Schizophrenia Research, 8 (1992) 171- 18 1 0 1992 Elsevier Science Publishers B.V. All rights
SCHIZO
171 reserved
0920-9964/92/$05.00
00263
Semantic priming of word pronunciation schizophrenia Sophia “Department
of Psychiatry.
Vinogradov”,
California
Pac$c
Beth
Medical
A. Oberb
Center
(Par@
and Campus),
and lexical decision in
Gregory
K. Shenaut”
and Medical
Research
Institute
qf San
Francisco,
San Francisco, CA, USA bDivi.~ion of Human Development, Department of Applied Behavioral Sciences, University qf’ California, Davis, CA, USA and Department of Veterans Affairs Medical Centers. Livermore and Martinez. CA, USA and ‘Department of Veterans Afairs Medical Center, Marlinez, CA, USA (Received
17 February
1992; revision
received
16 July 1992; accepted
27 July 1992)
Experimental assessments of semantic memory structure and function in schizophrenic subjects can be a useful approach for delineating some of the information processing deficits in schizophrenia. In this study, a pronunciation and a lexical decision semantic priming experiment were conducted with 19 schizophrenic subjects and 20 normal controls. A short stimulus-onset asynchrony (250 msec) and a relatively low proportion of related prime-target pairs were used in order to examine automatic priming and in order to avoid the contribution of attentional, controlled processes. On the pronunciation task, schizophrenic subjects showed a significant priming effect, equal to the priming shown by normal controls. However, on the lexical decision task, schizophrenics, unlike normal controls, did not show a priming effect which is significantly greater than zero, even though the group difference in priming effect (interaction of priming effect by group) was nonsignificant. The lack of priming on the lexical decision task is consistent with the hypothesis that schizophrenic subjects may show abnormalities in the realm of post-lexical, controlled information processing. The equal-to-normal priming for schizophrenic subjects indicates that the basic structure of the semantic network, including associations among related concepts, is intact in schizophrenia, and that spreading activation also occurs normally. Key words: Semantic memory;
Semantic
priming;
(Schizophrenia)
INTRODUCTION
Over the years, various attentional and associational disturbances have been proposed as major contributing factors to the phenomenology of schizophrenia (Frith, 1979; Reed, 1970; Callaway and Naghdi, 1982; Schwartz, 1982; Maher et al., 1983; Cohen and Servan-Schreiber, 1990). Many models of abnormal language and memory processing have been put forth to explain the disorgaCorrespondence to: S. Vinogradov, Outpatient Psychiatry Department, California Pacific Medical Center (Pacific Campus), 2323 Sacramento Street, San Francisco, CA 94115, USA,
nized speech patterns, auditory hallucinations, and cognitive dysfunctions observed in schizophrenic patients (Salzinger et al., 1970; Maher, 1983; Hoffman, 1986). Although there is widespread agreement that some sort of informationprocessing deficit exists in schizophrenia, a great deal of controversy remains as to the nature of the impairments and their role in schizophrenic behavior. The concept of semantic memory
Cognitive psychologists have been developing models of the associative network of permanently stored knowledge which builds up over an individual’s lifetime. This associative network is often
172
referred to as semantic memory (Schacter and Tulving, 1982; Tulving, 1972). This permanent, semantic memory system is accessed continually during all types of cognitive processing, including perception, learning, language production and comprehension, and problem solving. Therefore, any structural or functional abnormalities in the semantic memory system would be expected to have negative consequences for cognitive processing. Several researchers have demonstrated that schizophrenia patients show abnormalities in access and/or retrieval of concepts from semantic memory, in response to a contextual cue or ‘prime’ (Blanley, 1974; Chapman et al., 1976; Murchie and Weckowicz, 1980; Manschrek et al., 1988; Kwapil et al., 1990). Experimental assessments of semantic memory structure and function in schizophrenic patients would thus seem to be a useful approach for delineating some of the information processing deficits in schizophrenia. However, before providing the rationale for our semantic memory studies with schizophrenics, we will provide some background information about semantic memory models and the experimental paradigm of semantic priming. Spreading activation models of semantic memory The