Is Semantic Processing Impaired in Individuals With High Schizotypy?

Transcription

1 ORIGINAL ARTICLE Is Semantic Processing Impaired in Individuals With High Schizotypy? Celia J. A. Morgan, BSc, PhD,* Nick J. Bedford, BSc,* Alison O Regan, BA, PGradDip, and Susan L. Rossell, BSc, PhD* Abstract: Semantic processing deficits are present in schizophrenia. However, this research has often been criticized for methodological artifacts and confounds, including long hospitalizations and medication of patient samples. Utilizing high schizotypes (psychosis-prone individuals) can overcome these confounds. Previously, similar deficits have been reported in high schizotypes and schizophrenia on semantic priming tasks. In contrast to schizophrenia research, no other types of semantic processing have been examining in high schizotypes. Semantic processing is multifaceted, thus, deficits on semantic priming can not answer whether high schizotypes have difficulty with explicit semantic processing, that is, on tasks that require the conscious recollection of semantic information. In the current study, 24 high and 30 low scorers on the O-LIFE schizotypy scale were administered a battery of semantic processing measures. The high schizotypy group did not show global semantic processing impairments (as has been shown in schizophrenia), only impairments on a task designed to examine subtle categorization processing. Such deficits can be equated to those found on semantic priming tasks, in that both require quick and accurate access to semantic networks. Key Words: Schizotypy, semantic memory, categorization. (J Nerv Ment Dis 2009;197: ) Research on schizotypy or psychosis-proneness has suggested that the symptoms of schizophrenia are at the extreme end of a continuum that ranges from healthy, through to florid psychosis (Claridge and Hewitt, 1987). Individuals with high schizotypy are considered part of the normal diversity but are argued to mark a cognitive vulnerability towards psychosis, with these individuals performing many cognitive tasks similarly to individuals with schizophrenia (Peters et al., 1994). Hence, schizotypy has been associated with deficits in attention on the Continuous Performance Task (CPT; Lenzenweger et al., 1991; Obiols et al., 1992) and executive functioning, in the form of perseverative errors on the Wisconsin Card Sorting Task (Gooding et al., 1999; 2001; Tallent and Gooding, 1999; Suhr, 1997). In the last few years, semantic processing deficits have been suggested to be central to cognitive abnormalities in schizophrenia and are present on a wide range of tasks (Chen et al., 1994; Tamlyn *Section of Cognitive Neuropsychiatry, Psychological Medicine, Institute of Psychiatry, De Crespigny Park, London, United Kingdom; Clinical Psychopharmacology Unit, University College London, London, United Kingdom; and Cognitive Neuropsychiatry Laboratory, Alfred Psychiatry Research Centre, School of Psychology, Psychiatry, and Psychological Medicine, Monash University, Melbourne, Australia. Supported by British Academy to a small project. Send reprint requests to Susan L. Rossell, BSc, PhD, Cognitive Neuropsychiatry Laboratory, Alfred Psychiatry Research Centre, School of Psychology, Psychiatry and Psychological Medicine, Monash University, 1st Floor, Old Baker Building, The Alfred Hospital, Commercial Road, Melbourne VIC 3004, Australia. susan.rossell@med.monash.edu.au. Copyright 2009 by Lippincott Williams & Wilkins ISSN: /09/ DOI: /NMD.0b013e31819dc127 et al., 1992; Rossell et al., 1998, 1999, 2000). Despite this, only a few studies have examined the relationship between semantic processing and high schizotypy. Thus far, the majority of these studies have used a semantic priming task. This is an experimental technique that examines the nature and the relationship between the mental representations of words. It does so implicitly, that is, without being consciously directed to the precise rationale for the task (thus is not conscious), such that, when asked to engage in either naming or lexical decision semantic priming task, healthy controls illustrate a facilitated response for related or associated word pairs compared with unrelated word pairs. Previous experience with the related information is postulated to be the reason behind the facilitated priming effect. Semantic priming differences in participants with high schizotypy ratings have now been established in several studies (Beech et al., 1991; Kravetz et al., 1998; Moritz et al., 1999; Morgan et al., 2006). On the basis of these findings, it has been argued that semantic deficits in high schizotypes are similar to those reported in schizophrenia. There has been little research, however, on other aspects of semantic memory in schizotypy. Deficits in verbal fluency have been reported (Barrantes-Vidal et al., 2003; Duchene et al., 1998). Barrantes-Vidal reported an association between positive symptoms in schizotypy (i.e., unusual experiences on the Oxford-Liverpool Inventory of Feelings and the Experiences O-LIFE scale) and increased verbal fluency. While, Duchene demonstrated that high schizotypy scorers produced more rare words during fluency than low schizotypy scorers (similar results were also reported by Miller and Chapman 1983 on a word association task). However, both these studies used the letter version of verbal fluency, where participants are required to generate as many words as they can that begin with the same letter. This task requires participants to access their lexicon and then hold and maintain previously generated words on line. As such, this task probably indexes executive functioning rather than semantic memory per se. Category fluency can be viewed as a more appropriate semantic memory task, and while deficits on this task have been observed in