that includes a third response option so that participants can judge an underinformative statement as “true”, “false” or “both true and false”. If an increased tolerance for underinformativeness explains the results of the first experiment, we expect that all participants will prefer the response that allows them to indicate that an underinformative statement is actually perceived as underinformative. In that case the response patterns of patients and controls will be comparable. If on the other hand, the findings of Experiment 1 reflect a decreased ability to derive SIs in patients with psychosis, the use of a ternary statement-evaluationtask will not alter the response pattern observed in Experiment 1. We hypothesize that patients will give fewer pragmatic answers than control subjects despite the adaptation of the statement-evaluation-task. A concept that is often addressed in the context of pragmatic language is Theory of Mind (ToM). ToM refers to a person's ability to infer mental states of the self and others. Intact ToM is necessary for successful communication (Martin and McDonald, 2003). After all, the ability to understand non-literal language requires that one comprehends what are a speaker's feelings and intentions and what the speaker thinks the hearer knows. In other words, pragmatic language skills seem closely intertwined with ToM abilities. There exists a broad literature on the presence of ToM deficits in several clinical populations with manifest pragmatic impairments (for an overview, see e.g., Cummings, 2017). More specifically, numerous studies illustrated ToM impairments in psychosis (Bora et al., 2009a, 2009b; Brüne and Bodenstein, 2005; Harrington et al., 2005; Sprong et al., 2007) and demonstrated an association of ToM-deficiencies with pragmatic language deficits in this population (Champagne-Lavau and Stip, 2010; Langdon et al., 2002b; Mazza et al., 2008). The association between ToM-deficits and pragmatic language difficulties is however far from straightforward as for example Langdon et al. (Langdon et al., 2002a) showed that impaired ToM was associated with poor understanding of irony but not with poor understanding of metaphors. Also the previously mentioned literature showing that people with ASD have no real problems interpreting SIs, suggests that the derivation of SIs not necessarily depends on reasoning about mental states (Chevallier et al., 2010; Pijnacker et al., 2009). Given the central role allocated to ToM in the context of pragmatic language, it is of interest to examine whether ToM-ability and the capacity to derive SIs are related. Our statement-evaluation-task does
3
not seem to invoke ToM-abilities because the statements are not embedded in a conversational context and contain no reference to an interlocutor whose beliefs, intentions or desires might be relevant. Therefore, we included a ToM-task in addition to the statement-evaluation-task. 3. Experiment 2 3.1. Method 3.1.1. Participants The patient group consisted of 17 young psychotic patients between 16 and 31 years old, who were diagnosed with psychosis according to DSM-IV by an experienced psychiatrist. The large majority of patients were male and all but two patients were hospitalized. The control group was matched to the patient group based on age and educational level (Table 3). 3.1.2. Study material We used the same two stimulus sets as in Experiment 1 but now participants had to judge the truthfulness of the statements on a 3point scale with anchor points: “true” (logical), “both true and false” (neutral), and “false” (pragmatic). In addition, we assessed participants' ToM with the Dutch version of the Faux-Pas-test. We opted for the Faux-pas test, because it was explicitly developed for adults (Stone et al., 1998) (unlike the Strange Stories test, which was developed originally for children). Furthermore, participants have to evaluate vignettes with social scenarios that look like daily-life situations, which leads to an ecologically valid test (Chaytor et al., 2006). Participants have to detect violations of social rules, “faux-pas”, in 9 short stories, 5 of which contain a “faux-pas” and 4 don't (Spek, A., Faux-Pas Test). Two things are necessary to detect such a faux-pas: First, an understanding that one person has knowledge that the other person is unaware of (or that one person has a mistaken belief), and an empathic understanding about what someone would find hurtful or offensive. For each story the participant is asked to indicate whether one of the characters in the story has said or done something that he/she had better not said or done. If the participant's answer is affirmative, 5 additional question are asked to enquire about the character's motives, feelings etc. If on the other hand, the participant's answer is negative, no additional questions are asked. Two easy, general control questions are asked about each story to verify whether the participant registered and remembers the necessary story details. 3.1.3. Scoring and statistical analysis The total faux-pas-score was obtained as follows. Each correctly answered control question is 1 point. If a story contains a faux-pas that is correctly detected by the participant, the participant receives one point and he receives 1 point for each correctly answered additional question. If the faux-pas is not detected, the participant receives 0 points. If participants indicate correctly that a story does not contain a faux-pas, they receive 2 points, if they incorrectly detect a faux-pas, they receive 0 points. The scores obtained for each story are summed Table 3 Characteristics of the two participant groups in Experiment 2.
Age (in years) Education level Lower secondary school Higher secondary school Higher education University Gender male
Patient group
Control group
Mean (STD)
Mean (STD)
23.4 (3.3) % 23.5 64.7 5.88 5.88 76.5
23.0 (2.6) % 23.5 64.7 5.88 5.88 82.4
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
4
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
to produce a total score that varies between 0 and 56, where a higher score is indicative of a better ToM. For the analysis of the pragmatic, logical and neutral answers, similar tests were used as in Experiment 1. A Pearson correlation coefficient was computed to study the association between ToM-ability as measured by the Faux-Pas-test and the number of logical responses to underinformative statements. 3.2. Results In the patients group, the number of logical answers was 4.9 (STD = 4.6, median = 6) on average compared to 1.7 (STD = 3.3, median = 0) in the control group (see Table 4). Patients gave significantly more logical answers in response to underinformative sentences than controls (Z = −1.67, p = 0.048). Both groups did not differ significantly with respect to the number of pragmatic (patients: mean = 2.8, STD = 3.8; controls: mean = 5.4, STD = 4.7) and neutral (patients: mean = 2.2, STD = 3.7; Controls: mean = 2.8, STD = 4.3) answers. There also was no significant difference between patients and controls in the number of incorrectly answered filler items (4.7% and 7.1% respectively). The ToM-ability of control participants was significantly higher than that of patients (Z = 3.56, p = 0.0004). Within the patient group, a significant negative correlation between the number of logical responses and the score on the Faux-Pas-test was observed (r = − 0.66, p = 0.0041). In the control group, no such correlation was found (r = 0.03, p = 0.90). The association between ToM-ability and the number of logical answers observed in the patient group was further explored, by dividing each participant group in two subgroups based on a median split of the faux-pas-scores. The different groups were then compared with respect to the number of logical, pragmatic and neutral responses with Mann-Whitney-U tests. Results revealed significant differences between the high- and lowToM patient groups. Patients of the low-ToM group gave significantly more logical answers than patients from the high-ToM group (Z = −2.35, p = 0.019) who gave significantly more neutral answers (Z = 2.19, p = 0.028). When comparing the low-ToM patient group with the low- and high-ToM control groups, it turned out that low-ToM patients gave significantly more logical answers than both low-ToM (Z = − 2.51, p = 0.012) and high-ToM controls (Z = −2.45, p = 0.014). Low-ToM controls gave significantly more pragmatic answers than low-ToM patients (Z = 1.97, p = 0.024). No further significant differences were observed neither between both control groups, nor between the high-ToM patient group and any of the control groups. 3.3. Discussion Switching from a binary to a ternary statement-evaluation-task did not fundamentally alter the result pattern observed in Experiment 1. In Experiment 2, patients with psychosis still gave significantly more
logical answers and thus derived fewer SIs than controls. The results of Experiment 1 apparently cannot be attributed to an increased tolerance for underinformativeness in the patient group. Consequently, the apparent dichotomy in the patient group revealed in the consistency analysis of Experiment 1 cannot be considered a by-product of a higher tolerance for underinformativeness in the patient group. This dichotomy could however be linked to differences in ToM-ability as is suggested by the negative correlation between number of logical interpretations of underinformative statements and ToM-ability that was observed in the patient group. ToM, as measured by the Faux-Pastest, may thus play a role in the interpretation of underinformative statements and the restoration of violations of the conversational maxims. This is consistent with the conclusions of other researchers who associated impaired understanding of non-literal language with defective ToM (Brüne and Bodenstein, 2005; Corcoran et al., 1997; Corcoran et al., 1995; Janssen et al., 2003; Langdon et al., 2002a). That patients with a low ToM-score gave significantly more logical answers than patients with high ToM and controls, suggests that a certain level of ToM-ability might be required to derive SIs. The observed differences between patients and controls may thus be related to differences in ToM-ability. That a correlation between ToM-ability and number of logical answers was not found in the control group is in line with the findings of other investigators (Mazza et al., 2008), and is probably due to a ceiling effect in the control group (Brüne and Bodenstein, 2005). Nevertheless, some caution is certainly required. We treated ToM as a unitary concept measured by a rather coarse total score while ToM is actually rather complex and comprises different aspects (Freeman, 2007; Frith, 2004; Montag et al., 2011; Shamay-Tsoory et al., 2007), which may each be linked to one or more aspects of pragmatic language.
