Journal of Abnormal Psychology 2000, Vol. 109, No. 4, 761-767
Copyright 2000 by the American Psychological Association, Inc. 0021-843X/00/$5.00 DOI: 10.1037//0021-843X.109.4.761
Contextual Insensitivity in Schizophrenic Language Processing: Evidence From Lexical Ambiguity Debra Titone, Deborah L. Levy, and Philip S. Holzman McLean Hospital and Harvard Medical School The authors investigated whether contextual failures in schizophrenia are due to deficits in the detection of context or the inhibition of contextually irrelevant information. Eighteen schizophrenia patients and 24 nonpsychiatric controls were tested via a cross-modal semantic priming task. Participants heard sentences containing homonyms and made lexical decisions about visual targets related to the homonyms' dominant or subordinate meanings. When sentences moderately biased subordinate meanings (e.g., the animal enclosure meaning of pen), schizophrenia patients showed priming of dominant targets (e.g., paper) and subordinate targets (e.g., pig). In contrast, controls showed priming only of subordinate targets. When contexts strongly biased subordinate meanings, both groups showed priming only of subordinate targets. The results suggest that inhibitory deficits rather than context detection deficits underlie contextual failures in schizophrenia.
One of the most interesting areas of cognitive function is the ability to contextually adjust behavior. If context is to be used effectively, relevant material must be appcehended and maintained in memory, and irrelevant material or impulses that may normally produce adaptive functioning must be suppressed. Although a general account of contextual processing is applicable to many cognitive domains, the domain of language is ideal for understanding its components. Many elements of language are ambiguous (e.g., words such as palm tropical vegetation or the surface of o n e ' s hand). Thus, one may place different sources of contextual constraint into conflict (e.g., palm in an immediate context that supports its less frequent or subordinate interpretation) and observe the ensuing behavior of an intact or impaired cognitive system. There are many reports of the disordered use of context in the language behavior of schizophrenia patients (Cameron, 1938; Bleuler, 1911/1950; Chapman & Chapman, 1973; Kraepelin, 1919/1971). Honigfeld (1963) found that schizophrenia patients performed less well than healthy adults in predicting missing words from highly predictable passages (see also Blaney, 1974; de Silva & Hemsley, 1977). Similarly, Chapman, Chapman, and Miller (1964) showed that schizophrenia patients consistently misinterpreted the meaning of ambiguous words such as pen when the
surrounding sentence context biased the subordinate (i.e., less frequent) meaning (e.g., "When the farmer bought a herd of cattle, he needed a new pen"; see also Chapman & Chapman, 1973; Chapman, Chapman, & Daut, 1976; Cohen, Barch, Carter, & Servan-Schreiber, 1999; Cohen, Braver, & O'Reilly, 1996; Cohen & Servan-Schreiber, 1992). More recently, Kuperberg and colleagues (Kuperberg, McGuire, & David, 1998) found that schizophrenia patients showed less disruption than control participants in processing sentences containing semantic violations (e.g., "The crowd waited eagerly: the young man drank [vs. grabbed] the
guitar"). Although these studies provide overwhelming support for the notion that schizophrenia patients have difficulty with context, the cognitive mechanisms responsible for contextual failures in schizophrenia are less understood. Many of the experimental tasks used previously permit measurement of only one kind of response given a particular context (e.g., participants choose either the contextually relevant or irrelevant meaning). Thus, although it is possible to assess t h e relative activation between contextually relevant and irrelevant meanings, it is unclear whether context affects activation of meanings that do not rise to the level of a response. For instance, schizophrenia patients may be able to detect contextually relevant material during comprehension, but they may be less able to use context to inhibit irrelevant material. Consistent with this notion, Done and Frith (1984) showed that schizophrenia patients' word recognition thresholds were reduced in the presence of context; however, their selection of words given a particular context was deficient. In the present study, we employed a task frequently used with schizophrenia patients that allows for the simultaneous assessment of contextually relevant and irrelevant meanings, the semantic priming task (Meyer & Schvaneveldt, 1971). In a basic version of this task, participants are presented with prime stimuli (e.g., doctor) and target stimuli (e.g., nurse) and are instructed to make a lexical decision (i.e., word-nonword decision) for each target. Responses are said to be primed (i.e., faster or more accurate, or both) to the extent that prime and target stimuli share a semantic