unaffected siblings of people with schizophrenia (Laurent et al., 1999) and people with schizotypal personality disorder (Trestman et al., 1995), it has not thus far been administered to healthy individuals scoring high on schizotypy. Kiang and Kutas (2005) report the only other study of semantic processing in high and low schizotypy participants. They assessed responses to a category-verification task and found no behavioral group differences. Thus, while considerable effort has been devoted to examining executive functioning and performance on semantic priming tasks in this psychosis-prone population, as yet, there has been no study that has fully characterized the semantic processing performance of high schizotypy individuals on a comprehensive semantic battery of explicit memory measures. Explicit semantic functions are those that require conscious recollection of semantic attributes, such as naming everyday objects, understanding meanings of words, or understanding meaning of sentences. Therefore, this study set out to fully characterize the explicit semantic processing performance of individuals scoring highly on schizotypy scales. We chose to take extreme scorers on the STA (an The Journal of Nervous and Mental Disease Volume 197, Number 4, April 2009

2 The Journal of Nervous and Mental Disease Volume 197, Number 4, April 2009 Semantics and Schizotypy overall schizotypy measure included in the O-LIFE most related to the positive symptoms) rather than adopting a symptom cluster approach, as the former has yielded some of the most consistent results in the literature (Avons et al., 2003). We also aimed to correlate the schizotypy subfactors derived from the O-LIFE with any emergent semantic processing deficits. To this end, 7 explicit semantic processing tasks were selected and performance was compared between high and low schizotypy groups. People with schizophrenia show performance deficits on all 7 measures. The definition task examines whether participants can generate and recognize the meanings of single words. On this task, patients with schizophrenia are impaired in both the generation and recognition of meanings (Rossell and David, 2006; Rossell and Batty, 2008). The Boston naming task examines whether participants can name visual line drawings, patients also demonstrate deficits on naming on this task (Joyce et al., 1996). A semantic fluency task was included, which tests a participants ability to correctly produce single words to category cues. This provides a more relevant index of semantic memory than letter fluency. On this task schizophrenia patients make more errors and generate fewer exemplars (Rossell et al., 1999). A semantic categorization task was used, which requires individuals to verify words as members or nonmembers of a conceptual category under conditions of increasing uncertainty. Again, deficits in patients with schizophrenia have been observed on this task, with patients being slower and less accurate at categorizing words (Chen et al., 1994; Rossell and David, 2006). We also included a word association task that requires participants to identify associated words and a synonym detection task, which involves participants identifying synonyms. People with schizophrenia are impaired on both of these tasks (Rossell and David, 2006). Lastly, a nonsense sentences task that examines participants ability to correctly identify real from nonsense sentences was included. Patients with schizophrenia have been found to be less accurate at distinguishing real from nonsense sentences (Rossell et al., 1998). Research utilizing high schizotypes (psychosis-prone individuals) also overcomes some of the confounding factors involved in studying actively symptomatic schizophrenics, since the former are nonpsychotic, unmedicated, and unhospitalised. If the symptoms manifest in people scoring highly on schizotypy scales are related to those exhibited in patients with clinical psychotic symptoms, then it is possible that these individuals may demonstrate some of the semantic processing deficits observed in schizophrenia. Using this population is useful as any semantic deficits that are present before the onset of clinical symptoms could potentially be more informative with regards to the cognitive causes rather than consequences of the disorder. As little research has examined semantic processing deficits in schizotypy we tentatively hypothesized that nonclinical adults who rate highly on a self-report measure of schizotypy will perform more poorly than low schizotypal participants on the 7 different semantic processing tasks, however that deficits would be more subtle than those observed in schizophrenia. METHODS Participants Two hundred fifty-one native English-speaking students completed the O-LIFE (Mason et al., 1995) questionnaire initially. The O LIFE is a 159-item questionnaire based on the Combined Schizotypal Traits Questionnaire (Bentall et al., 1989), and yields 4 factors: unusual experiences, cognitive disorganisation, introvertive anhedonia, and impulsive nonconformity. Unusual experiences relates to unusual perceptual events, hallucinatory experiences, and magical thinking. This is the most consistent factor emerging in analyses of schizotypal scales (Bentall et al., 1989; Nunn and Peters, 2001) and is represented by such questions as: I have felt that I have special, almost magical powers and Does your sense of smell sometimes become unusually strong? The factor of cognitive disorganization represents difficulties in sustaining attention and concentration, as well as moodiness and social anxiety, this is scored by questions such as: No matter how hard you try to concentrate, do unrelated thoughts always creep into your mind? and Do you frequently have difficulty starting to do things? Introvertive