4. General discussion and conclusion To the best of our knowledge, the present study is the first one to address SIs in the context of psychosis. Based on the literature about pragmatic language difficulties in patients with psychosis, we hypothesized that these patients would be less likely to derive SIs than controls. Our results confirmed this hypothesis: patients with psychosis derived fewer SIs than matched control subjects. In general, patients seem to prefer the logical interpretation of semantic expressions like “some” to the pragmatic one that is preferred by controls. In Experiment 2 the effect was statistically significant, in Experiment 1 almost (p = 0.0568). To make our evidence more compelling, we decided to conduct an additional analysis on the data of experiment 1 and 2 together. Therefore, we compared the number of logical responses of Experiment 1 and 2 combined between the patient group and the controls. We used a simple Bayesian t-test to quantify the strength of the evidence against the null hypothesis. In the patients group, the number of logical answers was 5.2 (STD = 4.6), in the control group 2.3 (STD = 3.8). The Bayes Factor = 28.66, which is considered to be strong evidence and it remains strong, also after a robustness check. Hence, subtle differences
Table 4 Percentage false, true/false and true answers on the underinformative questions, correct answers to the filler items and the faux-pas score of the patient and the control group. Type of response
Patient group
Control group
Proportion (N) Underinformative utterances
False True as well as false True
28.2 (48) 22.3 (38) 50.6 (84) 90.6 (154)
Correct filler items Patient group
54.1 (92) 28.8 (49) 17.1 (29) 95.3 (162) Control group
Median (Mean, STD) Faux-pas score (ToM)
39 (37, 9.10)
48 (48.3, 4.74)
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
in the interpretation of single words can apparently lead to different conclusions, which may influence daily communication. The ability to derive SIs seems linked to ToM-ability for we observed a significant correlation between both abilities in the patient group of Experiment 2. Whether the observed association is a causal one and if so, what the direction of this causality may be, remains under discussion (Martin and McDonald, 2003). Some evidence points to ToM as a necessary precursor of pragmatic language skills (Happé, 1993; Winner et al., 1998) whereas some authors argue that it is the experience with pragmatic language that leads to the development of proficient ToM (Peterson and Siegal, 2000). However, the statement-evaluation-task we used does not seem to rely heavily on ToM-ability because there's no reference to an interlocutor whose intention/emotions may be of relevance. One could therefore wonder whether the observed correlation might be an “artificial” one, i.e. the result of an underlying core cognitive process that determines both ToM-ability and the ability to interpret underinformative utterances (UUs) in a pragmatic way (e.g., working memory, weak central coherence, formal thought disorder). Although these processes have all been studied to some extent with respect to ToM and pragmatic language problems in psychosis, they remain largely unexplored when it comes to pragmatic inferences such as SIs. Our results indicate that this is a gap that is worth looking into. In the present study we explored how psychotic patients deal with UUs and which factors could influence this process. Our results indicated that psychotic patients have problems deriving SIs and that this could be linked to an impaired ToM-ability. Some study characteristics should however be kept in mind. In the present study we considered only one scalar expression (“some”). Obviously the generalizability of our findings to other scalars needs to be explored. Furthermore, the UUs were presented in isolation, without context. Context may influence the interpretation of scalar expressions, so it would be interesting to examine whether the conversational context influences the way people with psychosis deal with underinformativeness. Also, we only studied the link between ToM and SIs but, as discussed above, other processes may be relevant and future research should look into this. Additionally, as discussed above as well, we treated ToM as a unitary concept Approaching ToM in a multidimensional way may be fruitful to gain more insight in its role in pragmatic language in general and in SIs in particular. A last, intriguing issue that needs to be considered, concerns a potential difference between the two participant groups with respect to their experience with standardized tests. The patients probably have more experience with standardized tests and therefore may have tackled the implicature task somewhat differently than controls. Patients may have been more inclined to interpret our experimental task as, for instance an intelligence test, a test of logical reasoning. With this mindset, they might have focused more on whether the sentences were logically correct, and less on spontaneously interpreting each sentence. Papafragou and Musolino (2003) nicely demonstrated that how an implicature task is framed or interpreted strongly influences whether implicatures are drawn or not. They showed that children's ability to derive scalar implicatures is affected by their awareness of the goal of the task. Typically, young children do not reject underinformative sentences in a statement-evaluation-task. However, some pre-training and more specific instructions emphasizing the importance to detect pragmatic infelicity lead to an increase in the rejection rates of children. Furthermore, Skordos and Papafragou (2016) showed that when the alternatives are made more accessible and relevant, children also reject underinformative sentences more often. This may be of relevance for our study. Although the implicature task was presented as a language comprehension task which focused on interindividual differences in sentence comprehension, we cannot rule out that some patients, given their experience with standardized tests, unlike controls, may have considered the stronger alternatives (e.g., “all oaks are trees” for “some oaks are trees”) as not relevant here and therefore decided not
5
to reject the underinformative sentences. A manipulation of the accessibility and relevance of the response alternatives is an interesting topic for future research. Contributors Martien Wampers and Walter Schaeken designed the study and wrote the study protocol. Martien Wampers, Sofie Schrauwen, Marc De Hert and Leen Gielen were responsible for the recruitment and testing of participants. Martien Wampers, and Sofie Schrauwen performed the statistical analyses. Martien Wampers wrote the first draft of the manuscript and all other authors contributed to and have approved the final manuscript.
