- Email: [email protected]
Semantic priming in thought disordered schizophrenic patients
Schizophreniu Elsevier
Research,
I (1988) 61-66
SRS 00002
Semantic priming in thought disordered schizophrenic patients Theo C. Manschreck’, Brendan A. Maher2, James J. Milavetz3, C. Cecily Weisstein5 and Margaret L. Schneyer’
Donna
Ames4,
‘Massachusetts General Hospital, Harvard Medical School, Boston, MA, U.S.A., 2Hartlard University, Cambridge, MA, U.S.A., 3Carleton College, Nortl$eld, MN, U.S.A.. 4T@s Unioersit) School of Medicine, Boston, MA, U.S.A., and ‘Clark Unioer,yjty, Wowhester, MA, U.S.A. (Received
14 July 1987, revised received
28 August
19X7, accepted
24 September
1987)
Groups of thought disordered (TD) and non-thought disordered (NTD) schizophrenic patients, unipolar affective patients and normal controls performed a lexical decision task involving the recognition of words immediately preceded (primed) by either an associated or an unrelated word. Significant increments in recognition speed in the associated prime condition were found in all groups, with significantly greater gain by TD schizophrenics than by others. These findings are consistent with network models of associational activation and lend support to an attentional deficit hypothesis for schizophrenic language functioning. Key words: Semantic
priming;
Lexical decision;
Thought
disorder;
INTRODUCTION
Disordered associations may contribute to many components of schizophrenic psychopathology, including disturbed language (i.e., formal thought disorder) and motor behavior (Bleuler, 1950; Laffal, 1965; Cohen et al., 1974; Arieti, 1976; Manschreck, 1983). The specific underlying mechanisms in these relationships are poorly understood. Nevertheless, the view, that an attentional disorder may play a role in the pathologies in schizophrenia, has gained substantial clinical and research support (e.g., Salzinger et al., 1964; McGhie, 1969; Maher, 1972; Chapman and Chapman, 1973). The basic argument is as follows: the spoken language of some schizophrenic patients is strewn with inappropriate associations which compromise the understandability of what is said. Such disruptions occur generally, but especially at points of
Correspondence to: T.C. Manschreck, 25 Staniford Street, Rm 521, Erich Lindemann Mental Health Center, Boston, MA 02114, U.S.A.
0920-9964.‘88.:503.50
$’ 1988 Elsevier Science Publishers
Cognition;
(Schizophrenia)
attentional focus fluctuation, such as clause boundaries and sentence endings (Maher, 1972). These intrusions (including repetitions, idiosyncratic words, and other forms of Bleulerian loosened associations) decrease the coherence of speech and become a basis for the clinical diagnosis of formal thought disorder. Studies have demonstrated that formal thought disorder is correlated with measures of speech repetitiousness and reduced predictability, and that schizophrenic groups with and without thought disorder may be distinguished on a number of cognitive tasks requiring attention and use of associations (Salzinger et al., 1964: Manschreck et al., 1981, 1984; Maher et al., 1983). This and similar models of attentional and associational disturbance put forth to explain the thought disorder of schizophrenics are consistent with developments in cognitive psychology and psycholinguistics (Chapman et al., 1964; Maher, 1972; Cohen et al., 1974). According to one such development, the network model of semantic memory (based on the study of normal cognition), each component of an utterance activates associated semantic units within a neuronal network of logogens, or nodes, and these logogens remain activated for a finite period of time (Meyer et al., 1975;
B.V. (Biomedical
Division)
62
Shastri and Feldman, 1986). This process, known as ‘spread of activation’, results in an increased probability that related units will be incorporated into subsequent phrases and utterances. However, the time period over which such semantic priming can affect the content of speech is typically a matter of milliseconds. While the details of the mechanisms that underlie this effect are unclear, there are several possibilities. One obvious one has been that activated associations in the normal speaker, while strong, are at the same time subject to some process of rapid decay that limits their potential for intrusions to a very brief time period. The speed of this decay may occur purely as a function of strength of activation. Or it may result from an immediate and effective inhibitory process that helps to terminate the activated state of a logogen after a finite interval of time (Neely, 1977). If such an inhibitory process were impaired, the activated association might indeed decay less rapidly, and hence have a greater potential for intrusion into later thought and speech. The hypotheses that associations are susceptible to activation in normal subjects, and that activation of units undergoes decay after a brief interval, have been supported by the experimental work of several investigators (Collins and Quillian, 1969; Smith ct al., 1974; Meyer et al., 1975). They propose that neuronal excitation generated by the activation of a lexical unit of utterance spreads to other semantically related units that are assumed to be stored in neuronal connection with each other. With this in mind, cognitive psychologists have developed the technique of lexical decision as a measure of associational activation (Meyer and Schvaneveldt, 1971; Meyer et al., 1975). Subjects are shown a string of letters, the turget, and asked to decide whether or not the string is a word. The target is preceded by another string, the prims, to which no response is required. The prime may or may not be a word, and if a word, may or may not be semantically associated with the target. The degree of their association constitutes the independent variable of this method; the reaction time (RT) of the response is the main dependent variable. Where the prime and target are semantically related, RTs are typically shorter than when they are not. This gain in speed of recognition, known as ,fkilitution, is hypothesized to be due to
activation by the prime of an array of semantically associated words, one of which is the target. Recognition is thus made easier as the recognitionmatching task is focused upon a more circumscribed list of alternatives. Facilitation of word recognition through paired presentation of associated words is viewed as substantiation of the theory that semantic memory is physically organized in a network of associated logogens. The nature of semantic activation among schizophrenic patients may be a key to understanding the cognitive mechanisms that lead to thought disorder. It is possible, for instance, that inappropriate intrusions arise because processes governing the activation of associations in schizophrenic subjects last longer and/or are of greater magnitude than in normal subjects. Such effects could account for the occurrence in some schizophrenics of highly incoherent language or more specific deviations such as increased repetitiousness or changes in predictability of speech. Naturally, where such effects occur among schizophrenic patients, they will contribute to the clinical diagnosis of thought disorder. To examine experimentally the activation of semantic associations in schizophrenic groups, we used a lexical decision making task, with the following predictions: (I) response time will be shorter for all groups in the associated condition compared to the non-associated condition; that is, facilitation will be demonstrable in both experimental and control groups; and (2) schizophrenic patients showing concurrent thought disorder will show a greater increase in speed (facilitation effect) than nonthought disordered schizophrenics, and unipolar and normal controls. The latter prediction should result in a significant interaction between Group and Condition (associated x non-associated).
