The language of schizophrenia: An analysis ofmicro andmacrolinguisticabilities and their neuropsychological correlates
Andrea Marini a,b,⁎, Ilaria Spoletini b, Ivo Alex Rubino c, Manuela Ciuffa b, Pietro Bria d,Giovanni Martinotti d, Giulia Banfi b, Rocco Boccascino e, Perla Strom e,
Alberto Siracusano c, Carlo Caltagirone b,c, Gianfranco Spalletta b,c
a University of Udine, Udine, Italyb IRCCS Santa Lucia Foundation, Rome, Italy
c Department of Neuroscience, University of Rome “Tor Vergata”, Rome, Italyd Institute of Psychiatry, Catholic University of the Sacred Heart, Rome, Italy
e Department of Mental Health, ASL RMF, Rome, Italy
Received 17 December 2007; received in revised form 11 July 2008; accepted 20 July 2008Available online 2 September 2008
145A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
Covington et al., 2005). Abnormalities in language areashave been consistently reported in schizophrenicpatients (Spalletta et al., 2003; Honea et al., 2005;Koeda et al., 2006; Kuperberg et al., 2007). Language isa complex dynamic cognitive system which entailsintegration of multiple levels of linguistic and cognitiveprocessing (Kintsch and van Dijk, 1978; Caplan, 1992;Marini, 2001). Two main dimensions exist for language:1) awithin-sentence ormicrolinguistic dimension, respon-sible for intra-phrasal functions, and 2) a between-sentence or macrolinguistic dimension, responsible forinter-phrasal functions. Microlinguistic abilities involvethe organisation of phonological or graphemical patternsinto morphological strings and words (i.e. lexicalprocessing) and determine the syntactic context eachword requires for the generation of well-formed sentences(i.e. syntactic processing). Macroelaborative abilitiesdetermine the contextually appropriate meaning of aword or a sentence (i.e. pragmatic processing) and allowus to connect sentences or utterances bymeans of coherentand cohesive devices in order to integrate both linguisticand conceptual features of an uttered discourse or awrittentext and understand its main theme or gist (i.e. discourseprocessing) (Kintsch, 1994). Increasing evidence showsthat these levels can be analyzed separately (Davis et al.,1997; Marini et al., 2005a,b).
At the level of lexical processing selective problemshave been reported in schizophrenic speech (Barr et al.,1989; Allen et al., 1993; Spitzer, 1997; Marvel et al.,2004). Although normal at the levels of segmentalphonology (Chaika, 1974) and morphological organiza-tion (Chaika, 1990), their speech is characterized byflattened intonation (Cutting, 1985) and word-findingdifficulties (Andreasen, 1979; McKenna, 1994). Thesedata suggest preserved articulation skills and morpho-logical competence, but defective abilities to entail theirutterances with the correct suprasegmental pragmatic-prosodic contour. Furthermore, schizophrenic patients'disturbances include verbosity and empty content withverbal descriptions teaming with pronouns often with-out antecedents. Moreover, their productions ofteninclude deictic terms with no clear referents and verbsthat make their discourse vague and ambiguous.
At the level of syntactic processing, schizophrenicpatients' speech is usually normal, with no relevantaberrations (Andreasen, 1979; Covington et al., 2005).However, their sentences may appear somehow simpli-fied, characterized by reduced syntactic complexity(Morice and McNicol, 1986; Fraser et al., 1986; Thomaset al., 1990).
Schizophrenic patients usually have problems in thepragmatic use of language. Indeed, difficulties in
dealing with non-literal expressions (i.e. sarcasms,proverbs, metaphors, irony, idioms, indirect requests)have been widely reported (Chapman, 1960; Gibbs andBeitel, 1995; Mitchley et al., 1998; Sarfati and Hardy-Baylé, 1999; Langdon et al., 2002a,b; Tény et al., 2002;Corcoran and Frith, 2003; Lee et al., 2004). In somecases, such deficits have been related to an inability togenerate a correct “Theory Of Mind” (TOM: Sperberand Wilson, 2002), that is the ability to infer the mentalstates of their interlocutors (i.e. their perspective andtheir communicative intentions). In particular, TOMperformance predicted 39% of variance for proverbcomprehension in a group of schizophrenic patients(Brune and Bodenstein, 2005). From a general point ofview, the pragmatic deficits showed by these patientsmay be related to deficient TOM generation abilitywhich, in turn, may be the expression of a moregeneralized cognitive impoverishment in schizophrenia(Linscott, 2005).
The speech of schizophrenic patients is usuallyreported as disordered, filled with irrelevant pieces ofinformation and derailments (Andreasen, 1979). Sucherratic discourse may be linked to the inability to usepragmatic rules and/or by general cognitive deficitsinvolving factors such as attention (Neuchterlein et al.,2002), and/or action planning, ordering and sequencing(Woelwer and Gaebel, 2002; Docherty et al., 2006)which are crucial for efficient discourse processing.
To sum up, schizophrenic patients show linguisticdeficits which are very selective and subtle at themicrolinguistic level. However, deficits become morepervasive and severe at the macrolinguistic level whenpatients need to organize what they want to commu-nicate at the pragmatic-communicative level and gen-erate appropriate mental models. Unfortunately,patients' difficulties in these respects are not easy todetect and quantify. Furthermore, studies in the fieldhave not used a standard or comparable set of researchmethods. Finally, in order to adequately describe thelinguistic performance of a group of patients, it isessential to extend the analysis to all of the above-mentioned levels of complexity. Indeed, if a deficit atthe macrolinguistic level causes problems on themicrolinguistic level of text processing, the reversecan also be true, with microlinguistic deficits re-flecting on a more generalized difficulty in discourseorganization.