Funding body agreements and policies Not applicable. Conflict of interest None of the authors has any conflict of interest to disclose.
Acknowledgements We would like to thank the entire staff of Sint-Joris, the ward for young people with psychosis of the University Psychiatric Centre KU Leuven, for their support during the data collection process.
References Andreasen, N.C., Hoffman, R.E., Grove, W.M., 1985. Mapping abnormalities in language and cognition. In: Alpert, A. (Ed.), Controversies in Schizophrenia. Guilford Press, New York, NY, pp. 199–226. Bambini, V., Arcara, G., Bechi, M., Buonocore, M., Cavallaro, R., Bosia, M., 2016. The communicative impairment as a core feature of schizophrenia: frequency of pragmatic deficit, cognitive substrates, and relation with quality of life. Compr. Psychiatry 71, 106–120. Bora, E., Yucel, M., Pantelis, C., 2009a. Theory of mind impairment in schizophrenia: metaanalysis. Schizophr. Res. 109 (1–3), 1–9. Bora, E., Yücel, M., Pantelis, C., 2009b. Theory of mind impairment: a distinct trait-marker for schizophrenia spectrum disorders and bipolar disorder? Acta Psychiatr. Scand. 120 (4), 253–264. Bott, L., Noveck, I.A., 2004. Some utterances are underinformative: the onset and time course of scalar inferences. J. Mem. Lang. 51 (3), 437–457. Brüne, M., Bodenstein, L., 2005. Proverb comprehension reconsidered - “theory of mind” and the pragmatic use of language in schizophrenia. Schizophr. Res. 75 (2–3), 233–239. Champagne-Lavau, M., Stip, E., 2010. Pragmatic and executive dysfunction in schizophrenia. J. Neurolinguistics 23 (3), 285–296. Chaytor, N., Schimitter-Edgecombe, M., Burr, R., 2006. Improving the ecological validity of executive functioning assessment. Arch. Clin. Neuropsychol. 21 (3), 217–227. Chevallier, C., Wilson, D., Happé, F., Noveck, I., 2010. Scalar inferences in autism spectrum disorders. J. Autism Dev. Disord. 40 (9), 1104–1117. Colle, L., Angeleri, R., Vallana, M., Sacco, K., Bara, B.G., Bosco, F.M., 2013. Understanding the communicative impairments in schizophrenia: a preliminary study. J. Commun. Disord. 46 (3), 294–308. Corcoran, R., Mercer, G., Frith, C.D., 1995. Schizophrenia, symptomatology and social inference: investigating “theory of mind” in people with schizophrenia. Schizophr. Res. 17 (1), 5–13. Corcoran, R., Cahill, C., Frith, C.D., 1997. The appreciation of visual jokes in people with schizophrenia: a study of “mentalizing” ability. Schizophr. Res. 24 (3), 319–327. Cummings, L., 2017. Cognitive aspects of pragmatic disorders. In: Cummings, L. (Ed.), Research in Clinical Pragmatics. Springer, Cham, Switzerland, pp. 587–616. De Neys, W., Schaeken, W., 2007. When people are more logical under cognitive load dual task impact on scalar implicature. Exp. Psychol. 54 (2), 128–133. Freeman, D., 2007. Suspicious minds: the psychology of persecutory delusions. Clin. Psychol. Rev. 27 (4), 425–457. Frith, C.D., 2004. Schizophrenia and theory of mind. Psychol. Med. 34 (3), 385–389. Frith, C.D., Allen, H.A., 1988. Language disorders in schizophrenia and their implications for neuropsychology. In: Bebbington, P., McGuffin, P. (Eds.), Schizophrenia: The Major Issues. Heinemann Professional Pub, Oxford, UK, pp. 172–186. Grice, H.P., 1975. Logic and conversation. In: Cole, P., Morgan, J.L. (Eds.), Syntax and Semantics (Vol. 3): Speech Acts. Academia Press, New York, pp. 41–58. Guasti, T., Chierchia, G., Crain, S., Foppolo, F., Gualmini, A., Meroni, L., 2005. Why children and adults sometimes (but not always) compute implicatures. Lang. Cogn. Process. 20 (5), 667–696. Happé, F.G.E., 1993. Communicative competence and theory of mind in autism: a test of relevance theory. Cognition 48 (2), 101–119. Harrington, L., Siegert, R.J., McClure, J., 2005. Theory of mind in schizophrenia: a critical review. Cogn. Neuropsychiatry 10 (4), 249–286. Janssen, I., Krabbendam, L., Jolles, J., van Os, J., 2003. Alterations in theory of mind in patients with schizophrenia and non-psychotic relatives. Acta Psychiatr. Scand. 108 (2), 110–117. Katsos, N., Bishop, D.V.M., 2011. Pragmatic tolerance: implications for the acquisition of informativeness and implicature. Cognition 120 (1), 67–81.
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
6
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
Langdon, R., Coltheart, M., Ward, P.B., Catts, S.V., 2002a. Disturbed communication in schizophrenia: the role of poor pragmatics and poor mind-reading. Psychol. Med. 32 (7), 1273–1284. Langdon, R., Davies, M., Coltheart, M., 2002b. Understanding minds and understanding communicated meanings in schizophrenia. Mind Lang. 17 (1–2), 68–104. Linscott, R.J., 2005. Thought disorder, pragmatic language impairment, and generalized cognitive decline in schizophrenia. Schizophr. Res. 75 (2–3), 225–232. Martin, I., McDonald, S., 2003. Weak coherence, no theory of mind, or executive dysfunction? Solving the puzzle of pragmatic language disorders. Brain Lang. 85 (3), 451–466. Mazza, M., Di Michele, V., Pollice, R., Casacchia, M., Roncone, R., 2008. Pragmatic language and theory of mind deficits in people with schizophrenia and their relatives. Psychopathology 41 (4), 254–263. Montag, C., Dziobek, I., Richter, I.S., Neuhaus, K., Lehmann, A., Sylla, R., ... Gallinat, J., 2011. Different aspects of theory of mind in paranoid schizophrenia: evidence from a video-based assessment. Psychiatry Res. 186 (2–3), 203–209. Nieuwland, M.S., Ditman, T., Kuperberg, G.R., 2010. On the incrementality of pragmatic processing: an ERP investigation of informativeness and pragmatic abilities. J. Mem. Lang. 63 (3), 324–346. Papafragou, A., Musolino, J., 2003. Scalar implicatures: experiments at the semantics – pragmatics interface. Cognition 86 (3), 253–282. Peterson, C.C., Siegal, M., 2000. Insights into theory of mind from deafness and autism. Mind Lang. 15 (1), 123–145.