METHODS
Subjects consisted of 11 normal controls (six female, five male), nine unipolar affective patients (seven female, two male), and 18 schizophrenic patients, 12 thought disordered (TD) (11 male, one female) and six non-thought disordered (NTD) (all male). Patients were drawn from the outpatient clinics of the Massachusetts General Hospital, where they had been followed by the senior author for a
63
Apparatus
minimum of 2 years (range: 2-13), and had been diagnosed according to DSM III criteria. Normal controls were recruited from the Boston area. Thought disorder was diagnosed with standard ratings from the Schedule for Affective Disorders and Schizophrenia (SADS) (Spitzer and Endicott, 1977). Five features were rated: understandability, derailment, logic, poverty of information conveyed, and neologisms. Neologisms were rated as present or absent; all other features were rated from 0 (not present) to 5 (extremely severe). No neologisms were obtained from any subject. The classification of TD was made when a subject obtained a rating of 3 or more on any one feature during a psychiatric interview. The kappa coefficient for interrater reliability was within the range of 0.71-0.92 when two clinicians’ rating judgments were compared for ten randomly selected subjects. None of the unipolar patients showed evidence of thought disorder. Demographic characteristics for the four groups appear in Table 1. Analyses of variance showed no significant group differences in age, handedness, or length of illness. Similarly, chlorpromazine equivalent dosages in the TD and NTD groups were not different. Two TD and four unipolar patients were taking no medication. Other unipolar patients were receiving tricyclic medications. As is frequently the case in clinical samples, the mean education in the schizophrenic group was significantly lower than in the control groups (F = 3.10, P < 0.05). However, the TD and NTD groups did not differ from each other on any of these demographic features. Because sample availability did not permit matching of groups for sex and education (factors which may be related to or influence cognitive performance by schizophrenics), post-hoc correlations and analyses of covariance were performed to verify the result in the main contrast. TABLE
Subjects were seated at a Televideo 9 12c computer terminal, next to which were two RT keys. The basic elements of the lexical decision task were as follows. A prime-target sequence was presented in which the prime was displayed for 250 ms, followed immediately by the target for 1000 ms, giving an inter-stimulus interval of 250 ms. Changes in this interval can reduce or increase facilitation (Meyer et al., 1975; Neely, 1977; Anderson, 1983). The 250 ms interval was selected for our task, as it has reliably produced maximal facilitation in reports of the investigators mentioned above. For each trial the computer recorded accuracy and RT (the interval between time of onset of the target and the depression of a response key). If the RT did not occur within 1000 ms, ‘No Response’ was recorded. Completion of the sequence for one prime-target unit was followed directly by another. Preliminary analysis on the number of ‘No Response’ trials revealed no significant main effects for Group or Condition, and no Group x Condition interaction. Word pairs were the same as those used by Meyer et al. (1975). There were ten displays in each of the two experimental prime/target conditions: prime word/associated target word (A) and prime word/non-associated target word (NA). Because lexical decision tasks require subjects to make a word-non-word choice, trials from three additional conditions (ten displays each) were randomly presented along with the A and NA display pairs: prime non-word/target word; prime word/target non-word; and prime non-word/target non-word. Thus, each subject viewed a total of 50 prime/target trials, with no indication of which type of trial would be presented at any time. A 10 s rest period was permitted between every ten trials.
1
Demographic characteristics of groups (age, education and length Group
n
Education in years
Age
Mean
of illness,handedness
~SDJ
Mean
(SD)
Normal
11
25.5
(8.6)
15.3
(2.1)
Unipolar NTD TD
9 6 12
38.1 33.8 37.0
(12.8) (5.3) (12.4)
15.4 13.1 12.7
(1.5) (2.0) (3.3)
and chlorpromazine equivalents) Length of illness
Handedness {no. left)
11.4 (8.1) 12.1 (6.8) 12.3 (11.1)
0 0 2
2
CPZ (doselday)
_ 322 (246) 450 (348)
64
Procedure
TABLE 2
Subjects were seated in front of the computer given the following instructions by the tester:
and
‘On this screen you are going to see words flashing on and off. They will come in pairs, one right after the other. Sometimes the words will be real English words. and sometimes they will be nonsense words, just jumbled letters. Your job is to decide whether or not the second word in each pair IS really a word. If you think it is, press the “YES” key; if you think it isn’t, press the “NO” key. You should try to do this as quickly and accurately as you can. Remember, you need to decide whether the second word is a real word; just ignore the first word.’
Each subject then viewed 20 practice prime/target pairs and acknowledged enough comfort with the task to proceed to the experimental conditions.