The main aim of the present study was to describe indetail the micro- and macrolinguistic abilities in a groupof schizophrenic patients on a set of single-picture andcartoon story description tasks. The narratives producedby the group of schizophrenic patients were compared to
146 A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
those provided by a comparison group of healthyparticipants. The elicited narratives were analyzed usingtechniques that allowed the evaluation of microlinguis-tic, macrolinguistic and informative aspects of discourseprocessing. In order to better capture the exact nature oftheir linguistic performance, additional analyses eval-uated the effect of macrolinguistic variables on micro-linguistic (i.e. lexical and sentence-level processing)performance and the neuropsychological predictors ofschizophrenic patients' linguistic performance. In parti-cular, it was hypothesized that selective problems inmicrolinguistic processing may be related to a moregeneral problem in discourse planning and organizationwhich, in turn, may be the consequence of a deficit inexecutive functions.
2. Methods
2.1. Participants and diagnostic assessment
Seventy-seven Italian-speaking participants wereincluded in the study. All participants gave writteninformed consent to participate in the study after allprocedures had been fully explained. Approval for thestudy had previously been obtained from the local ethicscommittee.
The experimental group consisted of 29 schizophre-nic patients diagnosed according with DSM-IV criteria(American Psychiatric Association, 1994). For allpatients included in this study, clinicians who treatedthe patients and knew their clinical history used DSM-IV criteria to make a preliminary diagnosis of schizo-phrenia. All diagnoses were then confirmed by onesenior clinical psychiatrist (G.S.) using the structuredclinical interview for DSM-IV (SCID-P; First et al.,1997a,b). Reproducibility of the preliminary andconfirmed clinical diagnoses showed substantial agree-ment with interrater reliability levels of kN0.80.
All schizophrenic patients included in the study werein the phase of clinical stability and had been receivingstable oral doses of atypical antipsychotics for at leastone month.
Exclusion criteria included: substance abuse ordependence during the foregoing year; a history oftraumatic brain injury or any other neurological illness;any past or present major medical illness that may affectbrain structures such as diabetes, cerebrovasculardisease, etc; any brain pathology identified on T2- orFLAIR-scans; and mental retardation.
The control group, 48 healthy participants, wasincluded in order to compare the narratives produced byschizophrenic participants with those uttered by healthy
individuals. All comparison subjects were carefullyscreened for a current or lifetime diagnosis of any axis Ior II disorder using SCID-I and SCID-II (First et al.,1997a,b). Schizophrenia among first-degree relativeswas an exclusion criterion as well as the other exclu-sion criteria mentioned above for patients withschizophrenia.
The two groups were matched for chronological age,gender and level of formal education. The socio-demographic characteristics of patients with schizo-phrenia and healthy comparison subjects are describedin Table 1.
2.2. Neuropsychological assessment
Two trained neuropsychologists (I.S. and G.B.)interviewed patients by administering the Mini MentalState Examination (MMSE: Folstein et al., 1975) todefine the global cognitive impairment, and a modifiedversion of the Mental Deterioration Battery (MDB) toassess selective cognitive performances. The MDB is astandardized and validated neuropsychological instru-ment (Carlesimo et al., 1996). The tests were selected toprovide information about the functionality of differentareas of cognition: language (Phonological VerbalFluency; Categorical Verbal Fluency), verbal memory(Rey's 15-word Immediate Recall and Delayed Recall),visual memory (Immediate Visual Memory), and logicalreasoning (Raven's Progressive Matrices '47). Con-structional praxis was assessed through the Copy of theRey–Osterrieth Complex Figure Test (Osterrieth, 1944),executive functioning by administering the WisconsinCard Sorting Test (WCST achieved categories, perse-verative errors and non-perseverative errors) (Heatonet al., 1993). Finally, information about psychomotorspeed, attention, and inhibitory processing was obtainedby using the Trail Making Test, Part A and Part B(Reitan, 1992). Two of the 29 schizophrenic patients,did not complete the neuropsychological examinationdue to non-compliance. Neuropsychological scores areshown in Table 2.
2.3. Assessment of narrative abilities
The narrative assessment was performed on thestory-tellings elicited with a picture-story descriptiontask. Three narratives were elicited with the help of asingle-picture stimulus (“picnic”) and two cartoonstories with six pictures each (“flower pot” and“quarrel”). The single picture “picnic”, taken from theWestern Aphasia Battery (WAB; Kertesz, 1982),describes the scene of a picnic on a lake. The two
Table 1Sociodemographic and clinical characteristics of 77 subjects with or without schizophrenia diagnosis
SD: Standard Deviation; degrees of freedom (75 for t-tests and 1 for chi-square tests).
147A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
cartoon picture sequenceswere used byHuber andGleber(1982) and by Nicholas and Brookshire (1993) respec-tively, to analyze textual competence and discourseinformation content. The former is about a man who iswalking with his dog and is hit on the head by a flowerpot. The latter story is about a quarrel between a husbandand wife. All three stories were very easy to describe andthe characters were clearly depicted. On average, it tookapproximately 1 min to describe each story. Eachparticipant described all three stories. The order ofpresentation was counterbalanced across subjects.
Since for this picture description task no normativedata are currently available, the narratives produced by
Table 2Neuropsychological scores of 27 patients with schizophrenia
Wisconsin Card Sorting Test, non-perseverative errors 3.8±4.1Trail Making Test, Part A (seconds) 102.0±63.4Trail Making Test, Part B (seconds) 255.0±168.6
MMSE: Mini Mental State Examination; SD: Standard Deviation.
the group of 29 schizophrenic patients were compared tothose uttered by the group of 48 healthy comparisonsubjects. In order to avoid poor performance due toshort-term memory limitations (Anderson et al., 1988;Wicksell et al., 2004), the cartoon story remained visibleuntil the participant completed the description. Eachstory-telling was tape-recorded and subsequently tran-scribed verbatim by one of the authors (A.M.) and by anexperienced speech therapist (M.C.) including phono-logical fillers, pauses, and false starts. The stories weresubjected to a quantitative textual analysis. The analysisfocused on four main aspects: verbal productivity,lexical and morpho-syntactic organization, informative-ness and textual organization (Marini and Carlomagno,2004). The scoring procedure was performed indepen-dently by two raters and then compared. Acceptableinterrater reliability was defined as k≥0.80, and the tworesearchers, after an appropriate training, achieved thislevel between them prior to the beginning of the studyand for all narrative values of the present study. Theresidual differences were resolved through discussion.