Debra Titone, Deborah L. Levy, and Philip S. Holzman, Psychology Research Laboratory, McLean Hospital, Belmont, Massachusetts, and Department of Psychiatry, Harvard Medical School. This study was supported in part by National Institutes of Health Grants R03 MH60272, 5T32MH16259-19, P01 MH31154, R01 MH49487, R01 MH31340, and K05 MH010121. We are grateful to Robert D. Vincent for statistical programming assistance; Michael Coleman, Steven Matthysse, John Vuchetich, and Yue Chen for commenting on the manuscript; and Laurie Teraspulsky for recruiting participants. Correspondence concerning this article should be addressed to Debra Titone, Psychology Research Laboratory, Mailman Research Center, McLean Hospital, Belmont, Massachusetts 02478. Electronic mail may be sent to
[email protected]. 761
762
TITONE, LEVY, AND HOLZMAN
relationship. Two cognitive mechanisms thought to underlie semantic priming are automatic and controlled semantic processing (Neely, 1991; Neely & Keefe, 1989; Posner & Snyder, 1975a, 1975b). Automatic processes are fast-acting responses to stimuli that are not mediated by expectation or overt cogitation. In contrast, controlled processes occur relatively later in time than automatic processes or require more effort. If schizophrenia patients fall to detect contextual information, one might expect a global deficiency in automatic semantic priming. This is not the case, however. Semantic priming for singleword stimuli in schizophrenia is often normal or larger than that of healthy adults (Barch et al., 1996; Chapin, McCown, Vann, Kenn e y , & Youssef, 1992; Chapin, Vann, Lycaki, Josef, & Meyendorff, 1989; Henik, Nissimov, Priel, & Umansky, 1995; Kwapil, Hegley, Chapman, & Chapman, 1990; Moritz et al., 1999; Spitzer, Braun, Hermle, & Maier, 1993; Spitzer, Braun, Maier, Hermle, & Maher, 1993; Vinogradov, Ober, & Shenaut, 1992; Vinogradov et al., 1996; Weisbrod, Maier, Harig, Himmelsbach, & Spitzer, 1998). Moreover, other work suggests that schizophrenia patients have a broader semantic sensitivity than controls (Spitzer, Braun, Hermle, & Maier, 1993; Spitzer, Braun, Maier, 1993; Weisbrod et al., 1998; see also Mitchell et al., 1991; Koyama et al., 1991; and Niznikiewicz et al., 1997, for similar conclusions using eventrelated potentials). Such results are inconsistent with the existence of an automatic processing deficit in schizophrenia. In contrast to automatic processing, performance on tasks of controlled or inhibitory processing is impaired in schizophrenia. Schizophrenia patients have been shown to produce longer color naming latencies than do controls on the Stroop task (Abramczyk, Jordan, & Hegel, 1983; Boucart, Mobarek, Cuervo, & Danion, 1999; Cohen et al., 1999; Hepp, Maier, Hermle, & Spitzer, 1996; Schooler, Neumann, Caplan, & Roberts, 1997). They perform worse than controls on latent inhibition tasks (i.e., learning that a previously irrelevant stimulus is now relevant; Baruch, Hemsley, & Gray, 1988; McDowd, Filion, Harris, & Braff, 1993; Swerdlow, Braff, Hartston, Perry, & Geyer, 1996), and they show reduced negative priming (i.e., responding to a stimulus that was previously inhibited; Beech & Claridge, 1987; Claridge, Clark, & Beech, 1992; David, 1995; Laplante, Everett, & Thomas, 1992; Salo, Robertson, Nordahl, & Kraft, 1997). In addition, many researchers report schizophrenia-related deficits on tasks requiring the inhibition of a dominant mental set (e.g., Wisconsin Card Sorting Task) or behavioral tendency (e.g., Trail Making Task; Flashman, Flaum, Gupta, & Andreasen, 1996; McGrath, Scheldt, Welham, & Clair, 1997; Radant, Claypoole, Wingerson, Cowley, & Roy-Byme, 1997; Robert et al., 1997; Yurgelun-Todd & Kinney, 1993). Finally, schizophrenia patients are impaired in semantic priming when long stimulus-onset asynchronies (SOAs) between prime and target stimuli are present, a condition that is believed to reflect controlled processing (Barch et al., 1996). We examined whether the detection of context via automatic semantic processing or the inhibition of contextually irrelevant material accomplished via controlled semantic processing is deficient in schizophrenia using the cross-modal semantic priming task. Control participants and schizophrenia patients listened, over headphones, to prime stimuli consisting of 64 noun-noun homonyms (e.g., jam) embedded in contexts biased toward the homonyms' less frequent, subordinate meanings. Immediately after offset of the spoken prime, participants saw visual targets related