RESULTS
RT means and standard deviations for each group in the A and NA conditions appear in Table 2. Raw scores (ms) were used in the data analysis on RTs. Because sample distributions were skewed, we performed identical analyses using log RTs, according to the transformation method described by Shoben (1983). The results for the two analyses were comparable, and we therefore present only the raw data results, as these are more straightforward and simpler to interpret. Facilitation
and uccuracy
A repeated measures ANOVA was conducted with Group (Normals x Affective x NTD x TD) and Condition (A x NA) as the main effects, and RT as the dependent measure. A significant main effect for Condition was obtained (F( 1,34) = 26.23, P < O.OOl), as was a significant interaction for Group x Con-
Reaction time (RT) means and standard deviationsfor reported in milliseconds Prime-real target-real (associatedj
GrOUpS
Normal Unipolar NTD TD
all groups
Prime-real target-real (non-associated)
Mean
(SD)
Mean
605 556 594 478
(92) (116) (82) (176)
642 612. 630 561
(SD! (85)
( 146) (66) (178)
dition (F(3, 34) = 3.12, P < 0.01). The main effect for Group was not significant (F(3, 34) = 1.39). A second, identical repeated measures ANOVA was performed with accuracy as the dependent measure, which yielded no significant main effects or interaction. The accuracy means for each group in the A and NA conditions appear in Table 3. Because the task was cognitive, and because groups were not matched for sex or education, we repeated the ANOVA on RT with these factors as covariates. Similarly, despite adequate matching for age, we elected to examine age as a covariate, because of its potential impact on cognitive performance. In all three analyses the main effects for Condition and the Group x Condition interactions remained significant at the levels originally obtained, with no main effects for Group. As a final cautionary measure, we obtained Pearson correlation coefficients for these factors. The results for the experimental groups and the total subject pool revealed no relationship of either age or education with RT or accuracy scores.
TABLE 3 Mean accuracy
and error data for all groups. including mean number of trials with no response Prime-real target-real
Normal Unipolar NTD TD
Prime-real target-real
[associated)
(non-associated)
Hits
Misses
No response
Hits
Misses
No response
9.3 6.8 8.3 7.2
0.1 0.4 0.5 0.7
0.4 I.1 0.7 2.2
9.0 1.6 8.8 7.0
0.2 0.4 0.0 0.6
0.7 1.4 1.0 2.4
65
Speed-accuracy
tradeoff
The results of the above analyses support the hypotheses that RTs in the A condition would be reduced from those in the NA condition, and that TD schizophrenics would show significantly greater facilitation than the other groups. However, in choice/RT tasks it is often the case that elements in experimental conditions that might lead to shorter RTs might also create accuracy deficits. This effect is termed a speed-accuracy tradeoff, and should be avoided as it confounds results. The discovery of a significant increase in facilitation by the TD but not the NTD group with no corresponding significant differences in accuracy, suggests that the results do not reflect a speed-accuracy tradeoff. Nonetheless, as additional verification, we examined the RT x accuracy correlations directly. Here the speed-accuracy effect would appear as a positive correlation between RT and number of correct items. Pearson correlations for Normal (r= -0.31, n.s.), TD (r = -0.65, NTD (r = -0.89, P < O.Ol), and P < 0.05), Unipolar (r = -0.87, P < 0.05) subjects do not reflect such a relationship. It appears that as speed was gained (increasingly smaller RTs), accuracy was, if anything, enhanced. These data provide no support for the inference of a speed-accuracy tradeoff.
DISCUSSION
In this study greater facilitation in lexical decision was found in schizophrenic patients with concurrent thought disorder than was found in comparison groups. Facilitation occurred as predicted for all groups between the A and NA conditions. It is not wise to emphasize the importance of the main effect for Condition, as it was qualified by a significant interaction with Group. It should be noted, however, that this main effect verifies that facilitation did occur in the sample overall, and that this effect was simply more pronounced for the TD group. In addition, accuracy of decisions was unaffected by the level of association between prime and target which was varied in our conditions, and the demographic factors of age, education and sex do not appear to have confounded our results. These findings are congruent, then, with the hypothesis
that the language anomalies that contribute to the diagnosis of thought disorder in schizophrenia are related to deviations in the processes that mediate associational activation and/or inhibition. That TD schizophrenics showed the greatest degree of semantic facilitation is in accord with the ‘spreading activation’ model of semantic memory, and with the hypothesis that activation lasts longer or is of greater magnitude in TD schizophrenics than in normal subjects (Collins and Quillian, 1969). These results are not amenable to a speed-accuracy tradeoff interpretation. It is possible that other interpretations might be offered, and replication of this finding with other, larger samples is necessary. However, the data are encouraging and indicate that the technique of lexical decision has value for the study of associational anomalies in psychopathology. Schizophrenics may indeed be slow at problem solving in general, which requires complex manipulation of information. However, network activation and semantic facilitation, which may involve automatic as well as conscious processes (DenHeyer et al., 1983) may well be greater and/or longer lasting in some schizophrenic patients (i.e., those with TD). Hence a general deficit