2.3.1. Microlinguistic analysisVerbal productivity was measured as the number of
units produced by each subject during the story-telling.The unit count included all verbalizations, irrespectiveof their linguistic or contextual correctness or appro-priateness (Haravon et al., 1994; Marini et al., 2005b).The total number of well-formed words, excludingphonological fillers, phonemic paraphasias and phoneticerrors, was then computed (Words in Table 3).
Table 3Levels of productivity and accuracy of microlinguistic processing of 29 schizophrenic patients and 48 comparison subjects, and comparisons betweengroups for microlinguistic variables on the three story-telling tasks
SD: Standard Deviation; ⁎Significant after Bonferroni correction for multiple comparisons.
148 A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
The lexical and syntactic organization was measuredin terms of lexical and morpho-syntactic processing.Lexical processing was assessed in three ways. At thefirst step, the analysis focused on the lexeme level ofword processing (Levelt et al., 1999) and a ratio ofphonological selection was computed. The ratio ofphonological selection (Marini et al., 2005a, 2007a,b)was obtained by dividing the number of words by thenumber of units. Units included all verbalizations,irrespective of their linguistic or contextual correctnessor appropriateness. Word count included only well-formed words, i.e. excluding phonological fillers andphonological errors. Thus, the ratio of phonologicalselection allowed assessing the ability to retrievephonologically well-formed words (% Phonologicalselection in Table 3).
The lemma level, i.e. the level at which semantic,morpho-syntactic and morphological features of a wordbecome available, was assessed in terms of production ofsemantic paraphasias and paragrammatic errors. Theratio of semantic paraphasias was calculated by dividingthe number of semantic paraphasias by the number ofwords (% Semantic paraphasias in Table 3) (Marini et al.,2005b). These errors were scored when a target wordwas substituted by a semantically related word (Haravonet al., 1994). An example for a semantic paraphasia is theword “mother” in the sentence “here he's talking to hismother”, where the speaker implied “wife”.
For each story, the ratio of paragrammatic errorsincluded grammatical errors with bound morphemes(for example, “questo è una coppia”: “this [masc inItalian] is a couple [fem]” – in Italian questo should bequesta) and function words (for example, batte da una
porta “he is knocking from a door” – in Italian dainstead of a). The ratio of paragrammatic errors wascalculated by dividing the number of paragrammaticerrors by the number of words (% Paragrammatic errorsin Table 3).
As a gross measure of syntactic organization the Meanlength of utterance was calculated (MLU, in Table 3). Anutterance is a unit of speech surrounded by silence. In thetranscription, utterance boundaries were determined byintonation contours as well as by pause length. For eachstory description, the total number of utterances wasassessed following the criteria established in the ShewanSpontaneous LanguageAnalysis System (Shewan, 1988).The ratio measuring the MLU was calculated by dividingthe total number of words by the number of utterances.
2.3.2. Macrolinguistic analysisThe informative content of each narrative was
measured in order to obtain a numerical evaluation oflexical–semantic appropriateness in describing the gistof the story. The lexical–semantic appropriateness wasdetermined counting the Lexical Information Units(LIUs), i.e. words that were not only phonologicallywell-formed but also appropriate from a grammaticaland pragmatic point of view (Marini et al., 2005a,b).Therefore, all those words that were classified assemantic paraphasias, fillers, paragrammatisms or pre-sent in tangential utterances (i.e. utterances that weresomehow deviating from the gist of the story) wereexcluded from the LIUs count. The ratio of lexicalinformativeness was obtained dividing the amount ofLIUs by the amount of words (% Lexical informa-tiveness in Table 4).
Table 4Levels of accuracy and informativeness of macrolinguistic processing of 29 schizophrenic patients and 48 comparison subjects, and comparisonsbetween groups for macrolinguistic variables on the three story-telling tasks
SD: Standard Deviation; ⁎Significant after Bonferroni correction for multiple comparisons.
149A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
For each story the total number of possible thematicunits was identified (Marini et al., 2005a). Therefore, itwas possible to measure a ratio of thematic informa-tiveness for each story dividing the number of thematicunits in the subjects' speech by the total amount ofthematic units that could be obtained by the story (%Thematic informativeness in Table 4).
The macrolinguistic measure of discursive organiza-tion included ratios of cohesive errors as well as bothlocal and global coherence errors. Cohesive errorsmeasure the structural connectivity across utterances,whereas local coherence errors are a measure of failureto link utterances conceptually. A ratio of cohesiveerrors was obtained by dividing the total amount ofcohesive errors by the number of utterances that formedthe description (% Cohesive errors in Table 4). Acohesive error was scored each time cohesive functionwords were used incorrectly or whenever utteranceswere abruptly interrupted but at the condition that thesubsequent utterances were completing the idea pre-viously introduced. The ratio of local coherence errorswas measured dividing the total number of localcoherence errors by the number of utterances produced(% Local coherence errors in Table 4). Missing orambiguous referents and semantic shifts were countedas local coherence errors. A semantic shift was scoredwhenever there was an abrupt stop within an utterance,after which the flow of thoughts was not continued inthe following utterance but, instead, a new conceptbegan. For instance, in “he's trying to… and here he is ona landing” the first utterance remained unfinished, thesecond utterance introduced a new scene. Missingreferents were instances in which the referent of a
pronoun or the implicit subject of a verb (Italian is a pro-drop language) were not unambiguously clear or wereincorrect. For example, consider the following sequenceof utterances: “Qui stanno litigando furiosamente. Poidice: (…)” (“Here they are fighting furiously. Then[implicit pronoun] says: (…)”). In the second utterancethere was a missing referent because it was not clearwhom the verb “dice (says)” referred to. A ratio of globalcoherence errors was calculated dividing the totalnumber of tangential utterances by the amount ofutterances that formed each description (% Globalcoherence errors in Table 4). An utterance where aderailment in the flow of discourse was detected wasconsidered tangential. For instance, while describing the“picnic” story, a schizophrenic participant said: “It is apicnic / I like picnics / I have had several picnics in mylife /”. Obviously, with respect to the gist of the story,utterances such as “I like picnics” and “I have had severalpicnics in my life” are not relevant to telling the story perse and were probably triggered by semantic–conceptualassociations that can be considered as tangential.