anhedonia is conceptualized as difficulties in gaining enjoyment from social or other sources, a lack of enjoyment of physical and emotional intimacy, and a preference for spending time alone. These items are thought to relate particularly to the negative symptoms of schizophrenia and are identified by such items as: There are just not many things that I have ever really enjoyed and Are people better off if they stay aloof from emotional involvement with people? The factor of impulsive nonconformity taps into asocial behaviors, impulsivity, and nonconformity. Items on this factor include: Do you ever have the urge to break or smash things? and Do you often feel like doing the opposite of what people suggest, even though you know they are right? In addition to providing separate scores for each of the 4 factors, analysis of the O-LIFE also generates a composite STA score. The questionnaire also generates an Extraversion score; these questions were included in the O LIFE as filler items and the data from these were not analyzed. Participants for the experimental stage were selected on the basis of their scores on the STA subscale of the O-LIFE. The groups were separated on the basis of this scale as it gives a general measure of schizotypy, as the authors aimed to correlate scores on the O-LIFE subfactors with any semantic deficits. The STA scores were trichotomized on the basis of pilot work into groups corresponding to, for the low scorers the bottom 10th percentile, and for the high scorers the top 90th percentile. The high schizotypy group comprised 26 (13 male) students who scored between 24 and 36 and the low schizotypy group comprised 32 (12 male) students who scored between 0 and10. Design and Procedure An independent groups design was used to compared the high and low schizotypy groups. All participants gave written, informed, witnessed consent and were paid for participation. Participants in the experimental stage were tested on the battery of semantic processing tasks described below. They were also administered the National Adult Reading Test (NART: Nelson, 1982) to estimate premorbid verbal IQ, the Edinburgh Handedness Inventory (Oldfield, 1971) and Peter s Delusional Inventory (Peters et al., 1999). Lastly, they completed a semantic priming task, the results of which are reported elsewhere (Morgan et al., 2006). Semantic Tests Definition Task For the definition task, participants were asked to (1) generate definitions of single target words. This established whether the information for that word was available. This was compared with (2) a recognition-forced choice task. The recognition task used the same targets and participants were asked to select the correct definition of each word from 4 choices. Participants always generated definitions first, and then after a delay of approximately 90 minutes, during which the participant performed other cognitive tests, completed the recognition task. The stimuli were 42 words, between 4 and 8 letters long, they were a selection of nouns, verbs, and adjectives and were taken from the MRC Oxford Psycholinguistic Database Lippincott Williams & Wilkins 233

3 Morgan et al. The Journal of Nervous and Mental Disease Volume 197, Number 4, April 2009 In the generate condition the words were presented to the participants. The experimenter read the word to the participant, who was then required to generate a definition of each word, i.e. what the participant thought the word meant. The score sheet was placed on the table in front of the participant so they could see the spellings of each word. The participants performance was monitored for the first 5 words, if they were not completing the task correctly the task instructions were repeated and an example was given of a correct definition. These definitions were then marked against definitions given in the Oxford Concise English Dictionary and the Merriam- Webster Online Dictionary, and each answer was awarded between 2 and 0 points depending on its accuracy and completeness. The mean % accurate was calculated out of a possible score of 84. The second version of the task was a recognition-forced choice paradigm. All the words were presented with accompanying definitions. The definitions were created using the Oxford Concise English Dictionary and the Merriam-Webster Online Dictionary and Thesaurus. Two foil answers, 1 incorrect answer and 1 correct answer, were allocated to each word. The foil answers were created using a thesaurus and finding definitions of similar words or concepts. For example, college, correct an educational establishment providing higher education, foils a school for young children/an institution for educating children and incorrect an alcoholic drink resembling gin. The participants were required to indicate which definition they thought best fitted the word; they were told to take as long as they needed to complete this task. The mean percentage of correctly selected items was calculated. Boston Naming Task Participants were asked to name 60 progressively difficult visual line drawings (Kaplan et al., 1983). The mean % of correctly named items was recorded. Semantic Fluency This was examined using 4 different categories: animals, furniture, fear, and happy. Participants were asked to give as many category exemplars as they could for each category. For the 2 affective categories a category example was defined as objects or events that can result in fear or happiness. The following examples were provided, fear crocodile and happy dancing. Sixty seconds was allowed for each category; the examiner recorded participants responses. The total number of words generated for the 4 categories was calculated minus errors (i.e., category inappropriate words) and perseverations. Semantic Categorization We used a revised version