explanation for schizophrenic cognitive deficits in general is inadequate, inasmuch as it would fail to explain the accelerated performance (speed) by the TD subjects in all conditions on this task. The relationships of various clinical variables (e.g., neurological features, ventricular brain ratio, specific language anomalies and motor disturbances) and the facilitation characteristic need to be explored. For example, how does enhanced facilitation relate to Type I or Type II patterns of psychopathology (Crow, 1980), or to evidence of abnormal involuntary movements (i.e., tardive dyskinesia), with which disordered cognition has been associated (Waddington and Youssef, 1986)? Certain details of priming in schizophrenic disorders also remain to be elucidated. As mentioned earlier, changes in the interval between onset of the prime and target (termed stimulus onset asynchrony (SOA)) can affect the magnitude of facilitation effects. Experimental variation of SOAs in lexical decision studies of schizophrenia might produce more precise estimates of the time intervals involved in the activation of semantic units. Further investigations with schizophrenic patients might examine word frequency effects (i.e., typicality
66
of usage), which are commonly used in the study of normal semantic memory (Shoben, 1983). Such research would place these lexical decision results into the broader context of semantic memory for this and other populations. The observation of enhanced facilitation among the TD schizophrenics is consistent with evidence in such patients for a locus of difficulty at the level of associations (Maher, 1983). Whether altered activation, decay, or inhibition in TD schizophrenics accounts for the facilitation differences we cannot conclude at present. Nevertheless, it is tempting to speculate that Bleuler’s description (1950) of schizophrenic thinking as ‘caught in the same circle . . . of ideas’ may be based on deviant processes operating in semantic memory. REFERENCES Anderson, J. (1983) The Architecture of Cognition. Harvard University Press, Cambridge, MA. Arieti, S. (1976) The Intrapsychic Self. Basic Books, New York. Bleuler, E. (1950) Dementia Praecox, or the Group of Schizophrenias (translated by J. Zincs from the original German edition, Dementia praecox oder die Gruppe der Schizophrenien, 1911) International Universities Press, New York. Chapman, L.J. and Chapman, J.P. (1973) Disordered Thought in Schizophrenia. Prentice Hall, Englewood Cliffs, NJ. Chapman, L.J., Chapman, J.P. and Miller, G.A. (1964) A theory of verbal behavior in schizophrenia. In: B.A. Maher (Ed.), Progress in Experimental Personality Research, Vol. 1. Academic Press, New York, pp. 49-77. Cohen, B.D., Nachmani, G. and Rosenberg, S. (1974) Referent communication disturbances in schizophrenia. J. Abnorm. Psycho]. X3, 1-13. Collins, A. and Quillian, M. ( 1969) Retrieval time from semantic memory. J. Verbal Learn. Verbal Behav. 8, 240-247. Crow, T.J. (1980) Molecular pathology of schizophrenia: more than one disease process? Br. Med. J. 280, 66-6X. DenHeyer, K., Briand, K. and Dannenbring, G.L. (1983) Strategic factors in a lexical decision task: Evidence for automatic and attention-driven processes. Mem. Cognit. 13, 22X-232. Laffal, J. (1965) Normal and Pathological Language. Atherton, New York.
McGhie, A. (1969) Pathology of Attention, Penguin, Baltimore, MD. Maher, B.A. (1972) The language of schizophrenia: a review and interpretation. Br. J. Psychiatry 120, 4-17. Maher, B.A. (1983) A tentative theory of schizophrenic uttcrante. In: B.A. Maher and W.B. Maher (Eds.), Progress in Experimental Personality Research, Vol. 12. Academic Press, New York, pp. l-52. Maher, B.A., Manschreck, T.C. and Molino, M.A.C. (1983) Redundancy, pause distribution and thought disorder in schizophrenia. Lang. Speech 26, 191-199. Manschreck, T.C. (1983) Psychopathology of motor behavior in schizophrenia. In: B.A. Maher and W.B. Maher (Eds.), Progress in Experimental Personality Research, Vol. 12. Academic Press, New York, pp. 53-99. Manschreck, T.C., Maher, B.A. and Ader, D.N. (1981) Formal thought disorder, the type-token ratio, and disturbed voluntary movement in schizophrenia. Br. J. Psychiatry 139, 7 15. Manschreck, T.C., Maher, B.A., Hoover, T.M. and Ames, D. ( 1984) The type-token ratio in schizophrenic disorders: clinical and research value. Psychol. Med. 14, 151-157. Meyer, D. and Schvaneveldt, R. (1971) Facilitation in recognizing pairs of words: evidence of a dependence between retrieval operations. J. Exp. Psycho]. 90, 2277234. Meyer, D., Schvaneveldt, R. and Ruddy, M.G. (1975) Loci of contextual effects on visual word recognition. In: P. Rabbit and S. Dornic (Eds.), Attention and Performance. Academic Press, New York. Neely, J.H. (1977) Semantic priming and retrieval from lexical memory: roles of inhibitionless spreading activation and limited capacity attention. J. Exp. Psychol. Gen. 106.226-254. Salzinger, K., Portnoy, S. and Feldman. R.S. (1964) Verbal behavior of schizophrenics and normals. Ann. N.Y. Acad. Sci. 105, 8455860. Shastri, L. and Feldman, J.A. (1986) Neural nets, routines, and semantic networks. In: N.E. Sharkey (Ed.), Advances in Cognitive Science, Vol. 1. Ellis Horwood, New York. Shoben, E. (1983) Semantic and lexical decisions. In: CR. PUB (Ed.), Handbook of Research Methods in Human Memory and Cognition. Academic Press, New York. Smith, E.E., Shoben, J. and Rips. L. (1974) Structure and process in semantic memory: a featural model for semantic decisions. Psycho]. Rev. Xl, 214-241. Spitzer, R. and Endicott, J. (1977) Schedule for Affective Disorders and Schizophrenia (SADS). Biometric Research, New York. Waddington, J.L. and Youssef, H.A. ( 1986) An unusual cluster of tardive dyskinesia in schizophrenia: association with cognitive dysfunction and negative symptoms. Am. J. Psychiatry 143, 1162-l 165.