2.4. Statistical analyses
Comparisons of sociodemographic variables weremade using chi-square test for gender and Student's t-test for continuous variables (i.e. age and educationallevel).
A series of 10 t-tests was performed in order tocompare the two groups for all individual micro- andmacrolinguistic performance variables (i.e. Words,% Phonological Selection, % Semantic paraphasias,% Paragrammatic errors, Mean length of utterance,
150 A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
% Cohesive errors, % Local coherence errors, % Globalcoherence errors, % Lexical informativeness, % The-matic informativeness). These comparisons wererepeated separately for the three story-telling tasks. Inthese analyses the level of statistical significance wasdefined as pb0.05 after Bonferroni correction formultiple comparisons (alphab0.05/30–alphab0.0016).
In order to control for the effect of macrolinguisticvariables involved in discourse processing – often foundto be impaired in schizophrenia – (i.e. % Localcoherence errors and % Global coherence errors) onlinguistic measures examining the microlinguistic levelof processing (namely % Semantic paraphasias, that iserrors at the lexico-semantic level, and % Paragram-matic errors, that is errors at the morpho-syntactic level),we performed two univariate analyses of covariance(ANCOVAs) with the two macrolinguistic variablescores as covariates and the microlinguistic variablescores as dependent variables. The cartoon story “flowerpot” was used as an index for linguistic performance inthese ANCOVAs. To determine the neuropsychologicalvariables which correlated with those linguistic vari-ables that were impaired in schizophrenic patients on atleast two story-telling tasks, we performed a series ofstepwise multiple regression analyses, with thoseindividual linguistic variables which were impaired inschizophrenic patients (i.e. % Semantic paraphasias,% Paragrammatic errors, % Local coherence errors,% Global coherence errors, % Lexical informativeness,and % Thematic informativeness) as dependent vari-ables, and the neuropsychological variable scores asindependent variables, using a forward procedure and anF to enter of 4. This was done only in the schizophrenicgroup. The linguistic scores obtained on the cartoonstory “quarrel” were used as an index for linguisticperformance in these multiple regression analyses. Inaddition, pre-selection of neuropsychological variablesto include in the stepwise regression models was done byusing correlation analyses and Fisher's r to z transforma-tion in order to determine the significance of correla-tions. In the stepwise multivariate models only variableswith pb0.05 in the pre-selection univariate correlationanalyses were included.
3. Results
3.1. Assessment of narrative abilities
Tables 3 and 4 show the differences between groupsfor all the micro- and macrolinguistic variablesseparately for the three story-telling tasks. Apart from“Words”, “% Phonological Selection” and “Mean length
of utterance”, which did not differ between schizo-phrenic and control participants, the remaining linguisticvariables did differ in at least one of the three story-telling tasks in the schizophrenic group compared to thecontrol subjects group. As shown in Tables 3 and 4, thefollowing six linguistic variables in at least two of thestory-telling tasks were significantly different betweengroups: % Semantic paraphasias, % Paragrammaticerrors, % Local coherence errors, % Global coherenceerrors, % Lexical informativeness, and % Thematicinformativeness.
3.2. Effect of macrolinguistic abilities on microlinguisticperformance
To further clarify the relationship between micro- andmacrolinguistic variable performances, we compared% Semantic paraphasias and % Paragrammatic errors inthe “flower pot” story-telling of the patient and healthycomparison groups. Two ANCOVAs were run with% Local coherence errors and %Global coherence errorsas covariates in both analyses. Results revealed nodifferences between the two groups on % Semanticparaphasias (F = 0.104; df= 1,69; p= 0.748) and% Paragrammatic errors (F=2.025; df=1,69; p=0.159).
3.3. Neuropsychological predictors of linguistic perfor-mance in schizophrenia
As mentioned earlier, six out of the ten linguisticvariables were significantly different between theschizophrenic and control subjects groups in at leasttwo story-telling tasks. The “quarrel” story-tellingscores of these six linguistic variables were used as anindex to analyze the relationship between linguisticability and cognition in schizophrenic patients only.Cognitive variables which in the univariate correlationanalyses were related (pb0.05) to the linguisticvariables (see Table 5) were selected as independentvariables in the further stepwise multiple regressionanalyses.
None of the neuropsychological variables correlatedsignificantly with % Thematic informativeness or %Local coherence errors in the univariate correlationanalyses. Thus, a series of four stepwise multipleregression analyses was performed to identify theneuropsychological predictors of % Semantic parapha-sias, % Paragrammatic errors, % Global coherence errors,and % Lexical informativeness scores as dependentvariables, because only these four linguistic variablescores correlated with one or more cognitive variablescores (as shown in bold in Table 5). Results indicated that
Table 5Relationships between % Semantic paraphasias, % Paragrammatic errors, % Global coherence errors, and % Lexical informativeness and cognitivelevel in the schizophrenic group as emerged from the univariate correlation analysis
Semantic paraphasias Paragrammatic errors Global coherence errors Lexical informativeness
Cognitive variables Pearson's r (p-value) Pearson's r (p-value) Pearson's r (p-value) Pearson's r (p-value)
MMSE: Mini Mental State Examination; WCST: Wisconsin Card Sorting Test; TMA: Trail Making A; TMB: Trail Making B.
151A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
the MMSE score was the only predictor of % Semanticparaphasias (F=4.793; Standard coefficient=−0.401;df=1,25; r2=0.161; p=0.0381). Secondly, number ofnon-perseverative errors on the WCST was the onlypredictor of % Paragrammatic errors (F=15.546; Stan-dard coefficient=0.619; df=1,25; r2=0.383; p=0.0006).Furthermore, Trail Making B score was the only predictorof % Global coherence errors (F=23.035; Standardcoefficient=0.692; df=1,25; r2 =0.480; pb0.0001).Lastly, Raven's Progressive Matrices '47 score was theonly predictor of % Lexical informativeness (F=9.104;Standard coefficient = 0.517; df=1,25; r2 = 0.267;p=0.0051).