Pijnacker, J., Hagoort, P., Buitelaar, J., Teunissen, J.P., Geurts, B., 2009. Pragmatic inferences in high-functioning adults with autism and Asperger syndrome. J. Autism Dev. Disord. 39 (4), 607–618. Shamay-Tsoory, S.G., Shur, S., Barcai-Goodman, L., Medlovich, S., Harari, H., Levkovitz, Y., 2007. Dissociation of cognitive from affective components of theory of mind in schizophrenia. Psychiatry Res. 149 (1–3), 11–23. Skordos, D., Papafragou, A., 2016. Children's derivation of scalar implicatures: alternatives and relevance. Cognition 153, 6–18. Sponheim, S.R., Surerus-Johnson, C., Leskela, J., Dieperink, M.E., 2003. Proverb interpretation in schizophrenia: the significance of symptomatology and cognitive processes. Schizophr. Res. 65 (2–3), 117–123. Sprong, M., Schothorst, P., Vos, E., Hox, J., van Engeland, H., 2007. Theory of mind in schizophrenia: meta-analysis. Br. J. Psychiatry 191 (1), 5–13. Stone, V.E., Baron-Cohen, S., Knight, R.T., 1998. Frontal lobe contributions to theory of mind. J. Cogn. Neurosci. 10 (5), 640–656. Winner, E., Brownell, H., Happé, F., Blum, A., Pincus, D., 1998. Distinguishing lies from jokes: theory of mind deficits and discourse interpretation in right hemisphere brain-damaged patients. Brain Lang. 62 (1), 89–106.
Web references Spek, A., d. Faux Pas Test. Retrieved from. http://www.anneliesspek.nl/pagina27a.html.
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
Contents lists available at ScienceDirect
Schizophrenia Research journal homepage: www.elsevier.com/locate/schres
Patients with psychosis struggle with scalar implicatures Martien Wampers a,b,⁎, Sofie Schrauwen c, Marc De Hert a,b, Leen Gielen a, Walter Schaeken c a b c
University Psychiatric Hospital KU Leuven, Leuvensesteenweg 517, 3070 Kortenberg, Belgium University of Leuven, Faculty of Medicines, Department of Neurosciences, Herestraat 49, Box 721, 3000 Leuven, Belgium University of Leuven, Faculty of Psychology and Educational Sciences, Laboratory of Experimental Psychology, Tiensestraat 101, Leuven, Belgium
a r t i c l e
i n f o
Article history: Received 27 March 2017 Received in revised form 6 July 2017 Accepted 28 August 2017 Available online xxxx Keywords: Psychosis Language Pragmatics Reasoning Scalar implicatures
a b s t r a c t Pragmatic language difficulties in people with psychosis have been demonstrated repeatedly but one of the most studied types of pragmatic language, i.e. scalar implicatures (SIs), has not yet been examined in this population. SIs are a special kind of pragmatic inferences, based on linguistic expressions like some, or, must. Such expressions are part of a scale of informativeness organized by informativity (e.g. some/many/all). Although semantically the less informative expressions imply the more informative ones, pragmatically people generally infer that the use of a less informative expression implies that the more informative option is not applicable. Based on the pragmatic language difficulties of people with psychosis we hypothesized that they may be less likely to derive these pragmatic SIs. We conducted two studies in which the ability of people with psychosis to derive SIs was compared to that of healthy controls matched for age and educational level. In the second study we additionally explored the possible link between the capacity to derive SIs and theory of mind (ToM) ability. In general, people with psychosis were less likely to derive SIs than controls. However, the patient group was not homogeneous: half had problems deriving SIs, the other half did not. This dichotomization seems linked to ToM ability because in the patient group, better ToM was associated with a higher ability to derive SIs. Based on the nature of the stimuli used in the SI-task we speculate that this link may not be a direct but an indirect one. © 2017 Published by Elsevier B.V.
1. Introduction Successful social interactions rely heavily on one's ability to go beyond the explicit, literal content of conversational statements and grasp the actual, intended meaning for in daily communication, the message that one wants to express is often not explicitly mentioned. For decades, researchers have illustrated the difficulties patients diagnosed with psychosis experience when they have to decode the non-literal content of conversational statements. These difficulties include trouble grasping the figurative meaning of proverbs and metaphors and problems with understanding humor and irony (e.g., Bambini et al., 2016; Brüne and Bodenstein, 2005; Sponheim et al., 2003). Langdon et al. (2002b) showed for instance that patients with psychosis made more errors than controls in a story comprehension task, when the speech was metaphorical or ironic. More broadly, Linscott (2005) demonstrated that such patients were less compliant with Gricean conversational rules. A broad assessment of the pragmatic skills of patients with psychosis, using the Assessment Battery of Communication, showed that the impairments in patients with psychosis is extended to many domains of communicative skills (Colle et al., 2013). It's important to note that even when syntax and semantics are more or less ⁎ Corresponding author at: Leuvensesteenweg 517, 3070 Kortenberg, Belgium. E-mail address: [email protected] (M. Wampers).
intact, people with psychosis sometimes still show difficulties at the discourse level (e.g., Andreasen et al., 1985; Frith and Allen, 1988). Although pragmatic language has been studied extensively in the context of psychosis, one of the most studied types of pragmatic inferences namely scalar implicatures (SIs) (Guasti et al., 2005; Papafragou and Musolino, 2003), to the best of our knowledge, have not yet been studied in people with psychosis. In the present study, we therefore aim to gain insight in the ability of people with psychosis to derive SIs. We believe that the study of scalar stimuli in the context of psychosis has the potential to advance our understanding of the mechanisms underlying pragmatic language deficit in patients with psychosis. Scalar stimuli allow a high level of control over both stimulus content and context, which may be problematic in more complex language stimuli like stories or in stimuli with fixed content like proverbs. This high level of manipulability of scalar stimuli creates new ways to study the influence of minor stimulus changes on pragmatic language comprehension. SIs are based on linguistic expressions like some, or, must etc. Such expressions are part of a scale of informativeness that is ordered from less informative to more informative. Examples of such scales are: b Some/many/allN, b May/mustN, b Sometimes/often/alwaysN. The statement (1) Some patients were attentive will be generally interpreted as (2) Some but not all patients were attentive
http://dx.doi.org/10.1016/j.schres.2017.08.053 0920-9964/© 2017 Published by Elsevier B.V.
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
2
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
And not as (3) All patients were attentive.
However, on a strictly semantic level “some” means “some and possibly all”. The (implicit) addition of “but not all” does not follow logically but is the result of a SI. These SIs occur because interlocutors regard human conversation as a cooperative process, guided by a number of communicative principles or maxims (Grice, 1975). One of these maxims, the maxim of quantity, requires utterances to provide no more and no less information than is necessary for the purpose and clarity of the conversation. This maxim underlies the SI above. A sentence like (1) might be considered underinformative: it's not clear whether or not all patients were attentive. The speaker thus seems to violate the maxim of quantity. But since the receiver/listener assumes the speaker to be cooperative and to obey the maxim of quantity, he infers that the speaker used the weaker “some” because the stronger “all” was not applicable. He therefore derives a SI and concludes that “some but not all” patients were attentive. According to Nieuwland et al. (2010), people with pragmatic language difficulties, would be less sensitive to violations of the conversational maxims and hence less likely to derive pragmatic inferences including SIs. Based on data about pragmatic language difficulties in people with psychosis, we hypothesize that people with psychosis will be less prone to derive SIs. However, we don't want to make too strong a claim, because not all work on scalar expressions in clinical populations points in this direction. For instance, Chevallier et al. (2010) reported no significant differences in the number of pragmatic responses on the disjunction (interpreting “or” as “A or B, but not both”, instead of the logical “A or B or both) between children with and without autism spectrum disorders (ASD). In the present study, we test how people with psychosis respond to underinformative statements containing “some”. We expect patients with psychosis to have problems deriving SIs: they will respond less pragmatically when confronted with the scalar expression “some” than controls.