Research,
I (1988) 61-66
SRS 00002
Semantic priming in thought disordered schizophrenic patients Theo C. Manschreck’, Brendan A. Maher2, James J. Milavetz3, C. Cecily Weisstein5 and Margaret L. Schneyer’
Donna
Ames4,
‘Massachusetts General Hospital, Harvard Medical School, Boston, MA, U.S.A., 2Hartlard University, Cambridge, MA, U.S.A., 3Carleton College, Nortl$eld, MN, U.S.A.. 4T@s Unioersit) School of Medicine, Boston, MA, U.S.A., and ‘Clark Unioer,yjty, Wowhester, MA, U.S.A. (Received
14 July 1987, revised received
28 August
19X7, accepted
24 September
1987)
Groups of thought disordered (TD) and non-thought disordered (NTD) schizophrenic patients, unipolar affective patients and normal controls performed a lexical decision task involving the recognition of words immediately preceded (primed) by either an associated or an unrelated word. Significant increments in recognition speed in the associated prime condition were found in all groups, with significantly greater gain by TD schizophrenics than by others. These findings are consistent with network models of associational activation and lend support to an attentional deficit hypothesis for schizophrenic language functioning. Key words: Semantic
priming;
Lexical decision;
Thought
disorder;
INTRODUCTION
Disordered associations may contribute to many components of schizophrenic psychopathology, including disturbed language (i.e., formal thought disorder) and motor behavior (Bleuler, 1950; Laffal, 1965; Cohen et al., 1974; Arieti, 1976; Manschreck, 1983). The specific underlying mechanisms in these relationships are poorly understood. Nevertheless, the view, that an attentional disorder may play a role in the pathologies in schizophrenia, has gained substantial clinical and research support (e.g., Salzinger et al., 1964; McGhie, 1969; Maher, 1972; Chapman and Chapman, 1973). The basic argument is as follows: the spoken language of some schizophrenic patients is strewn with inappropriate associations which compromise the understandability of what is said. Such disruptions occur generally, but especially at points of
Correspondence to: T.C. Manschreck, 25 Staniford Street, Rm 521, Erich Lindemann Mental Health Center, Boston, MA 02114, U.S.A.
0920-9964.‘88.:503.50
$’ 1988 Elsevier Science Publishers
Cognition;
(Schizophrenia)
attentional focus fluctuation, such as clause boundaries and sentence endings (Maher, 1972). These intrusions (including repetitions, idiosyncratic words, and other forms of Bleulerian loosened associations) decrease the coherence of speech and become a basis for the clinical diagnosis of formal thought disorder. Studies have demonstrated that formal thought disorder is correlated with measures of speech repetitiousness and reduced predictability, and that schizophrenic groups with and without thought disorder may be distinguished on a number of cognitive tasks requiring attention and use of associations (Salzinger et al., 1964: Manschreck et al., 1981, 1984; Maher et al., 1983). This and similar models of attentional and associational disturbance put forth to explain the thought disorder of schizophrenics are consistent with developments in cognitive psychology and psycholinguistics (Chapman et al., 1964; Maher, 1972; Cohen et al., 1974). According to one such development, the network model of semantic memory (based on the study of normal cognition), each component of an utterance activates associated semantic units within a neuronal network of logogens, or nodes, and these logogens remain activated for a finite period of time (Meyer et al., 1975;
B.V. (Biomedical
Division)
62
Shastri and Feldman, 1986). This process, known as ‘spread of activation’, results in an increased probability that related units will be incorporated into subsequent phrases and utterances. However, the time period over which such semantic priming can affect the content of speech is typically a matter of milliseconds. While the details of the mechanisms that underlie this effect are unclear, there are several possibilities. One obvious one has been that activated associations in the normal speaker, while strong, are at the same time subject to some process of rapid decay that limits their potential for intrusions to a very brief time period. The speed of this decay may occur purely as a function of strength of activation. Or it may result from an immediate and effective inhibitory process that helps to terminate the activated state of a logogen after a finite interval of time (Neely, 1977). If such an inhibitory process were impaired, the activated association might indeed decay less rapidly, and hence have a greater potential for intrusion into later thought and speech. The hypotheses that associations are susceptible to activation in normal subjects, and that activation of units undergoes decay after a brief interval, have been supported by the experimental work of several investigators (Collins and Quillian, 1969; Smith ct al., 1974; Meyer et al., 1975). They propose that neuronal excitation generated by the activation of a lexical unit of utterance spreads to other semantically related units that are assumed to be stored in neuronal connection with each other. With this in mind, cognitive psychologists have developed the technique of lexical decision as a measure of associational activation (Meyer and Schvaneveldt, 1971; Meyer et al., 1975). Subjects are shown a string of letters, the turget, and asked to decide whether or not the string is a word. The target is preceded by another string, the prims, to which no response is required. The prime may or may not be a word, and if a word, may or may not be semantically associated with the target. The degree of their association constitutes the independent variable of this method; the reaction time (RT) of the response is the main dependent variable. Where the prime and target are semantically related, RTs are typically shorter than when they are not. This gain in speed of recognition, known as ,fkilitution, is hypothesized to be due to
activation by the prime of an array of semantically associated words, one of which is the target. Recognition is thus made easier as the recognitionmatching task is focused upon a more circumscribed list of alternatives. Facilitation of word recognition through paired presentation of associated words is viewed as substantiation of the theory that semantic memory is physically organized in a network of associated logogens. The nature of semantic activation among schizophrenic patients may be a key to understanding the cognitive mechanisms that lead to thought disorder. It is possible, for instance, that inappropriate intrusions arise because processes governing the activation of associations in schizophrenic subjects last longer and/or are of greater magnitude than in normal subjects. Such effects could account for the occurrence in some schizophrenics of highly incoherent language or more specific deviations such as increased repetitiousness or changes in predictability of speech. Naturally, where such effects occur among schizophrenic patients, they will contribute to the clinical diagnosis of thought disorder. To examine experimentally the activation of semantic associations in schizophrenic groups, we used a lexical decision making task, with the following predictions: (I) response time will be shorter for all groups in the associated condition compared to the non-associated condition; that is, facilitation will be demonstrable in both experimental and control groups; and (2) schizophrenic patients showing concurrent thought disorder will show a greater increase in speed (facilitation effect) than nonthought disordered schizophrenics, and unipolar and normal controls. The latter prediction should result in a significant interaction between Group and Condition (associated x non-associated).