4. Discussion
The present work investigated linguistic abilities andtheir neuropsychological correlates in a group ofschizophrenic patients. The main result indicates amild problem at the level of microlinguistic processingand a severe deficit on the macrolinguistic dimension ofprocessing. Furthermore, three linguistic parameters(i.e. Paragrammatic errors, Lexical informativeness, andGlobal coherence errors) were explained by specificneuropsychological domains whereas one additionallinguistic parameter (i.e. Semantic paraphasias) waspredicted by a measure of global cognitive level.
At the microlinguistic level, the speech of patientswith schizophrenia was quite variable, as there wereaspects of processing which did not statistically differfrom the comparison group (i.e. productivity) andaspects of processing in which their performance was
not uniform. As for productivity (i.e. production ofwords), despite the non significance, it is noteworthythat schizophrenic participants were somehow moreverbose, producing on average more words than thecomparison subjects in all three stories. Lexicalprocessing analysis showed an absence of problems atthe phonetic or phonological level. Indeed, % Phono-logical Selection was similar in both groups. However,schizophrenic participants produced more Semanticparaphasias and Paragrammatic errors, suggesting thepresence of a selective problem in dealing with someaspects of the lemma level of lexical processing.Interestingly, the score of Paragrammatic errors wasexplained by a measure of attention (specifically, set-shifting ability), i.e. WCST non-perseverative errors.This index likely reflects a failure in the use ofcontextual information, which is necessary to setshifting (Barceló and Knight, 2002). Such associationbetween Paragrammatic errors and WCST non-perse-verative errors is consistent with models of lexicalprocessing (Indefrey and Levelt, 2004; Levelt et al.,1999) where the role of contextual information inspeech production has been highlighted. Indeed, thelatter has been thought to involve a multi-step process:a) a phase of pre-linguistic conceptual planning of theintended message, in which the speaker retrieves allavailable data from long-term declarative memory toformulate a mental model coherent with the context;b) a linguistic phase, in which the preverbal messageis converted into a speech plan by means of a pre-liminary stage of lexical selection followed by one oflexical access; and c) an articulation phase in which
152 A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
production actually takes place. The stage of lexicalselection allows speakers to select the lexical itemsthat correspond to the intended meaning through anactivation/inhibition mechanism (Green, 1986, 1998;Paradis and Goldblum, 1989; Paradis, 1993). At theend of the lexical selection process, the target word hasbeen activated and the linguistic system gains access toits semantic, morpho-syntactic and morphologicalfeatures (lemma level of word representation) andthen to its phonological form (lexeme level of wordrepresentation). In case of sentence production, themorpho-syntactic structures requested by the selectedlemma (i.e. its argumental structure) guide the processof sentence generation by means of thematic roleassignments and phrase generation, and the informationcontained in the lemmas of the selected lexical items isused to build up well-formed syntactic representations(Chomsky, 1988, 1995; Caplan, 1992; Pinker, 1999).Accordingly, the present results show that, althoughschizophrenic speech is not impaired at the level oflexeme processing, it seems to be characterized bylemma-level deficits, that is in lexico-semantic andmorpho-syntactic aspects of lexical processing. Indeed,abnormalities in associative connections betweenwords and concepts have long been considered a corefeature of schizophrenia (Manschreck et al., 1988;Spitzer et al., 1993, 1994; Kuperberg et al., 2007) and,in chronic schizophrenic patients, particularly thosewith positive thought disorder, have been associated toanomalous increases in activity within inferior pre-frontal and temporal cortices (Kuperberg et al., 2007).As for syntactic processing, the speech of patients withschizophrenia was usually normal, with no relevantaberrations (Andreasen, 1979; Covington et al., 2005)even though the uttered sentences were somehowsimplified and characterized by reduced syntacticcomplexity (see also Fraser et al., 1986; Morice andMcNicol, 1986; Thomas et al., 1990, DeLisi, 2001).
As to the macrolinguistic dimension of languageprocessing, the narratives of schizophrenic patients wereless informative and more tangential than those utteredby the group of comparison participants. Indeed, thestory-descriptions provided by the schizophrenicpatients were barely informative and characterized byempty speech, filled with semantic paraphasias and,mostly, by local coherence errors such as the frequentuse of pronouns without antecedents, deictic terms withno clear referents and frequent derailments that maketheir discourse vague and ambiguous. Interestingly,DeLisi (2001) reports reduced use of clausal embeddingand connectivity between sentences in the narrativespeech of chronic schizophrenic patients noting that the
number of conjoined and embedded clauses wasinversely correlated to poor content. However, in thatcase it was not possible to discern whether the reducedinter-sentential connectivity were linked to an aberrantuse of linguistic devices (which would cause cohesiveerrors) or to the lack of conceptual connections betweencontinguous sentences (local coherence errors). In ourexperiment, however, a strong group-related differencewas found in the production of local coherence errorswhereas only slight problems were due to cohesivedisturbances. This suggests that the reduced inter-sentential connectivity might have been determined bydifficulties in focusing on the connections between thepropositions conveyed by the sentences. Interestingly,once measures of macrolinguistic discourse processing(i.e. % Local coherence errors and % Global coherenceerrors) have been included in the analysis as covariates,no differences between schizophrenic and controlparticipants in the production of % Semantic para-phasias and % Paragrammatic errors were detectable.This suggests that the lexico-semantic errors producedby the schizophrenic patients included in the study arerelated to more general difficulties in global reasoningand sequencing as if a general problem in generating acoherent mental model were responsible for theirsemantic confusion. This is also consistent with ourfinding of a correlation between Lexical informative-ness and Raven's Progressive Matrices '47, whichmeasures logical reasoning, defined as “the ability todiscriminate and perceive relations between any funda-mentals, new or old” (Cattell, 1943). In the conceptualstage of message planning, declarative memory pro-vides knowledge not only about what to say, but alsoabout what has previously been said (linguistic context)and about the particular situation, place and timein which the communicative exchange takes place(Johnson-Laird, 1980; Levinson, 1983). It is assumedthat, during the phase of message planning, a super-visory attentional system (SAS: Green, 1998) is incharge of controlling the speakers' communicativeintentions and modulating the amount of informationthat the speaker intends to communicate as well as itsrelevance with respect to what has previously been said(Grice, 1975). Accordingly, our data show that variancein Global coherence errors was explained by perfor-mance on the Trail Making B task, therefore confirmingour hypothesis that executive (specifically, inhibitoryprocessing) deficits may be related to difficulties inconceptual message planning and, at the lexical level,to selective problems in the process of lemma-levelsemantic selection. This is also in accordance withthe dysexecutive model of formal thought disorder in
ContributorsGianfranco Spalletta and Andrea Marini designed the study.
that could inappropriately bias theirwork, is reported by anyof the authors.