semantic memory process of spreading activation has been the focus of numerous studies with young normal subjects and more recently, with elderly normals (Burke et al., 1987; Howard et al., 1981), Alzheimer’s disease (Chertkow et al., 1989; Nebes et al., 1984; Ober and Shenaut, 1988; Ober et al., 1991), and aphasia (Milberg and Blumstein, 1981) subjects. The semantic memory network is generally viewed as a network of conceptual nodes connected by relational links. For example, the concepts APPLE, PEAR, and ORANGE are all linked to the superordinate concept FRUIT. Furthermore, a set of features is presumed to be stored with each conceptual node, including sensory and semantic features, as well as verbal labels (Anderson, 1983; Anderson, 1984; Collins and Loftus, 1975). This semantic memory structure constitutes a network for spread of activation between connected conceptual nodes (Fig. 1). When a specific node (e.g., APPLE) is activated in memory, say, by external stimulation, such as hearing the concept or word referred to in speech,
activation spreads to related concepts or nodes (ORANGE, PEAR, etc.) through the relational links connecting them. The closer a given concept or node is to an activated node, the more activation it will receive. Useful models for understanding some aspects of human cognitive dysfunction, such as those that occur in Alzheimer’s disease and in schizophrenia, can thus be generated by postulating changes in the characteristics of such spreading activation networks (Ober and Shenaut. 1988; Cohen and Servan-Schreiber, 1992). Semantic priming experiments and semantic priming t$ects Semantic priming experiments provide one means of gathering empirical data based on predictions made from a spreading activation network model of cognition. Semantic priming describes the phenomenon of increasing the speed with which a target word is pronounced, or with which a string of letters is recognized as a word (referred to as a lexical decision task), by presenting to the subject a semantically related word or ‘prime’ prior to the appearance of the target word (Neely, 1991). One widely accepted explanation of this phenomenon is that activation has spread from the prime word’s conceptual node (for example, APPLE) to the related target word’s conceptual node (PEAR) through links between nodes in the semantic network, thus making the target word more accessible for pronunciation or recognition (Collins and Loftus, 1975). The reaction times for pronouncing or recognizing target words preceded by related versus unrelated primes can be measured and compared. Semantic priming effects occur when there is a reduction in reaction time to a target word preceded by a related prime as compared to a target word preceded by an unrelated prime (Figs. 2 and 3). When semantic priming effects are observed, they are assumed to be the result of two levels of processes: an automatic process, the spread of activation in the semantic network as we have already described above; and at least two controlled processes (Posner and Snyder, 1975; Neely, 1977) (Fig, 4). The automatic process of activation spread requires no attentional capacity and begins immediately after access of the prime. The controlled processes, involving attentional capacity and strategic factors, consist of: (1) a pre-lexical
173
Fruit
node
Fig. I. Example
of a hypothetical
process, expectancy, which can operate in both pronunciation and lexical decision tasks if there is at least 400 msec between prime and target onset, especially if there is a high proportion of related pairs among the experimental trials; and (2) a postlexical process, semantic matching (Neely, 1977, 199 I). (These controlled processes will be discussed in greater detail in the Discussion section.) Rationale for the study In this pilot study, we conducted a pronunciation and a lexical decision semantic priming experiment on 19 schizophrenics and 20 normal controls. We used a short stimulus-onset asynchrony (SOA) of 250 msec and a relatively low proportion of related prime-target pairs in order to examine automatic priming (i.e. spreading activation), and in order to avoid, as much as possible, the contribution of controlled processes. Our aim was to answer the following two questions: (1) do schizophrenic subjects show normal automatic spread of activation in the semantic network when compared to normal control subjects? and (2) will the schizophrenic subjects show normal priming effects on a lexical decision task as well as a pronunciation task, even though the lexical decision task may afford the
semantic
network
(category
of ‘fruit’).