to either the homonyms' dominant or subordinate meanings. Participants were instructed to make lexical decisions (i.e., wordnonword judgments) for the visual targets and to press an appropriately labeled button to indicate their decision. To examine the effects of contextual strength, we embedded homonyms in sentence contexts that moderately biased the subordinate meaning or strongly biased the subordinate meaning. Because the sentences always biased the subordinate meaning, visual targets related to subordinate meanings were always contextually relevant, and targets related to dominant meanings were always contextually irrelevant. Thus, responses to subordinate targets reflect the extent to which listeners detect contextually relevant information contained in the sentence contexts. In contrast, responses to dominant targets reflect the extent to which listeners inhibit contextually irrelevant information. We expect facilitated responding to contextually relevant subordinate targets for controls but impaired responding to contextually irrelevant dominant targets. Given our hypothesis that contextual failures in schizophrenia are associated with impairments in the inhibition of contextually irrelevant information, we predict schizophrenia patients to show more priming than healthy controls for dominant, contextually irrelevant meanings when context biases subordinate meanings. Given our hypothesis that contextual failures in schizophrenic language processing are not associated with impairments in the detection of contextually relevant information, we expect schizophrenia patients to perform similarly to controls in the priming of contextually relevant subordinate meanings. Method
Participants The participants (n = 18) were a random selection of patients who had been discharged during the previous year from a psychiatric hospital and who met criteria of the Diagnostic and Statistical Manual of Mental Disorders (4th edition; DSM-IV; American Psychiatric Association, 1994) for schizophrenia. Diagnoses were assigned by experienced clinicians based on a review of a standardized interview (Structured Clinical Interview for the DSM-IV) conducted by trained interviewers and a review of all hospital records. The control participants (n = 24), who were selected from a medical outpatient clinic, were matched to the patients on age, parental socioeconomic status, and education. The following additional inclusion criteria applied to all participants: fluency in English, no diagnosed organic brain disease, no substance-alcohol abuse or dependence within the past 2 years, no tardive dyskinesia, no use of alcohol or recreational drugs within 1 week of testing, and verbal IQ above 70. Written informed consent was obtained from all participants. Table 1 presents clinical and demographic information for the patient and control samples. The groups did not differ in terms of age, number of years of education, or estimated verbal IQ (based on the vocabulary scale of the Wechsler Adult Intelligence Scale, 1981). The groups did differ in level of functioning as measured by the Global Assessment Scale (Endicott, Spitzer, Fliess, & Cohen, 1976); as would be expected, the schizophrenia patients had a significantly lower level of functioning than the control participants (p < .001). The groups also differed in amount of clinically assessed thought disorder, as measured by the Thought Disorder Index (p < .001 ; Holzman, Shenton, & Solovay, 1986; Shenton, Solovay, & Holzman, 1987; Shenton, Solovay, Holzman, Coleman, & Gale, 1989; Solovay et al., 1986; Solovay, Shenton, & Holzman, 1987). As would also be expected, schizophrenia patients had a significantly higher total amount of thought disorder than control participants.
763
CONTEXTUAL INSENSITIVITY IN SCHIZOPHRENIA Table 1 Participant Characteristics Schizophrenia patients (n = 18)
Controls (n = 24)
Characteristic
M
SD
M
SD
p
Age (years) Years of education Estimated verbal IQ Global Assessment Scale score Duration of illness (years) Clorpromazine equivalent (rag) Total Thought Disorder Index score Brief Psychiatric Rating Scale score
34.3 13.7 104.4 38.4 11.1 547.3
8.5 1.8 11.9 11.7 6.0 815.4
37.3 14.5 105.4 77.6
11.8 2.5 11.2 10.2
ns ns ns