2. Experiment 1 2.1. Method 2.1.1. Participants Two groups of participants took part in the study. The patient group consisted of 25 adults diagnosed with schizophrenia according to DSMIV by an experienced psychiatrist. All patients were outpatients. The second group, the control group, was matched to the patient group with respect to age and educational level (see Table 1). All participants were of Dutch literacy and provided written informed consent.
Table 1 Characteristics of the two groups in Experiment 1.
Age (in years) Education level Primary school Lower secondary school Higher secondary school Higher education University Gender male
Patient group (N = 25)
Control group (N = 25)
Mean (STD)
Mean (STD)
49.29 (8.97) % 8 12 48 24 8 52
50.93 (9.72) % 0 16 40 36 8 36
2.1.2. Procedure and study material We used a binary statement-evaluation-task. Each participant received one of two equivalent stimulus sets each of which contained 20 statements. Half of the statements were underinformative through the use of “some” and could lead to SIs. These statements always took the form “some bexemplar N are b category N”, e.g. “Some oaks are trees” (Bott and Noveck, 2004; De Neys and Schaeken, 2007). Participants were asked to judge them as either true (logical) or false (pragmatic). Participants also evaluated 10 filler items (containing “some” or “all”). They were either clearly true (4 items, e.g. “All parrots are birds”) or clearly false (6 items e.g. “Some poplars are fish”) and allowed us to verify whether participants were attentive and able to correctly perform the task. The 20 statements were bundled in random order in booklets that displayed one sentence per page. 2.1.3. Statistical analysis To test our hypothesis, we compared the number of pragmatic answers between both groups with a one-sided Mann-Whitney-U test. A two-sided Mann-Whitney-U test was used to compare both groups with respect to the number of correctly answered filler-items. Withingroup comparisons were performed with a Wilcoxon signed-rank-test. Additionally, we verified whether participants remained consistent in their response to the underinformative items, i.e. whether they stuck to either pragmatic or logical answers. Participants who chose the same response for at least 7 of the 10 target statements were considered consistent. They were then classified as either consistently pragmatic (at least 7 pragmatic answers) or consistently logical (at least 7 logical answers). A chi-square test of independence was used to test whether the prevalence of these categories differed between study groups. 2.2. Results In the control group, participants gave an average of 7.2 (STD = 4.1, Median = 10) pragmatic responses. This number was lower in the patient group in which an average of 4.6 (STD = 4.7, Median = 1) pragmatic responses was given. This difference was marginally significant (Z = − 1.58, p = 0.0568). Patients thus tend to derive fewer SIs than controls. Overall, participants were consistent in their response to underinformative statements (Table 2) for all of them chose the same response alternative for at least 7 of the 10 target items. However there was a clear difference between patient and control group (χ2(1) = 5.33, p = 0.021). The control group was rather homogeneous since the vast majority of control subjects (76%) were classified as consistently pragmatic. The patient group on the other hand could be divided into two groups of comparable size with approximately half of them (56%) classified as consistently logical and the other half (44%) as consistently pragmatic. The results of the within groups comparisons are in line with those of the consistency analysis. In the patient group, pragmatic (45.6%) and logical (54.4%) answers occurred with comparable frequency whereas pragmatic responses (72.4%) were significantly more frequent than logical ones (27.6%; Z = −2.12, p = 0.017) in the control group. Participants made few mistakes in responding to the filler items with the patients even making fewer errors (2%) than the controls (6%; Z = 2.48, p = 0.014). Table 2 Number (%) of patients and controls with a consistently pragmatic or logical response pattern. Group
Consistently pragmatic
Consistently logical
Patients Controls
11 (44%) 19 (76%)
14 (56%) 6 (24%)
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
2.3. Discussion On average, patients derived fewer SIs than controls, which is in line with our hypothesis and this difference was marginally significant. Moreover, the number of participants that consistently derives SIs, is significantly higher in the control group than in the patient group. These differences are not due to differences in the ability to perform the task since both patients and controls attain high levels of accuracy on the filler items. Rather, these findings suggest that patients with schizophrenia are less likely to derive SIs. However, a decreased pragmatic ability to derive SIs in patients with psychosis is not the only possible explanation. Alternatively, the lower number of SIs derived in the patient group may be a consequence of their higher tolerance for underinformativeness. According to Katsos and Bishop (2011), underinformative statements are true but suboptimal. When people are confronted with underinformative statements in the context of a binary judgement task, they cannot express their sensitivity to underinformativeness because they are forced to choose between “true” and “false”. They will then consider underinformative statements as either true or false depending on whether they focus respectively on the fact that these statements are true or on the fact that they are suboptimal. If patients with psychosis are more tolerant for underinformativeness than controls, they will, more often than controls, respond logically to an underinformative statement. Testing the validity of this alternative explanation is the purpose of our next experiment. We replaced the binary statement-evaluationtask of Experiment 1 by a ternary statement-evaluation-task that includes a third response option so that participants can judge an underinformative statement as “true”, “false” or “both true and false”. If an increased tolerance for underinformativeness explains the results of the first experiment, we expect that all participants will prefer the response that allows them to indicate that an underinformative statement is actually perceived as underinformative. In that case the response patterns of patients and controls will be comparable. If on the other hand, the findings of Experiment 1 reflect a decreased ability to derive SIs in patients with psychosis, the use of a ternary statement-evaluationtask will not alter the response pattern observed in Experiment 1. We hypothesize that patients will give fewer pragmatic answers than control subjects despite the adaptation of the statement-evaluation-task. A concept that is often addressed in the context of pragmatic language is Theory of Mind (ToM). ToM refers to a person's ability to infer mental states of the self and others. Intact ToM is necessary for successful communication (Martin and McDonald, 2003). After all, the ability to understand non-literal language requires that one comprehends what are a speaker's feelings and intentions and what the speaker thinks the hearer knows. In other words, pragmatic language skills seem closely intertwined with ToM abilities. There exists a broad literature on the presence of ToM deficits in several clinical populations with manifest pragmatic impairments (for an overview, see e.g., Cummings, 2017). More specifically, numerous studies illustrated ToM impairments in psychosis (Bora et al., 2009a, 2009b; Brüne and Bodenstein, 2005; Harrington et al., 2005; Sprong et al., 2007) and demonstrated an association of ToM-deficiencies with pragmatic language deficits in this population (Champagne-Lavau and Stip, 2010; Langdon et al., 2002b; Mazza et al., 2008). The association between ToM-deficits and pragmatic language difficulties is however far from straightforward as for example Langdon et al. (Langdon et al., 2002a) showed that impaired ToM was associated with poor understanding of irony but not with poor understanding of metaphors. Also the previously mentioned literature showing that people with ASD have no real problems interpreting SIs, suggests that the derivation of SIs not necessarily depends on reasoning about mental states (Chevallier et al., 2010; Pijnacker et al., 2009). Given the central role allocated to ToM in the context of pragmatic language, it is of interest to examine whether ToM-ability and the capacity to derive SIs are related. Our statement-evaluation-task does