opportunity for some post-lexical, controlled information processing? Schizophrenic patients are known to show significantly slower simple reaction times (RTs) overall than control subjects (De Amicis and Cromwell, 1979; Manuzza, 1980). Therefore, we also expected to see slower overall RTs for the schizophrenics compared to normal subjects in these priming experiments. However, our main interest was in the priming effects, (which are RT d@rences between targets preceded by related versus unrelated primes (Figs. 2 and 3) and particularly, in the comparison of priming effects across two different tasks.
METHODS Subjects Seventeen male and two female schizophrenic outpatients were recruited from Mental Hygiene Clinics at the Department of Veterans Affairs Medical Centers in Palo Alto and Livermore, CA. Their ages ranged from 21 to 40 years with a mean of 3 1.58 (SD = 5.35). All subjects carried long-term diagnoses of chronic schizophrenia based on the
174
Related prime-target
pair
Unrelated
prime-target
pair
Word shown first to subject on computer screen (the “prime”)
stimulus onset asynchrony (SOA)
Word shown second to subject (the “target”)
reaction time (RT)
Subject reads last word aloud
Word read aloud:
Semantic priming effect
Word read aloud:
Fig. 2. Schematic
illustration
of a semantic-priming
DSM-IIIR and further received Research and Diagnostic Criteria (RDC) diagnoses of schizophrenia, either chronic undifferentiated subtype or chronic paranoid subtype (Spitzer et al., 1978). RDC diagnoses were based on chart review and at least two clinical interviews by one of the authors (S.V.). The subjects carried no other psychiatric or neurologic diagnoses. Individuals with a history of head trauma or with current psychoactive substance use elicited by self-report, chart review, or clinical impression of primary treating psychiatrist, were excluded from the study. Mean length of illness was eight years (range 3-20 years). Schizophrenic subjects had an average educational level of 12.63 years (SD= 1.46). In this pilot study, we did not attempt to control for the effect of medication status; however, all subjects were maintained at the lowest possible dose of antipsychotic medications for at least two weeks prior to testing. Six subjects were medication-free for one week prior to testing, while the
experiment
involving
pronunciation
remaining 13 received antipsychotic medication at an average daily dose of 300 mg of chlorpromazine equivalents, with a range of 100~600 mg. Five of these individuals also received low-dose adjunctive anticholinergic medication to control extrapyramida1 side effects. No subject was on any other psychotropic agent. All subjects showed symptoms of at least mild to moderate chronic psychosis, and at the time of testing, the mean score on the Clinical Global Impressions Scale was 3-moderate, with a range of 2-5. The most frequent symptoms displayed during clinical interviews included paranoid ideation and overinclusive thinking (19 subjects), bizarre or unusual ideas (17 subjects), chronic auditory hallucinations (12 subjects) and disorganized speech patterns (IO subjects). All subjects gave informed consent and were paid for their participation. Twenty-two control subjects (10 males, 12 females) participated in the pronunciation experiment, and 20 of these subjects also participated in
175 Related primetarget pair
Neutral primetarget pair
Unrelated primetarget pair
=T “blank”
Stimulus
onset
Word shown second to subject: the “target”
“orange”
reaction time
reaction time CRT)
CRT)
Subject decides if this last word is a real word or a non-word
total priming
reaction time (RT)
-
effect
Total Priming Effect - RT (unrelated) - RT (related)
Interference = RT (unrelated) - RT (neutral) Fig. 3. Example
Semantic
Automatic Processes Spreadof BCIIVBllO”
Priming
Effects
Controlled Processes
of lexical decision
tasks.
the lexical decision experiment. The control subjects were recruited from two local junior colleges. For the 22 young controls, the mean years of age was 20.14 (SD=2.73) and the mean years of education was 13.3 1 (SD = 1.09). Informed consent was obtained from all control subjects and they were paid for their participation. The two subject groups differed significantly in mean years of age (t=8.81, p
SCHIZO
171 reserved
0920-9964/92/$05.00
00263
Semantic priming of word pronunciation schizophrenia Sophia “Department
of Psychiatry.