METHODS
Subjects consisted of 11 normal controls (six female, five male), nine unipolar affective patients (seven female, two male), and 18 schizophrenic patients, 12 thought disordered (TD) (11 male, one female) and six non-thought disordered (NTD) (all male). Patients were drawn from the outpatient clinics of the Massachusetts General Hospital, where they had been followed by the senior author for a
63
Apparatus
minimum of 2 years (range: 2-13), and had been diagnosed according to DSM III criteria. Normal controls were recruited from the Boston area. Thought disorder was diagnosed with standard ratings from the Schedule for Affective Disorders and Schizophrenia (SADS) (Spitzer and Endicott, 1977). Five features were rated: understandability, derailment, logic, poverty of information conveyed, and neologisms. Neologisms were rated as present or absent; all other features were rated from 0 (not present) to 5 (extremely severe). No neologisms were obtained from any subject. The classification of TD was made when a subject obtained a rating of 3 or more on any one feature during a psychiatric interview. The kappa coefficient for interrater reliability was within the range of 0.71-0.92 when two clinicians’ rating judgments were compared for ten randomly selected subjects. None of the unipolar patients showed evidence of thought disorder. Demographic characteristics for the four groups appear in Table 1. Analyses of variance showed no significant group differences in age, handedness, or length of illness. Similarly, chlorpromazine equivalent dosages in the TD and NTD groups were not different. Two TD and four unipolar patients were taking no medication. Other unipolar patients were receiving tricyclic medications. As is frequently the case in clinical samples, the mean education in the schizophrenic group was significantly lower than in the control groups (F = 3.10, P < 0.05). However, the TD and NTD groups did not differ from each other on any of these demographic features. Because sample availability did not permit matching of groups for sex and education (factors which may be related to or influence cognitive performance by schizophrenics), post-hoc correlations and analyses of covariance were performed to verify the result in the main contrast. TABLE
Subjects were seated at a Televideo 9 12c computer terminal, next to which were two RT keys. The basic elements of the lexical decision task were as follows. A prime-target sequence was presented in which the prime was displayed for 250 ms, followed immediately by the target for 1000 ms, giving an inter-stimulus interval of 250 ms. Changes in this interval can reduce or increase facilitation (Meyer et al., 1975; Neely, 1977; Anderson, 1983). The 250 ms interval was selected for our task, as it has reliably produced maximal facilitation in reports of the investigators mentioned above. For each trial the computer recorded accuracy and RT (the interval between time of onset of the target and the depression of a response key). If the RT did not occur within 1000 ms, ‘No Response’ was recorded. Completion of the sequence for one prime-target unit was followed directly by another. Preliminary analysis on the number of ‘No Response’ trials revealed no significant main effects for Group or Condition, and no Group x Condition interaction. Word pairs were the same as those used by Meyer et al. (1975). There were ten displays in each of the two experimental prime/target conditions: prime word/associated target word (A) and prime word/non-associated target word (NA). Because lexical decision tasks require subjects to make a word-non-word choice, trials from three additional conditions (ten displays each) were randomly presented along with the A and NA display pairs: prime non-word/target word; prime word/target non-word; and prime non-word/target non-word. Thus, each subject viewed a total of 50 prime/target trials, with no indication of which type of trial would be presented at any time. A 10 s rest period was permitted between every ten trials.
1
Demographic characteristics of groups (age, education and length Group
n
Education in years
Age
Mean
of illness,handedness
~SDJ
Mean
(SD)
Normal
11
25.5
(8.6)
15.3
(2.1)
Unipolar NTD TD
9 6 12
38.1 33.8 37.0
(12.8) (5.3) (12.4)
15.4 13.1 12.7
(1.5) (2.0) (3.3)
and chlorpromazine equivalents) Length of illness
Handedness {no. left)
11.4 (8.1) 12.1 (6.8) 12.3 (11.1)
0 0 2
2
CPZ (doselday)
_ 322 (246) 450 (348)
64
Procedure
TABLE 2
Subjects were seated in front of the computer given the following instructions by the tester:
and
‘On this screen you are going to see words flashing on and off. They will come in pairs, one right after the other. Sometimes the words will be real English words. and sometimes they will be nonsense words, just jumbled letters. Your job is to decide whether or not the second word in each pair IS really a word. If you think it is, press the “YES” key; if you think it isn’t, press the “NO” key. You should try to do this as quickly and accurately as you can. Remember, you need to decide whether the second word is a real word; just ignore the first word.’
Each subject then viewed 20 practice prime/target pairs and acknowledged enough comfort with the task to proceed to the experimental conditions.