3
not seem to invoke ToM-abilities because the statements are not embedded in a conversational context and contain no reference to an interlocutor whose beliefs, intentions or desires might be relevant. Therefore, we included a ToM-task in addition to the statement-evaluation-task. 3. Experiment 2 3.1. Method 3.1.1. Participants The patient group consisted of 17 young psychotic patients between 16 and 31 years old, who were diagnosed with psychosis according to DSM-IV by an experienced psychiatrist. The large majority of patients were male and all but two patients were hospitalized. The control group was matched to the patient group based on age and educational level (Table 3). 3.1.2. Study material We used the same two stimulus sets as in Experiment 1 but now participants had to judge the truthfulness of the statements on a 3point scale with anchor points: “true” (logical), “both true and false” (neutral), and “false” (pragmatic). In addition, we assessed participants' ToM with the Dutch version of the Faux-Pas-test. We opted for the Faux-pas test, because it was explicitly developed for adults (Stone et al., 1998) (unlike the Strange Stories test, which was developed originally for children). Furthermore, participants have to evaluate vignettes with social scenarios that look like daily-life situations, which leads to an ecologically valid test (Chaytor et al., 2006). Participants have to detect violations of social rules, “faux-pas”, in 9 short stories, 5 of which contain a “faux-pas” and 4 don't (Spek, A., Faux-Pas Test). Two things are necessary to detect such a faux-pas: First, an understanding that one person has knowledge that the other person is unaware of (or that one person has a mistaken belief), and an empathic understanding about what someone would find hurtful or offensive. For each story the participant is asked to indicate whether one of the characters in the story has said or done something that he/she had better not said or done. If the participant's answer is affirmative, 5 additional question are asked to enquire about the character's motives, feelings etc. If on the other hand, the participant's answer is negative, no additional questions are asked. Two easy, general control questions are asked about each story to verify whether the participant registered and remembers the necessary story details. 3.1.3. Scoring and statistical analysis The total faux-pas-score was obtained as follows. Each correctly answered control question is 1 point. If a story contains a faux-pas that is correctly detected by the participant, the participant receives one point and he receives 1 point for each correctly answered additional question. If the faux-pas is not detected, the participant receives 0 points. If participants indicate correctly that a story does not contain a faux-pas, they receive 2 points, if they incorrectly detect a faux-pas, they receive 0 points. The scores obtained for each story are summed Table 3 Characteristics of the two participant groups in Experiment 2.
Age (in years) Education level Lower secondary school Higher secondary school Higher education University Gender male
Patient group
Control group
Mean (STD)
Mean (STD)
23.4 (3.3) % 23.5 64.7 5.88 5.88 76.5
23.0 (2.6) % 23.5 64.7 5.88 5.88 82.4
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
4
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
to produce a total score that varies between 0 and 56, where a higher score is indicative of a better ToM. For the analysis of the pragmatic, logical and neutral answers, similar tests were used as in Experiment 1. A Pearson correlation coefficient was computed to study the association between ToM-ability as measured by the Faux-Pas-test and the number of logical responses to underinformative statements. 3.2. Results In the patients group, the number of logical answers was 4.9 (STD = 4.6, median = 6) on average compared to 1.7 (STD = 3.3, median = 0) in the control group (see Table 4). Patients gave significantly more logical answers in response to underinformative sentences than controls (Z = −1.67, p = 0.048). Both groups did not differ significantly with respect to the number of pragmatic (patients: mean = 2.8, STD = 3.8; controls: mean = 5.4, STD = 4.7) and neutral (patients: mean = 2.2, STD = 3.7; Controls: mean = 2.8, STD = 4.3) answers. There also was no significant difference between patients and controls in the number of incorrectly answered filler items (4.7% and 7.1% respectively). The ToM-ability of control participants was significantly higher than that of patients (Z = 3.56, p = 0.0004). Within the patient group, a significant negative correlation between the number of logical responses and the score on the Faux-Pas-test was observed (r = − 0.66, p = 0.0041). In the control group, no such correlation was found (r = 0.03, p = 0.90). The association between ToM-ability and the number of logical answers observed in the patient group was further explored, by dividing each participant group in two subgroups based on a median split of the faux-pas-scores. The different groups were then compared with respect to the number of logical, pragmatic and neutral responses with Mann-Whitney-U tests. Results revealed significant differences between the high- and lowToM patient groups. Patients of the low-ToM group gave significantly more logical answers than patients from the high-ToM group (Z = −2.35, p = 0.019) who gave significantly more neutral answers (Z = 2.19, p = 0.028). When comparing the low-ToM patient group with the low- and high-ToM control groups, it turned out that low-ToM patients gave significantly more logical answers than both low-ToM (Z = − 2.51, p = 0.012) and high-ToM controls (Z = −2.45, p = 0.014). Low-ToM controls gave significantly more pragmatic answers than low-ToM patients (Z = 1.97, p = 0.024). No further significant differences were observed neither between both control groups, nor between the high-ToM patient group and any of the control groups. 3.3. Discussion Switching from a binary to a ternary statement-evaluation-task did not fundamentally alter the result pattern observed in Experiment 1. In Experiment 2, patients with psychosis still gave significantly more
logical answers and thus derived fewer SIs than controls. The results of Experiment 1 apparently cannot be attributed to an increased tolerance for underinformativeness in the patient group. Consequently, the apparent dichotomy in the patient group revealed in the consistency analysis of Experiment 1 cannot be considered a by-product of a higher tolerance for underinformativeness in the patient group. This dichotomy could however be linked to differences in ToM-ability as is suggested by the negative correlation between number of logical interpretations of underinformative statements and ToM-ability that was observed in the patient group. ToM, as measured by the Faux-Pastest, may thus play a role in the interpretation of underinformative statements and the restoration of violations of the conversational maxims. This is consistent with the conclusions of other researchers who associated impaired understanding of non-literal language with defective ToM (Brüne and Bodenstein, 2005; Corcoran et al., 1997; Corcoran et al., 1995; Janssen et al., 2003; Langdon et al., 2002a). That patients with a low ToM-score gave significantly more logical answers than patients with high ToM and controls, suggests that a certain level of ToM-ability might be required to derive SIs. The observed differences between patients and controls may thus be related to differences in ToM-ability. That a correlation between ToM-ability and number of logical answers was not found in the control group is in line with the findings of other investigators (Mazza et al., 2008), and is probably due to a ceiling effect in the control group (Brüne and Bodenstein, 2005). Nevertheless, some caution is certainly required. We treated ToM as a unitary concept measured by a rather coarse total score while ToM is actually rather complex and comprises different aspects (Freeman, 2007; Frith, 2004; Montag et al., 2011; Shamay-Tsoory et al., 2007), which may each be linked to one or more aspects of pragmatic language.