of the category task reported in Chen et al., (1994). Eighteen categories were selected from the norms of Battig and Montague (1969) and Hampton and Gardiner (1983), e.g. vehicles. For each category, 5 different exemplar words were selected to provide different degrees of relatedness, resulting in 90 trials. These exemplar words were either: (1) high frequency (vehicle bus), (2) low frequency (vehicle ferry), (3) borderline (vehicle ski), (4) related but outside the category (vehicle horse), and (5) unrelated (vehicle banker). The 5-exemplar types were matched for the number of letters, number of phonemes and syllables, and concreteness and imageability. They were not matched for frequency (Kucera and Francis, 1967) as this was a test variable (high-frequency category exemplars also had high lexical frequency and low-frequency category exemplars had low lexical frequency as reported in Kucera and Francis 1967 ): (1) high: ( 30) mean 86, (2) low: (1 29) mean 5, (3) borderline: mean 21, (4) related: mean 38, and (5) unrelated: mean 22. Stimuli were presented in random order centrally on a computer screen. First, category names appeared in capital letters for 1000 milliseconds. After a delay of 550 milliseconds exemplar words appeared in lower case for 200 milliseconds followed by a 2000 milliseconds response window. All letters were in Times New Roman with a 44-point font. After the exemplar word was shown, participants were asked to indicate whether it belonged to the category or not by pressing one of 2 buttons, yes or no. Participants were given 10 practice trials. For high and low-frequency categories the single obvious correct answer was yes. For related and unrelated categories the single obvious correct answer is no. The % of correct responses to each of these 4 conditions was recorded. For the borderline condition neither yes nor no was correct, however we decided to examine the % of yes responses. We therefore report the 5 accuracy scores from this task. Word Associations Using the stimuli from PALPA (Psycholinguistic Assessment of Language Processing in aphasia (Kay et al., 1992) task 51, we examined the participant s ability to correctly identify word associations. Thirty key words were compared with 4 other words; 1 related, 1 semantic foil, and 2 unrelated (e.g., key word-fog, relatedmist, semantic foil-steam, and unrelated-bolt and lock). This is a paper and pencil task with key words on the left hand side of the page in bold followed by the 4 possible associates on the same line. The participant was asked to indicate, by underlining, which of the 4 other words was the most closely related or associated by meaning to the key word. Accuracy was recorded. Synonyms Using the stimuli from PALPA task 49 we examined the participant s ability to correctly identify synonyms. Sixty word pairs (30 synonyms and 30 nonsynonyms) were randomly presented centrally on a computer screen. The first word in the pair was presented for 200 milliseconds; there was a 550-millisecond blank screen followed by the second word for 200 milliseconds. There was a blank screen between trials of 2500 milliseconds and the participants were able to respond for up to 2000 milliseconds after second word presentation. All letters were in lower case Times New Roman with a 44-point font. Participants were asked to indicate whether or not the pair was a synonym using a 2-button press (synonym start beginning, not a synonym tool crowd). Accuracy and RTs (using RTs to correct stimuli only) were recorded. Nonsense Sentences Participants were asked to classify 96 sentences as true or false. Thirty-two sentences were true, 32 true unlikely and 32 nonsense (Rossell et al., 1998 for an earlier version of this task). True sentences were correct representations of the world, i.e., A rose is a flower. Nonsense sentences were not accurate representations of the world and were produced by combining 2 true sentences, i.e., Doors can sing is the product of choir boys can sing and doors can close. The unlikely sentences were also plausible or possible representations of the world, but were unlikely or unusual constructions, i.e. Leaves are red. All the sentences were syntactically correct. The content of the 96 sentences was also manipulated. Five different content types were selected: neutral (48 sentences), persecutory (12), grandiose (12), religious (12), and somatic (12). The content of the emotional sentences were selected to relate to some of the common delusional ideas seen in patients with schizophrenia. When constructing the sentences, care was taken to match the number of words in each sentence, the length of the words, their frequency in the English language (Kucera and Francis, 1967), and the grammatical construction of each sentence (i.e., most had the form noun-verb-noun, noun-verb-adjective although there were some variations) across both the different sen Lippincott Williams & Wilkins

4 The Journal of Nervous and Mental Disease Volume 197, Number 4, April 2009 Semantics and Schizotypy tence types and content types. Rossell et al. (1998) used paper and pencil administration of this task. In the current project, we collected RT data, and therefore, used a computerized version. Sentences were presented in the center of a computer screen (white on black) for 2000 milliseconds, followed by a blank screen for 1800 milliseconds. Participants were given 3000 milliseconds from the presentation of the sentence to make a response. They were asked to indicate using a 2-button press whether they thought each sentence was true (something that could happen) or false (something that was not possible). All letters were lower case, were presented in Times New Roman with a 36-point font, and all sentences were presented on 1 line. We pilot tested this task with a