153A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
schizophrenia (Barrera et al., 2005). Furthermore,this finding supports previous reports where the schizo-phrenic speech disorder was attributed to deficient at-tention (Docherty et al., 2006). Overall, the present datasuggest that the problems observed at the microlinguisticlevel of processing in the schizophrenic group mayreflect more general deficits at the macrolinguistic level,such as the inability to take into account all availablecontextual data and select an adequate mental model.This latter finding is interesting in light of the hypothesisthat abnormalities within fronto-temporal systems playan important role in the aetiology and expression ofschizophrenia-spectrum disorders (Woods et al., 2007).Indeed, in previous studies deficits in context pro-cessing tasks have been associated with prefrontalcortical dysfunction in both medicated and non medi-cated schizophrenic patients (Wood and Flowers, 1990;Curtis et al., 2001; Barch et al., 2003; MacDonald et al.,2005). In particular, on tasks requiring high contextprocessing demands, MacDonald et al. (2005) foundincreased activity in the middle frontal gyrus (Brod-mann's area 9) in a group of healthy participants and in agroup of patients with non-schizophrenia psychosis, butnot in the group of patients with schizophrenia.
Some issues have to be considered before the con-clusion. Firstly, linguistic problems in schizophrenicpatients treated with psychotropic drugs may be sec-ondary to the treatment. However, re-running eachanalysis using antipsychotic drug dosages (in olanzapineequivalents) as an additional independent variable didnot change the significance of results (data availableupon request). Secondly, because of the cross-sectionaldesign of the study and the chronic nature of the illness inthe majority of patients we cannot specifically attributethese linguistic changes to psychosocial, neurodevelop-mental or neurodegenerative factors. It must also beconsidered that patients and comparison participantswere matched for educational level. Thus, one of thepossible social factors which can influence the linguisticresults was excluded.
In conclusion, the present data suggest that languageproduction in schizophrenia is impaired at bothmacro andmicrolinguistic levels of processing, with the strongestindicators being macrolinguistic. Patient's language isdisordered and filled with irrelevant pieces of informationand derailments (Andreasen, 1979). The pragmatic rulesgoverning local and global coherence structuring arecontinuously violated leading to tangential, poor, some-times even incoherent speech (see also Linscott, 2005).Such erratic discourse may be linked to the inability to usepragmatic rules and/or by cognitive deficits involvingfactors such as attention (Neuchterlein et al., 2002), and/or
action planning, ordering and sequencing (Docherty et al.,2006; Woelwer and Gaebel, 2002) which are crucial forefficient discourse processing. Such an impairment likelyreflects prefrontal cortical dysfunction. Further researchis required in order to deepen the cognitive nature oflinguistic performance in schizophrenic patients and itsneuro-anatomical correlates.
Role of funding sourceThis research was supported by IRCCS Fondazione “Santa Lucia”.
IRCCS did not have any further role in the study design; in the analysisand interpretation of the data; in the writing of the report, and in thedecision to submit the paper for publication.
Andrea Marini designed the linguistic protocol, supervised linguisticassessment and coding, managed the literature search, and wrotethe final version of the manuscript. Gianfranco Spalletta and CarloCaltagirone designed the neuropsychological protocol and super-vised neuropsychological assessment. Gianfranco Spalletta performedthe statistical analyses and contributed to the interpretation of thedata. Ilaria Spoletini and Giulia Banfi managed neuropsychologicaldata acquisition and processing. Manuela Ciuffa administered thelinguistic tests. Perla Strom, Ivo Alex Rubino, Pietro Bria, GiovanniMartinotti, Rocco Boccascino, and Alberto Siracusano coordinatedpatient recruitment.
Conflict of interestNo actual or potential conflict of interest including any financial,
personal or other relationships with other people or organizations
AcknowledgementsThis work was funded by IRCCS Fondazione “Santa Lucia”. The
authors would like to thank the two anonymous reviewers for theirinsightful comments.
References
Allen, H.A., Liddle, P.F., Frith, C.D., 1993. Negative features, retrievalprocesses and verbal fluency in schizophrenia. Br. J. Psychiatry163, 769–775.
American Psychiatric Association, 1994. Diagnostic and StatisticalManual of Mental Disorders, fourth ed. American PsychiatricAssociation, Washington.
Anderson, S.W., Routh, D.K., Ionanescu, V.V., 1988. Serial positionmemory of boys with Duchenne muscular dystrophy. Dev. Med.Child Neurol. 30, 328–333.
Andreasen, N.C., 1979. Thought, language, and communicationdisorders: clinical assessment, definition of terms, and assessmentof their reliability. Arch. Gen. Psychiatry 36, 1325–1330.
Andreasen, N.C., Grove, W.M., 1986. Thought, language, andcommunication in schizophrenia: diagnosis and prognosis. Schi-zophr. Bull. 12, 348–359.
Barceló, F., Knight, R.T., 2002. Both random and perseverative errorsunderlie WCST deficits in prefrontal patients. Neuropsychologia40, 349–356.
154 A. Marini et al. / Schizophrenia Research 105 (2008) 144–155
Barch, D., Sheline, Y.I., Csernansky, J., Snyder, A., 2003. Workingmemory and prefrontal cortex dysfunction: specificity to schizo-phrenia compared with major depression. Biol. Psychiatry 1 (53),376–384.