Vinogradov”,
California
Pac$c
Beth
Medical
A. Oberb
Center
(Par@
and Campus),
and lexical decision in
Gregory
K. Shenaut”
and Medical
Research
Institute
qf San
Francisco,
San Francisco, CA, USA bDivi.~ion of Human Development, Department of Applied Behavioral Sciences, University qf’ California, Davis, CA, USA and Department of Veterans Affairs Medical Centers. Livermore and Martinez. CA, USA and ‘Department of Veterans Afairs Medical Center, Marlinez, CA, USA (Received
17 February
1992; revision
received
16 July 1992; accepted
27 July 1992)
Experimental assessments of semantic memory structure and function in schizophrenic subjects can be a useful approach for delineating some of the information processing deficits in schizophrenia. In this study, a pronunciation and a lexical decision semantic priming experiment were conducted with 19 schizophrenic subjects and 20 normal controls. A short stimulus-onset asynchrony (250 msec) and a relatively low proportion of related prime-target pairs were used in order to examine automatic priming and in order to avoid the contribution of attentional, controlled processes. On the pronunciation task, schizophrenic subjects showed a significant priming effect, equal to the priming shown by normal controls. However, on the lexical decision task, schizophrenics, unlike normal controls, did not show a priming effect which is significantly greater than zero, even though the group difference in priming effect (interaction of priming effect by group) was nonsignificant. The lack of priming on the lexical decision task is consistent with the hypothesis that schizophrenic subjects may show abnormalities in the realm of post-lexical, controlled information processing. The equal-to-normal priming for schizophrenic subjects indicates that the basic structure of the semantic network, including associations among related concepts, is intact in schizophrenia, and that spreading activation also occurs normally. Key words: Semantic memory;
Semantic
priming;
(Schizophrenia)
INTRODUCTION
Over the years, various attentional and associational disturbances have been proposed as major contributing factors to the phenomenology of schizophrenia (Frith, 1979; Reed, 1970; Callaway and Naghdi, 1982; Schwartz, 1982; Maher et al., 1983; Cohen and Servan-Schreiber, 1990). Many models of abnormal language and memory processing have been put forth to explain the disorgaCorrespondence to: S. Vinogradov, Outpatient Psychiatry Department, California Pacific Medical Center (Pacific Campus), 2323 Sacramento Street, San Francisco, CA 94115, USA,
nized speech patterns, auditory hallucinations, and cognitive dysfunctions observed in schizophrenic patients (Salzinger et al., 1970; Maher, 1983; Hoffman, 1986). Although there is widespread agreement that some sort of informationprocessing deficit exists in schizophrenia, a great deal of controversy remains as to the nature of the impairments and their role in schizophrenic behavior. The concept of semantic memory
Cognitive psychologists have been developing models of the associative network of permanently stored knowledge which builds up over an individual’s lifetime. This associative network is often
172
referred to as semantic memory (Schacter and Tulving, 1982; Tulving, 1972). This permanent, semantic memory system is accessed continually during all types of cognitive processing, including perception, learning, language production and comprehension, and problem solving. Therefore, any structural or functional abnormalities in the semantic memory system would be expected to have negative consequences for cognitive processing. Several researchers have demonstrated that schizophrenia patients show abnormalities in access and/or retrieval of concepts from semantic memory, in response to a contextual cue or ‘prime’ (Blanley, 1974; Chapman et al., 1976; Murchie and Weckowicz, 1980; Manschrek et al., 1988; Kwapil et al., 1990). Experimental assessments of semantic memory structure and function in schizophrenic patients would thus seem to be a useful approach for delineating some of the information processing deficits in schizophrenia. However, before providing the rationale for our semantic memory studies with schizophrenics, we will provide some background information about semantic memory models and the experimental paradigm of semantic priming. Spreading activation models of semantic memory The semantic memory process of spreading activation has been the focus of numerous studies with young normal subjects