4. General discussion and conclusion To the best of our knowledge, the present study is the first one to address SIs in the context of psychosis. Based on the literature about pragmatic language difficulties in patients with psychosis, we hypothesized that these patients would be less likely to derive SIs than controls. Our results confirmed this hypothesis: patients with psychosis derived fewer SIs than matched control subjects. In general, patients seem to prefer the logical interpretation of semantic expressions like “some” to the pragmatic one that is preferred by controls. In Experiment 2 the effect was statistically significant, in Experiment 1 almost (p = 0.0568). To make our evidence more compelling, we decided to conduct an additional analysis on the data of experiment 1 and 2 together. Therefore, we compared the number of logical responses of Experiment 1 and 2 combined between the patient group and the controls. We used a simple Bayesian t-test to quantify the strength of the evidence against the null hypothesis. In the patients group, the number of logical answers was 5.2 (STD = 4.6), in the control group 2.3 (STD = 3.8). The Bayes Factor = 28.66, which is considered to be strong evidence and it remains strong, also after a robustness check. Hence, subtle differences
Table 4 Percentage false, true/false and true answers on the underinformative questions, correct answers to the filler items and the faux-pas score of the patient and the control group. Type of response
Patient group
Control group
Proportion (N) Underinformative utterances
False True as well as false True
28.2 (48) 22.3 (38) 50.6 (84) 90.6 (154)
Correct filler items Patient group
54.1 (92) 28.8 (49) 17.1 (29) 95.3 (162) Control group
Median (Mean, STD) Faux-pas score (ToM)
39 (37, 9.10)
48 (48.3, 4.74)
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
in the interpretation of single words can apparently lead to different conclusions, which may influence daily communication. The ability to derive SIs seems linked to ToM-ability for we observed a significant correlation between both abilities in the patient group of Experiment 2. Whether the observed association is a causal one and if so, what the direction of this causality may be, remains under discussion (Martin and McDonald, 2003). Some evidence points to ToM as a necessary precursor of pragmatic language skills (Happé, 1993; Winner et al., 1998) whereas some authors argue that it is the experience with pragmatic language that leads to the development of proficient ToM (Peterson and Siegal, 2000). However, the statement-evaluation-task we used does not seem to rely heavily on ToM-ability because there's no reference to an interlocutor whose intention/emotions may be of relevance. One could therefore wonder whether the observed correlation might be an “artificial” one, i.e. the result of an underlying core cognitive process that determines both ToM-ability and the ability to interpret underinformative utterances (UUs) in a pragmatic way (e.g., working memory, weak central coherence, formal thought disorder). Although these processes have all been studied to some extent with respect to ToM and pragmatic language problems in psychosis, they remain largely unexplored when it comes to pragmatic inferences such as SIs. Our results indicate that this is a gap that is worth looking into. In the present study we explored how psychotic patients deal with UUs and which factors could influence this process. Our results indicated that psychotic patients have problems deriving SIs and that this could be linked to an impaired ToM-ability. Some study characteristics should however be kept in mind. In the present study we considered only one scalar expression (“some”). Obviously the generalizability of our findings to other scalars needs to be explored. Furthermore, the UUs were presented in isolation, without context. Context may influence the interpretation of scalar expressions, so it would be interesting to examine whether the conversational context influences the way people with psychosis deal with underinformativeness. Also, we only studied the link between ToM and SIs but, as discussed above, other processes may be relevant and future research should look into this. Additionally, as discussed above as well, we treated ToM as a unitary concept Approaching ToM in a multidimensional way may be fruitful to gain more insight in its role in pragmatic language in general and in SIs in particular. A last, intriguing issue that needs to be considered, concerns a potential difference between the two participant groups with respect to their experience with standardized tests. The patients probably have more experience with standardized tests and therefore may have tackled the implicature task somewhat differently than controls. Patients may have been more inclined to interpret our experimental task as, for instance an intelligence test, a test of logical reasoning. With this mindset, they might have focused more on whether the sentences were logically correct, and less on spontaneously interpreting each sentence. Papafragou and Musolino (2003) nicely demonstrated that how an implicature task is framed or interpreted strongly influences whether implicatures are drawn or not. They showed that children's ability to derive scalar implicatures is affected by their awareness of the goal of the task. Typically, young children do not reject underinformative sentences in a statement-evaluation-task. However, some pre-training and more specific instructions emphasizing the importance to detect pragmatic infelicity lead to an increase in the rejection rates of children. Furthermore, Skordos and Papafragou (2016) showed that when the alternatives are made more accessible and relevant, children also reject underinformative sentences more often. This may be of relevance for our study. Although the implicature task was presented as a language comprehension task which focused on interindividual differences in sentence comprehension, we cannot rule out that some patients, given their experience with standardized tests, unlike controls, may have considered the stronger alternatives (e.g., “all oaks are trees” for “some oaks are trees”) as not relevant here and therefore decided not
5
to reject the underinformative sentences. A manipulation of the accessibility and relevance of the response alternatives is an interesting topic for future research. Contributors Martien Wampers and Walter Schaeken designed the study and wrote the study protocol. Martien Wampers, Sofie Schrauwen, Marc De Hert and Leen Gielen were responsible for the recruitment and testing of participants. Martien Wampers, and Sofie Schrauwen performed the statistical analyses. Martien Wampers wrote the first draft of the manuscript and all other authors contributed to and have approved the final manuscript.
Funding body agreements and policies Not applicable. Conflict of interest None of the authors has any conflict of interest to disclose.
Acknowledgements We would like to thank the entire staff of Sint-Joris, the ward for young people with psychosis of the University Psychiatric Centre KU Leuven, for their support during the data collection process.