Barr, W.B., Bilder, R.M., Goldberg, E., Kaplan, E., Mukherjee, S.,1989. The neuropsychology of schizophrenic speech. J. Comm.Dis. 22, 327–349.
Barrera, A., McKenna, P.J., Berrios, G.E., 2005. Formal thought disorderin schizophrenia: an executive or a semantic deficit? Psychol. Med.35, 121–132.
Brune, M., Bodenstein, L., 2005. Proverb comprehension recon-sidered—“Theory of Mind” and the pragmatic use of language inschizophrenia. Schizophr. Res. 75, 233–239.
Caplan, D., 1992. Language. Structure, Processing and Disorders. MITPress, Cambridge.
Carlesimo, G.A., Caltagirone, C., Gainotti, G., 1996. The MentalDeterioration Battery: normative data, diagnostic reliability andqualitative analyses of cognitive impairment. The Group for theStandardization of the Mental Deterioration Battery. Eur. Neurol.36, 378–384.
Cattell, R.B., 1943. The measurement of adult intelligence. Psychol.Bull. 40, 153–193.
Chaika, R., 1974. A linguist look at “schizophrenic” language. BrainLang. 1, 257–276.
Chaika, R., 1990. Understanding Psychotic Speech: Beyond Freud andChomsky. Charles C. Thomas, Springfield.
Chapman, L.J., 1960. Confusion of figurative and literal usages ofwords by schizophrenic and brain damaged patients. J. Abnorm.Soc. Psychol. 60, 412–416.
Chomsky, N., 1988. Language and Problems of Knowledge. MITPress, Cambridge.
Chomsky, N., 1995. The Minimalist Program. MIT Press, Cambridge.Corcoran, R., Frith, C.D., 2003. Autobiographical memory and theory
of mind: evidence of a relationship in schizophrenia. Psychol.Med. 33, 897–905.
Covington, M.A., He, C., Brown, C., Naçi, L., McClain, J.T., SirmonFjordbak, B., Semple, J., Brown, J., 2005. Schizophrenia and thestructure of language: the linguist view. Schizophr. Res. 77, 85–98.
Cutting, J., 1985. The Psychology of Schizophrenia. ChurchillLivingstone, Edinburgh.
Davis, G., O'Neil-Pirozzi, T., Coon, M., 1997. Referential cohesionand logical coherence of narration after right hemisphere stroke.Brain Lang. 56, 183–210.
DeLisi, E., 2001. Speech disorder in Schizophrenia: review of theliterature and exploration of its relation to the uniquely humancapacity for language. Schizophr. Bull. 27, 481–496.
Docherty, N.M., Strauss, M.E., Dinzeo, T.J., St-Hilaire, A., 2006. Thecognitive origins of Specific Types of Schizophrenic SpeechDisturbances. Am. J. Psychiatry 163, 2111–2118.
First,M.B., Spitzer, R.L., Gibbon,M.,Williams, J.B.W., 1997a. StructuredClinical Interview for DSM-IVAxis I Disorders (SCID-I), ClinicianVersion. American Psychiatric Press, Washington.
First, M.B., Gibbon,M., Spitzer, R.L.,Williams, J.B.W., Benjamin, L.S.,1997b. Structured Clinical Interview for DSM-IVAxis II PersonalityDisorders (SCID-II). American Psychiatric Press, Washington.
Folstein, M.F., Folstein, S.E., McHugh, P.R., 1975. “Mini-mentalstate”: a practical method for grading the cognitive state of patientsfor the clinician. J. Psychiatr. Res. 12, 189–198.
Fraser, W.I., King, K.M., Thomas, P., Kendell, R.E., 1986. The diagnosisof schizophrenia by language analysis. Br. J. Psychiatry 148, 275–278.
Gibbs Jr., R.W., Beitel, D., 1995. What proverb understanding revealsabout how people think. Schizophr. Bull. 133–154.
Green, D.W., 1986. Control, activation, and resource: a framework and amodel for the control of speech in bilinguals. Brain Lang. 27, 210–223.
Green, D.W., 1998. In: Paradis, Michel (Ed.), Aspects of BilingualAphasia. Brain Lang, vol. 59, pp. 531–535.
Grice, H.P., 1975. Logic and conversation. In: Cole, P., Morgan, J.P.(Eds.), Syntax and Semantics: Speech Acts. Academic Press, NewYork, pp. 41–58.
Haravon, A., Obler, L., Sarno, M., 1994. A method for microanalysisof discourse in brain-damaged patients. In: Bloom, R., Obler, L.,De Santi, S., Ehrlich, J. (Eds.), Discourse Analysis and Applica-tions: Studies in Adult Clinical Populations. Lawrence Erlbaum,Hillsdale, pp. 47–80.
Honea, R., Crow, T.J., Passingham, D., Mackay, C.E., 2005. Regionaldeficits in brain volume in schizophrenia: a meta-analysis of voxel-based morphometry studies. Am. J. Psychiatry 162, 2233–2245.
Huber, W., Gleber, J., 1982. Linguistic and non-linguistic processingof narratives in aphasia. Brain Lang. 16, 1–18.
Indefrey, P., Levelt, W.J., 2004. The spatial and temporal signatures ofword production components. Cognition 92, 101–144.
Kertesz, A., 1982. The Western Aphasia Battery. Grune & Stratton,New York.
Kintsch, W., 1994. Text comprehension, memory, and learning. Am.Psychol. 49, 294–303.
Kintsch, W., van Dijk, T., 1978. Toward a model of text comprehen-sion and production. Psychol. Rev. 85, 363–394.
Koeda, M., Takahashi, H., Yahata, N., Matsuura, M., Asai, K., Okubo,Y., Tanaka, H., 2006. Language processing and human voiceperception in schizophrenia: a functional magnetic resonanceimaging study. Biol. Psychiatry 15, 948–957.
Kuperberg, G.R., Deckersbach, T., Holt, D.J., Goff, D.,West,W.C., 2007.Increased temporal and prefrontal activity in response to semanticassociations in schizophrenia. Arch. Gen. Psychiatry 64, 138–151.