and more recently, with elderly normals (Burke et al., 1987; Howard et al., 1981), Alzheimer’s disease (Chertkow et al., 1989; Nebes et al., 1984; Ober and Shenaut, 1988; Ober et al., 1991), and aphasia (Milberg and Blumstein, 1981) subjects. The semantic memory network is generally viewed as a network of conceptual nodes connected by relational links. For example, the concepts APPLE, PEAR, and ORANGE are all linked to the superordinate concept FRUIT. Furthermore, a set of features is presumed to be stored with each conceptual node, including sensory and semantic features, as well as verbal labels (Anderson, 1983; Anderson, 1984; Collins and Loftus, 1975). This semantic memory structure constitutes a network for spread of activation between connected conceptual nodes (Fig. 1). When a specific node (e.g., APPLE) is activated in memory, say, by external stimulation, such as hearing the concept or word referred to in speech,
activation spreads to related concepts or nodes (ORANGE, PEAR, etc.) through the relational links connecting them. The closer a given concept or node is to an activated node, the more activation it will receive. Useful models for understanding some aspects of human cognitive dysfunction, such as those that occur in Alzheimer’s disease and in schizophrenia, can thus be generated by postulating changes in the characteristics of such spreading activation networks (Ober and Shenaut. 1988; Cohen and Servan-Schreiber, 1992). Semantic priming experiments and semantic priming t$ects Semantic priming experiments provide one means of gathering empirical data based on predictions made from a spreading activation network model of cognition. Semantic priming describes the phenomenon of increasing the speed with which a target word is pronounced, or with which a string of letters is recognized as a word (referred to as a lexical decision task), by presenting to the subject a semantically related word or ‘prime’ prior to the appearance of the target word (Neely, 1991). One widely accepted explanation of this phenomenon is that activation has spread from the prime word’s conceptual node (for example, APPLE) to the related target word’s conceptual node (PEAR) through links between nodes in the semantic network, thus making the target word more accessible for pronunciation or recognition (Collins and Loftus, 1975). The reaction times for pronouncing or recognizing target words preceded by related versus unrelated primes can be measured and compared. Semantic priming effects occur when there is a reduction in reaction time to a target word preceded by a related prime as compared to a target word preceded by an unrelated prime (Figs. 2 and 3). When semantic priming effects are observed, they are assumed to be the result of two levels of processes: an automatic process, the spread of activation in the semantic network as we have already described above; and at least two controlled processes (Posner and Snyder, 1975; Neely, 1977) (Fig, 4). The automatic process of activation spread requires no attentional capacity and begins immediately after access of the prime. The controlled processes, involving attentional capacity and strategic factors, consist of: (1) a pre-lexical
173
Fruit
node
Fig. I. Example
of a hypothetical
process, expectancy, which can operate in both pronunciation and lexical decision tasks if there is at least 400 msec between prime and target onset, especially if there is a high proportion of related pairs among the experimental trials; and (2) a postlexical process, semantic matching (Neely, 1977, 199 I). (These controlled processes will be discussed in greater detail in the Discussion section.) Rationale for the study In this pilot study, we conducted a pronunciation and a lexical decision semantic priming experiment on 19 schizophrenics and 20 normal controls. We used a short stimulus-onset asynchrony (SOA) of 250 msec and a relatively low proportion of related prime-target pairs in order to examine automatic priming (i.e. spreading activation), and in order to avoid, as much as possible, the contribution of controlled processes. Our aim was to answer the following two questions: (1) do schizophrenic subjects show normal automatic spread of activation in the semantic network when compared to normal control subjects? and (2) will the schizophrenic subjects show normal priming effects on a lexical decision task as well as a pronunciation task, even though the lexical decision task may afford the
semantic
network
(category
of ‘fruit’).