References Andreasen, N.C., Hoffman, R.E., Grove, W.M., 1985. Mapping abnormalities in language and cognition. In: Alpert, A. (Ed.), Controversies in Schizophrenia. Guilford Press, New York, NY, pp. 199–226. Bambini, V., Arcara, G., Bechi, M., Buonocore, M., Cavallaro, R., Bosia, M., 2016. The communicative impairment as a core feature of schizophrenia: frequency of pragmatic deficit, cognitive substrates, and relation with quality of life. Compr. Psychiatry 71, 106–120. Bora, E., Yucel, M., Pantelis, C., 2009a. Theory of mind impairment in schizophrenia: metaanalysis. Schizophr. Res. 109 (1–3), 1–9. Bora, E., Yücel, M., Pantelis, C., 2009b. Theory of mind impairment: a distinct trait-marker for schizophrenia spectrum disorders and bipolar disorder? Acta Psychiatr. Scand. 120 (4), 253–264. Bott, L., Noveck, I.A., 2004. Some utterances are underinformative: the onset and time course of scalar inferences. J. Mem. Lang. 51 (3), 437–457. Brüne, M., Bodenstein, L., 2005. Proverb comprehension reconsidered - “theory of mind” and the pragmatic use of language in schizophrenia. Schizophr. Res. 75 (2–3), 233–239. Champagne-Lavau, M., Stip, E., 2010. Pragmatic and executive dysfunction in schizophrenia. J. Neurolinguistics 23 (3), 285–296. Chaytor, N., Schimitter-Edgecombe, M., Burr, R., 2006. Improving the ecological validity of executive functioning assessment. Arch. Clin. Neuropsychol. 21 (3), 217–227. Chevallier, C., Wilson, D., Happé, F., Noveck, I., 2010. Scalar inferences in autism spectrum disorders. J. Autism Dev. Disord. 40 (9), 1104–1117. Colle, L., Angeleri, R., Vallana, M., Sacco, K., Bara, B.G., Bosco, F.M., 2013. Understanding the communicative impairments in schizophrenia: a preliminary study. J. Commun. Disord. 46 (3), 294–308. Corcoran, R., Mercer, G., Frith, C.D., 1995. Schizophrenia, symptomatology and social inference: investigating “theory of mind” in people with schizophrenia. Schizophr. Res. 17 (1), 5–13. Corcoran, R., Cahill, C., Frith, C.D., 1997. The appreciation of visual jokes in people with schizophrenia: a study of “mentalizing” ability. Schizophr. Res. 24 (3), 319–327. Cummings, L., 2017. Cognitive aspects of pragmatic disorders. In: Cummings, L. (Ed.), Research in Clinical Pragmatics. Springer, Cham, Switzerland, pp. 587–616. De Neys, W., Schaeken, W., 2007. When people are more logical under cognitive load dual task impact on scalar implicature. Exp. Psychol. 54 (2), 128–133. Freeman, D., 2007. Suspicious minds: the psychology of persecutory delusions. Clin. Psychol. Rev. 27 (4), 425–457. Frith, C.D., 2004. Schizophrenia and theory of mind. Psychol. Med. 34 (3), 385–389. Frith, C.D., Allen, H.A., 1988. Language disorders in schizophrenia and their implications for neuropsychology. In: Bebbington, P., McGuffin, P. (Eds.), Schizophrenia: The Major Issues. Heinemann Professional Pub, Oxford, UK, pp. 172–186. Grice, H.P., 1975. Logic and conversation. In: Cole, P., Morgan, J.L. (Eds.), Syntax and Semantics (Vol. 3): Speech Acts. Academia Press, New York, pp. 41–58. Guasti, T., Chierchia, G., Crain, S., Foppolo, F., Gualmini, A., Meroni, L., 2005. Why children and adults sometimes (but not always) compute implicatures. Lang. Cogn. Process. 20 (5), 667–696. Happé, F.G.E., 1993. Communicative competence and theory of mind in autism: a test of relevance theory. Cognition 48 (2), 101–119. Harrington, L., Siegert, R.J., McClure, J., 2005. Theory of mind in schizophrenia: a critical review. Cogn. Neuropsychiatry 10 (4), 249–286. Janssen, I., Krabbendam, L., Jolles, J., van Os, J., 2003. Alterations in theory of mind in patients with schizophrenia and non-psychotic relatives. Acta Psychiatr. Scand. 108 (2), 110–117. Katsos, N., Bishop, D.V.M., 2011. Pragmatic tolerance: implications for the acquisition of informativeness and implicature. Cognition 120 (1), 67–81.
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053
6
M. Wampers et al. / Schizophrenia Research xxx (2017) xxx–xxx
Langdon, R., Coltheart, M., Ward, P.B., Catts, S.V., 2002a. Disturbed communication in schizophrenia: the role of poor pragmatics and poor mind-reading. Psychol. Med. 32 (7), 1273–1284. Langdon, R., Davies, M., Coltheart, M., 2002b. Understanding minds and understanding communicated meanings in schizophrenia. Mind Lang. 17 (1–2), 68–104. Linscott, R.J., 2005. Thought disorder, pragmatic language impairment, and generalized cognitive decline in schizophrenia. Schizophr. Res. 75 (2–3), 225–232. Martin, I., McDonald, S., 2003. Weak coherence, no theory of mind, or executive dysfunction? Solving the puzzle of pragmatic language disorders. Brain Lang. 85 (3), 451–466. Mazza, M., Di Michele, V., Pollice, R., Casacchia, M., Roncone, R., 2008. Pragmatic language and theory of mind deficits in people with schizophrenia and their relatives. Psychopathology 41 (4), 254–263. Montag, C., Dziobek, I., Richter, I.S., Neuhaus, K., Lehmann, A., Sylla, R., ... Gallinat, J., 2011. Different aspects of theory of mind in paranoid schizophrenia: evidence from a video-based assessment. Psychiatry Res. 186 (2–3), 203–209. Nieuwland, M.S., Ditman, T., Kuperberg, G.R., 2010. On the incrementality of pragmatic processing: an ERP investigation of informativeness and pragmatic abilities. J. Mem. Lang. 63 (3), 324–346. Papafragou, A., Musolino, J., 2003. Scalar implicatures: experiments at the semantics – pragmatics interface. Cognition 86 (3), 253–282. Peterson, C.C., Siegal, M., 2000. Insights into theory of mind from deafness and autism. Mind Lang. 15 (1), 123–145.
Pijnacker, J., Hagoort, P., Buitelaar, J., Teunissen, J.P., Geurts, B., 2009. Pragmatic inferences in high-functioning adults with autism and Asperger syndrome. J. Autism Dev. Disord. 39 (4), 607–618. Shamay-Tsoory, S.G., Shur, S., Barcai-Goodman, L., Medlovich, S., Harari, H., Levkovitz, Y., 2007. Dissociation of cognitive from affective components of theory of mind in schizophrenia. Psychiatry Res. 149 (1–3), 11–23. Skordos, D., Papafragou, A., 2016. Children's derivation of scalar implicatures: alternatives and relevance. Cognition 153, 6–18. Sponheim, S.R., Surerus-Johnson, C., Leskela, J., Dieperink, M.E., 2003. Proverb interpretation in schizophrenia: the significance of symptomatology and cognitive processes. Schizophr. Res. 65 (2–3), 117–123. Sprong, M., Schothorst, P., Vos, E., Hox, J., van Engeland, H., 2007. Theory of mind in schizophrenia: meta-analysis. Br. J. Psychiatry 191 (1), 5–13. Stone, V.E., Baron-Cohen, S., Knight, R.T., 1998. Frontal lobe contributions to theory of mind. J. Cogn. Neurosci. 10 (5), 640–656. Winner, E., Brownell, H., Happé, F., Blum, A., Pincus, D., 1998. Distinguishing lies from jokes: theory of mind deficits and discourse interpretation in right hemisphere brain-damaged patients. Brain Lang. 62 (1), 89–106.
Web references Spek, A., d. Faux Pas Test. Retrieved from. http://www.anneliesspek.nl/pagina27a.html.
Please cite this article as: Wampers, M., et al., Patients with psychosis struggle with scalar implicatures, Schizophr. Res. (2017), http://dx.doi.org/ 10.1016/j.schres.2017.08.053