Langdon, R., Coltheart, M., Ward, P.B., Catts, S.V., 2002a. Disturbedcommunication in schizophrenia: the role of poor pragmatics andpoor mind-reading. Psychol. Med. 32, 1273–1284.
Langdon, R., Davies, M., Coltheart, M., 2002b. Understanding mindsand understanding communicated meanings in schizophrenia.Mind Lang. 17, 68–104.
Marini, A., 2001. Elementi di psicolinguistica generale. SpringerVerlag, Milano.
Marini, A., Carlomagno, S., 2004. Analisi del discorso e patologia dellinguaggio. Springer Verlag, Milano.
Marini, A., Carlomagno, S., Caltagirone, C., Nocentini, U., 2005a. Therole played by the right hemisphere in the organization of complextextual structures. Brain Lang. 93, 46–54.
Marini, A., Boewe, A., Caltagirone, C., Carlomagno, S., 2005b.Assessment of age-related differences in the processing of textualdescriptions. J. Psycholinguist. Res. 34, 439–463.
Marini, A., Caltagirone, C., Pasqualetti, P., Carlomagno, S., 2007a.Patterns of language retrieval in patients with non-chronic non-fluent aphasia following specific therapies. Aphasiology 21,164–186.
Marini, A., Lo russo, M.L., D'Angelo, G., Civati, F., Turioni, A.C.,Fabbro, F., Bresolin, N., 2007b. Evaluation of narrative abilities inpatients suffering from Duchenne Muscular Dystrophy. BrainLang. 102, 1–12.
Marvel, C.L., Schwartz, B.L., Isaacs, K.L., 2004. Word productiondeficits in Schizophrenia. Brain Lang. 89, 182–191.
McKenna, P.J., 1994. Schizophrenia and Related Syndromes. OxfordUniversity Press, Oxford.
Mitchley, N.J., Barber, J., Gray, Y.M., Brooks, N., Livingston, M.G.,1998. Comprehension of irony in schizophrenia. Cogn. Neurop-sychiatry 3, 127–138.
Morice, R., McNicol, D., 1986. Language changes in schizophrenia:a limited replication. Schizophr. Bull. 12, 239–251.
Neuchterlein, K.H., Asarnow, R.F., Subotnik, K.L., 2002. Thestructure of schizotypy: relationships between neurocognitiveand personality disorder features in relatives of schizophrenicpatients in the UCLA Family Study. Schizophr. Res. 54, 121–130.
Nicholas, L.E., Brookshire, R.H., 1993. A system for quantifying theinformativeness and efficiency of the connected speech of adultswith aphasia. J. Speech Hear. Res. 36, 338–350.
Osterrieth, P.A., 1944. Le test de copie d'une figure complexe. Arch.Psychol. 30, 206–356.
Paradis, M., 1993. Linguistic, psycholinguistic, and neurolinguisticaspects of “interference” in bilingual speakers: the ActivationThreshold Hypothesis. Int. J. Psycholinguist. 9, 133–145.
Paradis, M., Goldblum, M.C., 1989. Selective crossed aphasia in atrilingual aphasic patient followed by reciprocal antagonism. BrainLang. 36, 62–75.
Pinker, S., 1999. Words and Rules: The Ingredients of Language. BasicBooks, New York.
Reitan, R.M., 1992. Trail Making Test. Reitan NeuropsychologyLaboratory, Arizona.
Sarfati, Y., Hardy-Baylé, M.C., 1999. How do people with schizo-phrenia explain the behaviour of others? A study of theory of mindand its relationship to thought and speech disorganization inschizophrenia. Psychol. Med. 29, 613–620.
Shewan, C.M., 1988. The Shewan Spontaneous Language Analysissystem for aphasic adults: description, reliability and validity.J. Comm. Dis. 103–138.
Spalletta, G., Tomaiuolo, F., Marino, V., Bonaviri, G., Trequattrini, A.,Caltagirone, C., 2003. Chronic schizophrenia as a brain mis-connection syndrome: a voxel-based morphometry study. Schi-zophr. Res. 64, 15–23.
Sperber, D., Wilson, D., 2002. Pragmatics, modularity and mind-reading. Mind Lang. 17, 3–23.
Spitzer, M., 1997. A cognitive neuroscience view of schizophrenicthought disorder. Schizophr. Bull. 23, 29–50.
Spitzer, M., Braun, U., Maier, S., Hermle, L., Maher, B., 1993. Indirectsemantic priming in schizophrenic patients. Schizophr. Res. 11,71–80.
Spitzer, M., Weisker, I., Winter, M., Maier, S., Hermle, L., Maher, B.A.,1994. Semantic and phonological priming in schizophrenia. J.Abnorm.Soc. Psychol. 103, 485–494.
Tény, T., Herold, R., Szili, I.M., Trixler, M., 2002. Schizophrenicsshow a failure in the decoding of violations of conversationalimplicatures. Psychopathology 35, 25–27.
Thomas, P., King, K., Fraser, W.I., Kendell, R.E., 1990. Linguisticperformance in schizophrenia: a comparison of acute and chronicpatients. Br. J. Psychiatry 156, 204–210.
Wicksell, R.K., Kuhlgren, M., Melin, L., Eeg-Olofsson, E., 2004.Specific cognitive deficits are common in children with Duchennemuscular dystrophy. Dev. Med. Child Neurol. 46, 154–159.
Woelwer, W., Gaebel, W., 2002. Impaired Trail-Making Test-Bperformance in patients with acute schizophrenia is related toinefficient sequencing of planning and acting. J. Psychiatr. Res. 36,407–416.
Wood, F.B., Flowers, D.L., 1990. Hypofrontal vs hypo-sylvian bloodflow in schizophrenia. Schizophr. Bull. 16, 413–424.
Woods, S.P., Weinborn, M., Posada, C., O'Grady, J., 2007. Preliminaryevidence for impaired rapid verb generation in schizophrenia.Brain Lang. 102, 46–51.