student control group. They had an average response time of approximately 1500 milliseconds. Clare et al., (1993) had suggested that patients with schizophrenia took twice as long to respond as controls on a similar task, thus explaining why we decided to use a 3000-millisecond response window. RT and accuracy were recorded. Statistical Analysis Demographic data were analyzed with independent groups t tests, where data were nonparametric with a Mann-Whitney U test and frequency data with chi square. For RT data, any response that was more than 2.5 standard deviations (SDs) from the overall mean for each participant was excluded, as were incorrect responses. Univariate ANOVAs were conducted on the primary scoring variable or variables from each task with group (high schizotypy, low schizotypy) as the between participants factor. We Bonferroni corrected for multiple comparisons within each particular task if appropriate. Pearsons correlations were conducted where group differences emerged with the 4 subfactors of the O-LIFE and the PDI. TABLE 1. Demographics and OLIFE Scores of High and Low Schizotypy Groups High (N 26) Low (N 32) F p Male/female 12/14 13/19 NS Age 21.2 (1.7) 21.1 (2.7) NS No. yr education 15.5 (1.3) 15.3 (2.2) NS NART IQ (6.8) (5.3) NS OLIFE scores out of 36 Unusual 20.2 (4.3) 4.1 (2.9) Cognitive 18.0 (3.2) 6.9 (4.9) Introversive 6.2 (4.6) 3.6 (3.4) Impulsive 13.9 (2.7) 7.3 (3.5) STA 26.6 (2.7) 7.0 (2.6) RESULTS All participants were between the ages of 18 and 31 years, with estimated IQ scores ranging between 103 and 122 points (as scored by the NART). There was no significant difference in age, gender, education, or IQ between the 2 participant groups. There was, however, a trend for more of the high scorers to be left handed than low scorers ( , p 0.081) (Table 1). The PDI (Peters et al., 1999), revealed that the high schizotypy group showed a greater degree of delusional thinking than the low schizotypy group (scores out of 21: high 1.2 SD: 7.5 and low 3.8 (SD: 2.5): F (1, 57) 27.9, p 0.001). There were additionally significant differences between the groups on all factors of the schizotypy scale, as well as overall schizotypy scores (Table 1). Table 2 displays the data for low and high schizotypes on each of the 7 semantic processing measures. There were no significant group differences on the following 6 tasks: (1) definitions task (a) generate and (b) forced choice, (2) Boston naming task, (3) semantic fluency, (4) word association task, (5) synonyms, task and (6) nonsense sentences (a) RT and (b) accuracy. There was a group difference on the semantic categorization task. A 2 5 ANOVA with group (high, low) as between participant s and relatedness (high, low, borderline, related, unrelated) as the within participant s factors yielded main effects of group (F 1, , p 0.028) and relatedness (F 4, , p 0.001). There was a group relatedness interaction (F 4, , p 0.5). The main effect of group reflected poorer performance overall in the high scorers compared with the low scorers (high scorers mean 83% SD 4 and low scorers mean 79% SD 8 ). Due to our a priori interest in the effect of relatedness type we conducted further analyses to examine the strength of group differences to the different word types. The group difference was significant only for low (F 1, , p 0.02) frequency examples, reflecting poorer performance in the high schizotypes. The main effect of relatedness reflected a U-shaped performance curve with best performance for high and unrelated pairs, then low and related and poorest performance for borderline. An overall accuracy score was calculated for performance on the categorization task and was correlated with the O-LIFE subfactors and PDI. One significant correlation emerged between performance on the categories task and the cognitive disorganization subfactor of the O-LIFE (r 0.347, p 0.01). DISCUSSION Semantic processing was examined in high and low schizotypy groups using 7 different measures. Our hypothesis was only partially supported. The data revealed an impairment on only one of the measures: the semantic categorization task, in particular for low-frequency words. This deficit was found to be correlated with the cognitive disorganization subscale of the O-LIFE. High schizotypes did not have any difficulty with the other explicit semantic processing tasks like defining words, recognizing the correct definitions of words, recognizing word associations, category fluency, or detecting whether sentences were made with real or made up information. The impairment on the categories task is suggestive of idiosyncratic organization of semantic information in individuals with high schizotypy. In nonpsychosis-prone individuals, information or concepts are organized in a multidimensional semantic space, with information that has more overlapping features being closer together in semantic space; for example, dog and cat (4-legged animals) would be closer than dog and snake (a 4-legged animal and a no legged animal). This connectionist approach reduces the demands on processing incoming sensory and semantic stimuli (Hinton, 1981). Our current data imply that in people who exhibit high schizotypal traits such organization of information is different making them less effective on this task, which taps normal categorization of information. Supplementary experimentation will be needed to be completed to ascertain how high schizotypes organize information. Further, these data suggest that such deficits are very subtle, and hence are not evident on other categorical measures such as semantic fluency. The category task presumably requires quick and accurate access to semantic networks, and thus it may be this process that is also impaired or less effective in psychosis-prone individuals. However, the synonyms task is an alternative measure that presumably requires similar quick and accurate access to semantic networks, but did not demonstrate any differences between high and low scorers on the schizotypy scale. Both tasks require a simple binary decision to an explicit semantic relationship (categories is it a member or not; synonyms is it a synonym or not). The categories task is, 2009 Lippincott Williams & Wilkins 235