RESULTS
RT means and standard deviations for each group in the A and NA conditions appear in Table 2. Raw scores (ms) were used in the data analysis on RTs. Because sample distributions were skewed, we performed identical analyses using log RTs, according to the transformation method described by Shoben (1983). The results for the two analyses were comparable, and we therefore present only the raw data results, as these are more straightforward and simpler to interpret. Facilitation
and uccuracy
A repeated measures ANOVA was conducted with Group (Normals x Affective x NTD x TD) and Condition (A x NA) as the main effects, and RT as the dependent measure. A significant main effect for Condition was obtained (F( 1,34) = 26.23, P < O.OOl), as was a significant interaction for Group x Con-
Reaction time (RT) means and standard deviationsfor reported in milliseconds Prime-real target-real (associatedj
GrOUpS
Normal Unipolar NTD TD
all groups
Prime-real target-real (non-associated)
Mean
(SD)
Mean
605 556 594 478
(92) (116) (82) (176)
642 612. 630 561
(SD! (85)
( 146) (66) (178)
dition (F(3, 34) = 3.12, P < 0.01). The main effect for Group was not significant (F(3, 34) = 1.39). A second, identical repeated measures ANOVA was performed with accuracy as the dependent measure, which yielded no significant main effects or interaction. The accuracy means for each group in the A and NA conditions appear in Table 3. Because the task was cognitive, and because groups were not matched for sex or education, we repeated the ANOVA on RT with these factors as covariates. Similarly, despite adequate matching for age, we elected to examine age as a covariate, because of its potential impact on cognitive performance. In all three analyses the main effects for Condition and the Group x Condition interactions remained significant at the levels originally obtained, with no main effects for Group. As a final cautionary measure, we obtained Pearson correlation coefficients for these factors. The results for the experimental groups and the total subject pool revealed no relationship of either age or education with RT or accuracy scores.
TABLE 3 Mean accuracy
and error data for all groups. including mean number of trials with no response Prime-real target-real
Normal Unipolar NTD TD
Prime-real target-real
[associated)
(non-associated)
Hits
Misses
No response
Hits
Misses
No response
9.3 6.8 8.3 7.2
0.1 0.4 0.5 0.7
0.4 I.1 0.7 2.2
9.0 1.6 8.8 7.0
0.2 0.4 0.0 0.6
0.7 1.4 1.0 2.4
65
Speed-accuracy
tradeoff
The results of the above analyses support the hypotheses that RTs in the A condition would be reduced from those in the NA condition, and that TD schizophrenics would show significantly greater facilitation than the other groups. However, in choice/RT tasks it is often the case that elements in experimental conditions that might lead to shorter RTs might also create accuracy deficits. This effect is termed a speed-accuracy tradeoff, and should be avoided as it confounds results. The discovery of a significant increase in facilitation by the TD but not the NTD group with no corresponding significant differences in accuracy, suggests that the results do not reflect a speed-accuracy tradeoff. Nonetheless, as additional verification, we examined the RT x accuracy correlations directly. Here the speed-accuracy effect would appear as a positive correlation between RT and number of correct items. Pearson correlations for Normal (r= -0.31, n.s.), TD (r = -0.65, NTD (r = -0.89, P < O.Ol), and P < 0.05), Unipolar (r = -0.87, P < 0.05) subjects do not reflect such a relationship. It appears that as speed was gained (increasingly smaller RTs), accuracy was, if anything, enhanced. These data provide no support for the inference of a speed-accuracy tradeoff.
DISCUSSION
In this study greater facilitation in lexical decision was found in schizophrenic patients with concurrent thought disorder than was found in comparison groups. Facilitation occurred as predicted for all groups between the A and NA conditions. It is not wise to emphasize the importance of the main effect for Condition, as it was qualified by a significant interaction with Group. It should be noted, however, that this main effect verifies that facilitation did occur in the sample overall, and that this effect was simply more pronounced for the TD group. In addition, accuracy of decisions was unaffected by the level of association between prime and target which was varied in our conditions, and the demographic factors of age, education and sex do not appear to have confounded our results. These findings are congruent, then, with the hypothesis
that the language anomalies that contribute to the diagnosis of thought disorder in schizophrenia are related to deviations in the processes that mediate associational activation and/or inhibition. That TD schizophrenics showed the greatest degree of semantic facilitation is in accord with the ‘spreading activation’ model of semantic memory, and with the hypothesis that activation lasts longer or is of greater magnitude in TD schizophrenics than in normal subjects (Collins and Quillian, 1969). These results are not amenable to a speed-accuracy tradeoff interpretation. It is possible that other interpretations might be offered, and replication of this finding with other, larger samples is necessary. However, the data are encouraging and indicate that the technique of lexical decision has value for the study of associational anomalies in psychopathology. Schizophrenics may indeed be slow at problem solving in general, which requires complex manipulation of information. However, network activation and semantic facilitation, which may involve automatic as well as conscious processes (DenHeyer et al., 1983) may well be greater and/or longer lasting in some schizophrenic patients (i.e., those with TD). Hence a general deficit