5 Morgan et al. The Journal of Nervous and Mental Disease Volume 197, Number 4, April 2009 TABLE 2. Data From the Semantic Processing Tasks of High and Low Schizotypy Groups Task Variables High Low Definition Generate % correct 73.0 (10.4) 73.9 (12.5) Forced choice % correct 87.6 (9.7) 88.7 (6.5) BNT % correct 89.2 (11.1) 90.4 (8.2) Semantic fluency No. responses 66.1 (13.7) 69.0 (12.6) Semantic categorization % correct High 93.4 (14.7) 97.2 (4.0) a Low 83.3 (12.6) 89.6 (6.9)* Borderline 53.6 (16.3) 59.2 (12.6) a Related 72.4 (14.6) 72.4 (12.9) Unrelated 93.6 (5.8) 96.0 (4.3) Word associations % correct 87.1 (6.6) 88.2 (6.4) Synonyms % correct 93.4 (3.5) 91.3 (5.8) RT (141.5) (146.2) Nonsense task % correct 75.6 (9.5) 73.3 (8.2) RT (108.1) (121.5) a Trend toward significance. *p BNT indicates Boston naming task. however, a slightly more complex task, in that, it was designed to have difficult conditions where the answer was increasingly not obvious, i.e., high then low then borderline typicality. In this study, semantic deficits were only present in high schizotypes on a task examining subtle semantic associations. The results are unlikely to be a product of task difficulty as people with schizophrenia perform well above chance on the same task and all the words included would be expected to be in the lexicon of an individual of average IQ and education. Further, differences between the category and synonyms tasks relate to the fact that synonyms in this task are connected both by a category and by an associative relationship (e.g., shovel spade), whereas in the category task, word pairs are connected only by a category relationship (tool spade). This may have reduced the strength of semantic activation produced in the categories task, reinforcing the subtle nature of deficit, which is only evident in conditions of high demand on semantic processes. Interestingly, although there was an overall group difference in performance on the categories task, when we examined the data for each type of stimuli, a performance decrement was most evident in high schizotypes for low-frequency/typicality words. Speculatively, this finding may relate to both the amnesia literature and the work of Rossell and David (2006); they suggest that in an impaired semantic memory store the low-frequency words or concepts are lost first. It may be that in addition to a semantic store that is idiosyncratically organized, which results in group differences, there is some very low-level degradation of semantic memory in these high schizotypes. As there is evidence from the schizophrenia literature that such, yet much more profound, degradation occurs in this disorder; it would be interesting to examine the progression of any such deficits over time. It is also plausible that such deficits were not observed on other tasks as they were less demanding, and in such instances high schizotypy scorers may have invoked compensatory strategies. Previous research has suggested that semantic deficits are related to both thought disorder (Spitzer et al., 1993) and delusions (Rossell et al., 1999). The correlation that emerged in this study between performance on the categorization task and cognitive disorganization suggests that the former may be the case. Previous associations between semantic priming and language processing difficulties have also been found (Moritz et al., 2001), and have been interpreted as links between thought disorder and semantic processing abnormalities. However, other recent research has suggested that while there may be restricted higher order semantic processing in patients with thought disorder, executive deficits may better explain the development of these symptoms (Barrera et al., 2005; Stirling et al., 2006). Thus, additional work examining the relationship between semantic processing and schizophrenia symptoms is clearly needed. It is important to note that the factor of cognitive disorganization on the O-LIFE is thought to represent not only tangentiality of thought but also distractibility (Mason et al., 1997). Thus, the possibility remains that impairment on the category task could be related to an attentional disruption. However, it is difficult to explain why this would not impact the performance of the other cognitive measures if it were the case. In addition, it has been noted that components of the O-LIFE do not directly correspond to specific schizophrenia symptomatology (Mason et al., 1997); therefore, perhaps, assessing these individuals with measures of symptomatology used with patients with schizophrenia such as the Thought and Language Communication (TLC) scale may clarify this issue further. Recent work with people with high schizotypy and semantic priming has shown an association between impaired semantic priming and unusual perceptual experiences (Morgan et al., 2006). It may be that more automatic semantic processing, e.g., priming is related to different symptomatology than that which requires explicit responses, e.g., the categories task. The finding of impaired semantic categorization conflicts to some degree with previous research that found no behavioral differences on a similar task with high and low schizotypes (Kiang and Kutas, 2005). The task employed in this study was more demanding than that used by Kiang and Kutas (2005) with shorter SOAs and the inclusion of the borderline condition. Furthermore, despite no behavioral differences, the Kiang study did find evidence of abnormalities in the N400 event-related potential (ERP) in high scoring schizotypy participants compared with low scorers. The N400 ERP was reduced in response to the presentation of related stimuli compared with unrelated stimuli and is thought to represent the Lippincott Williams & Wilkins

6 The Journal of Nervous and Mental Disease Volume 197, Number 4, April 2009 Semantics and Schizotypy integration of stimuli into an unfolding context (Rossell et al., 2003). Kiang and Kutas suggested that there was an association between schizotypy and impaired context use, which lead to larger N400 potentials for both highly related category exemplars and low related exemplars. It has been suggested that such a deficit in maintaining internal representations may exist in schizophrenia (Cohen and Servan-Schreiber, 1992). While semantic priming in this group of schizotypal participants was also impaired (Morgan et al., 2006), other semantic tasks used in this battery require the use of context (e.g., synonyms task, word association task) but did not show deficits. It may be that in highly schizotypal individuals, deficits in the use of context are only evident on subtler measures, as would be supported by the absence of behavioral effects in a less demanding categorization task (Kiang and Kutas, 2005). The pattern of performance observed on the categorization task in high schizotypes is complementary to those results observed on the same task in patients with schizophrenia. Rossell and David (2006) reported that patients with schizophrenia had the following accuracy scores on the semantic categorization task: high 88.2%, low 73.2%, borderline 55.0%, related 49.3%, and unrelated 79.3%. The scores of the high schizotypy group in the current study fall between this schizophrenia sample and the low schizotypy group. Thus, our results also confirm that individuals with high schizotypy can be viewed as being on a continuum between normality and frank psychosis, overt semantic deficits are not present but subtle deficits are. No deficits were observed on any other of the explicit semantic processing measures. While most of these had not previously been conducted in the context of schizotypy, the verbal fluency results we observed here do conflict with previous research, which has found verbal fluency deficits in high schizotypes (Barrantes-Vidal et al., 2003). There are 2 possible reasons for this difference. First, previous research used a letter fluency task and we used a category fluency task. Second, previous research used a large sample. We chose to adopt a high vulnerability/low vulnerability approach with smaller groups as this has more consistently demonstrated patterns of impairment on specific cognitive tasks. It may have been that had we used a larger sample and a correlational analysis, then such differences would have emerged. The possibility also remains that due to the generally high IQ of participants in both groups there were ceiling effects for some of the measures. Participants were more accurate on the definition forcedchoice test, the Boston naming test, the word association test, and the synonyms test than on the categories test. However, this is not true for the nonsense task or the definitions generate task. Thus, our data is inconclusive with regards to possible ceiling effects. CONCLUSIONS In summary, this study demonstrated a very specific deficit on only 1 of 7 semantic tasks used. Further, this was particularly evident for low-frequency/typicality words. Such subtle semantic processing deficits may be a marker for vulnerability to development of psychotic symptomatology, in particular, disorganized symptoms. It is suggested that, as such, semantic deficits progress they may result in the complete experience of symptoms such as thought disorder or delusions. Interesting areas to address in future research are why the deficits do not progress in some of these individuals, and how they are manifest in the prodromal stages of schizophrenia. These data suggest that, unlike patients with schizophrenia, semantic processing is not globally impaired in individuals with high schizotypy. ACKNOWLEDGMENTS The authors thank Prof. Anthony David for his support. REFERENCES Avons SE, Nunn JA, Chan L, Armstrong H (2003) Executive function assessed by memory updating and random generation in schizotypal individuals. Psychiatry Res. 120: Barrantes-Vidal N, Fananas L, Rosa A, Caparros B, Dolors Riba M, Obiols JE (2003) Neurocognitive, behavioural and neurodevelopmental correlates of schizotypy clusters in adolescents from the general population. Schizophr Res. 61: Barrera A, McKenna PJ, Berrios GE (2005) Formal thought disorder in schizophrenia: An executive or a semantic deficit? Psychol Med. 35: Battig WF, Montague WE (1969) Category norms of verbal items in 56 categories: A replication and extension of the Connecticut category norms. J Exp Psychol. 80(3 pt 2):1 46. Beech A, McManus D, Baylis G, Tipper S, Agar K (1991) Individual differences in cognitive processes: Towards an explanation of schizophrenic symptomatology. Br J Psychol. 82(pt 4): Bentall RP, Claridge GS, Slade PD (1989) The multidimensional nature of schizotypal traits: A factor analytic study with normal subjects. Br J Clin Psychol. 28: Chen E, Wilkins AJ, McKenna PJ (1994) Semantic memory is both impaired and anomalous in schizophrenia. Psychol Med. 24: Clare L, McKenna PJ, Mortimer AM, Baddeley AD (1993) Memory in schizophrenia: What is impaired and what is preserved? Neuropsychologia. 