explanation for schizophrenic cognitive deficits in general is inadequate, inasmuch as it would fail to explain the accelerated performance (speed) by the TD subjects in all conditions on this task. The relationships of various clinical variables (e.g., neurological features, ventricular brain ratio, specific language anomalies and motor disturbances) and the facilitation characteristic need to be explored. For example, how does enhanced facilitation relate to Type I or Type II patterns of psychopathology (Crow, 1980), or to evidence of abnormal involuntary movements (i.e., tardive dyskinesia), with which disordered cognition has been associated (Waddington and Youssef, 1986)? Certain details of priming in schizophrenic disorders also remain to be elucidated. As mentioned earlier, changes in the interval between onset of the prime and target (termed stimulus onset asynchrony (SOA)) can affect the magnitude of facilitation effects. Experimental variation of SOAs in lexical decision studies of schizophrenia might produce more precise estimates of the time intervals involved in the activation of semantic units. Further investigations with schizophrenic patients might examine word frequency effects (i.e., typicality
66
of usage), which are commonly used in the study of normal semantic memory (Shoben, 1983). Such research would place these lexical decision results into the broader context of semantic memory for this and other populations. The observation of enhanced facilitation among the TD schizophrenics is consistent with evidence in such patients for a locus of difficulty at the level of associations (Maher, 1983). Whether altered activation, decay, or inhibition in TD schizophrenics accounts for the facilitation differences we cannot conclude at present. Nevertheless, it is tempting to speculate that Bleuler’s description (1950) of schizophrenic thinking as ‘caught in the same circle . . . of ideas’ may be based on deviant processes operating in semantic memory. REFERENCES Anderson, J. (1983) The Architecture of Cognition. Harvard University Press, Cambridge, MA. Arieti, S. (1976) The Intrapsychic Self. Basic Books, New York. Bleuler, E. (1950) Dementia Praecox, or the Group of Schizophrenias (translated by J. Zincs from the original German edition, Dementia praecox oder die Gruppe der Schizophrenien, 1911) International Universities Press, New York. Chapman, L.J. and Chapman, J.P. (1973) Disordered Thought in Schizophrenia. Prentice Hall, Englewood Cliffs, NJ. Chapman, L.J., Chapman, J.P. and Miller, G.A. (1964) A theory of verbal behavior in schizophrenia. In: B.A. Maher (Ed.), Progress in Experimental Personality Research, Vol. 1. Academic Press, New York, pp. 49-77. Cohen, B.D., Nachmani, G. and Rosenberg, S. (1974) Referent communication disturbances in schizophrenia. J. Abnorm. Psycho]. X3, 1-13. Collins, A. and Quillian, M. ( 1969) Retrieval time from semantic memory. J. Verbal Learn. Verbal Behav. 8, 240-247. Crow, T.J. (1980) Molecular pathology of schizophrenia: more than one disease process? Br. Med. J. 280, 66-6X. DenHeyer, K., Briand, K. and Dannenbring, G.L. (1983) Strategic factors in a lexical decision task: Evidence for automatic and attention-driven processes. Mem. Cognit. 13, 22X-232. Laffal, J. (1965) Normal and Pathological Language. Atherton, New York.
McGhie, A. (1969) Pathology of Attention, Penguin, Baltimore, MD. Maher, B.A. (1972) The language of schizophrenia: a review and interpretation. Br. J. Psychiatry 120, 4-17. Maher, B.A. (1983) A tentative theory of schizophrenic uttcrante. In: B.A. Maher and W.B. Maher (Eds.), Progress in Experimental Personality Research, Vol. 12. Academic Press, New York, pp. l-52. Maher, B.A., Manschreck, T.C. and Molino, M.A.C. (1983) Redundancy, pause distribution and thought disorder in schizophrenia. Lang. Speech 26, 191-199. Manschreck, T.C. (1983) Psychopathology of motor behavior in schizophrenia. In: B.A. Maher and W.B. Maher (Eds.), Progress in Experimental Personality Research, Vol. 12. Academic Press, New York, pp. 53-99. Manschreck, T.C., Maher, B.A. and Ader, D.N. (1981) Formal thought disorder, the type-token ratio, and disturbed voluntary movement in schizophrenia. Br. J. Psychiatry 139, 7 15. Manschreck, T.C., Maher, B.A., Hoover, T.M. and Ames, D. ( 1984) The type-token ratio in schizophrenic disorders: clinical and research value. Psychol. Med. 14, 151-157. Meyer, D. and Schvaneveldt, R. (1971) Facilitation in recognizing pairs of words: evidence of a dependence between retrieval operations. J. Exp. Psycho]. 90, 2277234. Meyer, D., Schvaneveldt, R. and Ruddy, M.G. (1975) Loci of contextual effects on visual word recognition. In: P. Rabbit and S. Dornic (Eds.), Attention and Performance. Academic Press, New York. Neely, J.H. (1977) Semantic priming and retrieval from lexical memory: roles of inhibitionless spreading activation and limited capacity attention. J. Exp. Psychol. Gen. 106.226-254. Salzinger, K., Portnoy, S. and Feldman. R.S. (1964) Verbal behavior of schizophrenics and normals. Ann. N.Y. Acad. Sci. 105, 8455860. Shastri, L. and Feldman, J.A. (1986) Neural nets, routines, and semantic networks. In: N.E. Sharkey (Ed.), Advances in Cognitive Science, Vol. 1. Ellis Horwood, New York. Shoben, E. (1983) Semantic and lexical decisions. In: CR. PUB (Ed.), Handbook of Research Methods in Human Memory and Cognition. Academic Press, New York. Smith, E.E., Shoben, J. and Rips. L. (1974) Structure and process in semantic memory: a featural model for semantic decisions. Psycho]. Rev. Xl, 214-241. Spitzer, R. and Endicott, J. (1977) Schedule for Affective Disorders and Schizophrenia (SADS). Biometric Research, New York. Waddington, J.L. and Youssef, H.A. ( 1986) An unusual cluster of tardive dyskinesia in schizophrenia: association with cognitive dysfunction and negative symptoms. Am. J. Psychiatry 143, 1162-l 165.