31: Claridge G, Hewitt JK (1987) A biometrical study of schizotypy in a normal population. Pers Individ Diff. 8: Cohen JD, Servan-Schreiber D (1992) Context, cortex and dopamine: A connectionist approach to behavior and biology in schizophrenia. Psychol Rev. 99: Duchene A, Graves RE, Brugger P (1998) Schizotypal thinking and associative processing: A response commonality alaysis of verbal fluency. J Psychiatry Neurosci. 23: Gooding DC, Kwapil TR, Tallent KA (1999) Wisconsin Card Sorting Test deficits in schizotypic individuals. Schizophr Res. 40: Gooding DC, Tallent KA, Hegyi JV (2001) Cognitive slippage in schizotypic individuals. J Nerv Ment Dis. 189: Hampton JA, Gardiner MM (1983) Measures of internal category structure: A correlational analysis of normative data. Br J Psychol. 74: Hinton GE (1981). Implementing semantic networks in parallel hardware. In GE Hinton, JA Anderson (Eds), Parallel Models of Associative Memory (pp ). Hillsdale (NJ): Erlbaum. Joyce EM, Collinson SL, Crichton P (1996) Verbal fluency in schizophrenia: Relationship with executive function, semantic memory and clinical alogia. Psychol Med. 26: Kaplan E, Goodglass H, Weintraub S (1983) Boston Naming Test. Malvern (PA): Lea & Febigcr. Kay J, Lesser R, Coltheart M (1992) Psycholinguistic Assessment of Language Processing in Aphasia. London: Lawrence Erlbaum. Kiang M, Kutas M (2005) Association of schizotypy with semantic processing differences: An event-related brain potential study. Schizophr Res. 77: Kravetz S, Faust M, Edelman A (1998) Dimensions of schizotypy and lexical decision in the two hemispheres. Pers Individ Diff. 25: Kucera H, Francis WN (1967) Computational Analysis of Present-Day American English. Providence (RI): Brown University Press. Laurent A, Moreaud O, Bosson JL, Naegele B, Boucharlat J, Saoud M, Dalery J, D Amato T (1999) Neuropsychological functioning among non-psychotic siblings and parents of schizophrenic patients. Psychiatry Res. 87: Lenzenweger MF, Cornblatt BA, Putnick M (1991) Schizotypy and sustained attention. J Abnorm Psychol. 100: Mason O, Claridge G, Clark K (1997) Electrodermal relationships with personality measures of psychosis-proneness in psychotic and normal subjects. Int J Psychophysiol. 27: Mason O, Claridge G, Jackson M (1995) New scales for the assessment of schizotypy. Pers Individ Diff. 18:7 13. Miller EN, Chapman LJ (1983) Continued word association in hypothetically psychosis-prone college students. J Abnorm Psychol. 92: Morgan CJA, Bedford N, Rossell SL (2006) Evidence of semantic disorganization using semantic priming in individuals with high schizotypy. Schizophr Res. 84: Moritz S, Mersmann K, Kloss M, Jacobsen D, Wilke U, Andresen B, Naber D, Pawlik K (2001) Hyper-priming in thought-disordered schizophrenic patients. Psychol Med. 31: Lippincott Williams & Wilkins 237

7 Morgan et al. The Journal of Nervous and Mental Disease Volume 197, Number 4, April 2009 Moritz S, Andresen B, Domin F, Martin T, Probsthein E, Kretschmer G, Krausz M, Naber D, Spitzer M (1999) Increased automatic spreading activation in healthy subjects with elevated scores in a scale assessing schizophrenic language disturbances. Psychol Med. 29: Nelson HE (1982) The National Adult Reading Test (NART). Windsor (UK): NFER. Nunn J, Peters E (2001) Schizotypy and patterns of lateral asymmetry on hemisphere-specific language tasks. Psychiatry Res. 103: Obiols JE, Clos M, Corbero E, Garcia-Domingo M, de Trincheria, I, Domenech E (1992). Sustained attention deficit in young schizophrenic and schizotypic men. Psychol Rep. 71(3 pt 2): Oldfield RC (1971) The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia. 9: Peters ER, Joseph SA, Garety PA (1999) Measurement of delusional ideation in the normal population: Introducing the PDI (Peters et al. Delusions Inventory). Schizophr Bull. 25: Peters ER, Pickering AD, Hemsley DR (1994) Cognitive inhibition and positive symptomatology in schizotypy. Br J Clin Psychol. 33(pt 1): Rossell SL, Batty R (2008) Evidence of semantic disorganisation in schizophrenia and bi-polar disorder using a new word definition task. Schizophr Res. In press. Rossell SL, David AS (2006) Are semantic processing deficits in schizophrenia an access or storage problem? Schizophr Res. 82: Rossell SL, Price CJ, Nobre AC (2003) The anatomy and time course of semantic priming investigated by fmri and ERPs. Neuropsychologia. 41: Rossell SL, Shapleske J, David AS (2000) Direct and indirect semantic priming with neutral and emotional stimuli in schizophrenia: Relationship with delusions. Cogn Neuropsychiatry. 5: Rossell SL, Rabe-Hesketh S, Shapleske J, David AS (1999) Is semantic fluency differentially impaired in schizophrenic patients with delusions? J Clin Exp Neuropsychol. 21: Rossell SL, Shapleske J, David AS (1998) Sentence verification and delusions: A content-specific deficit. Psychol Med. 28: Spitzer M, Braun U, Hermle L, Maier S (1993) Associative semantic network dysfunction in thought-disordered schizophrenic patients: Direct evidence from indirect semantic priming. Biol Psychiatry. 34: Stirling J, Hellewell J, Blakey A, Deakin W (2006) Thought disorder in schizophrenia is associated with both executive dysfunction and circumscribed impairments in semantic function. Psychol Med. 36: Suhr JA (1997) Executive functioning deficits in hypothetically psychosis-prone college students. Schizophr Res. 27: Tallent KA, Gooding DC (1999) Working memory and Wisconsin Card Sorting Test performance in schizotypic individuals: A replication and extension. Psychiatry Res. 89: Tamlyn D, McKenna PJ, Mortimer AM, Lund CE, Hammond S, Baddeley AD (1992) Memory impairment in schizophrenia: Its extent, affiliations and neuropsychological character. Psychol Med. 22: Trestman RL, Keefe RS, Mitropoulou V, Harvey PD, devegvar ML, Lees-Roitman S, Davidson M, Aronson A, Silverman J, Siever LJ (1995) Cognitive function and biological correlates of cognitive performance in schizotypal personality disorder. Psychiatry Res. 59: Lippincott Williams & Wilkins