opportunity for some post-lexical, controlled information processing? Schizophrenic patients are known to show significantly slower simple reaction times (RTs) overall than control subjects (De Amicis and Cromwell, 1979; Manuzza, 1980). Therefore, we also expected to see slower overall RTs for the schizophrenics compared to normal subjects in these priming experiments. However, our main interest was in the priming effects, (which are RT d@rences between targets preceded by related versus unrelated primes (Figs. 2 and 3) and particularly, in the comparison of priming effects across two different tasks.
METHODS Subjects Seventeen male and two female schizophrenic outpatients were recruited from Mental Hygiene Clinics at the Department of Veterans Affairs Medical Centers in Palo Alto and Livermore, CA. Their ages ranged from 21 to 40 years with a mean of 3 1.58 (SD = 5.35). All subjects carried long-term diagnoses of chronic schizophrenia based on the
174
Related prime-target
pair
Unrelated
prime-target
pair
Word shown first to subject on computer screen (the “prime”)
stimulus onset asynchrony (SOA)
Word shown second to subject (the “target”)
reaction time (RT)
Subject reads last word aloud
Word read aloud:
Semantic priming effect
Word read aloud:
Fig. 2. Schematic
illustration
of a semantic-priming
DSM-IIIR and further received Research and Diagnostic Criteria (RDC) diagnoses of schizophrenia, either chronic undifferentiated subtype or chronic paranoid subtype (Spitzer et al., 1978). RDC diagnoses were based on chart review and at least two clinical interviews by one of the authors (S.V.). The subjects carried no other psychiatric or neurologic diagnoses. Individuals with a history of head trauma or with current psychoactive substance use elicited by self-report, chart review, or clinical impression of primary treating psychiatrist, were excluded from the study. Mean length of illness was eight years (range 3-20 years). Schizophrenic subjects had an average educational level of 12.63 years (SD= 1.46). In this pilot study, we did not attempt to control for the effect of medication status; however, all subjects were maintained at the lowest possible dose of antipsychotic medications for at least two weeks prior to testing. Six subjects were medication-free for one week prior to testing, while the
experiment
involving
pronunciation
remaining 13 received antipsychotic medication at an average daily dose of 300 mg of chlorpromazine equivalents, with a range of 100~600 mg. Five of these individuals also received low-dose adjunctive anticholinergic medication to control extrapyramida1 side effects. No subject was on any other psychotropic agent. All subjects showed symptoms of at least mild to moderate chronic psychosis, and at the time of testing, the mean score on the Clinical Global Impressions Scale was 3-moderate, with a range of 2-5. The most frequent symptoms displayed during clinical interviews included paranoid ideation and overinclusive thinking (19 subjects), bizarre or unusual ideas (17 subjects), chronic auditory hallucinations (12 subjects) and disorganized speech patterns (IO subjects). All subjects gave informed consent and were paid for their participation. Twenty-two control subjects (10 males, 12 females) participated in the pronunciation experiment, and 20 of these subjects also participated in
175 Related primetarget pair
Neutral primetarget pair
Unrelated primetarget pair
=T “blank”
Stimulus
onset
Word shown second to subject: the “target”
“orange”
reaction time
reaction time CRT)
CRT)
Subject decides if this last word is a real word or a non-word
total priming
reaction time (RT)
-
effect
Total Priming Effect - RT (unrelated) - RT (related)
Interference = RT (unrelated) - RT (neutral) Fig. 3. Example
Semantic
Automatic Processes Spreadof BCIIVBllO”
Priming
Effects
Controlled Processes
of lexical decision
tasks.
the lexical decision experiment. The control subjects were recruited from two local junior colleges. For the 22 young controls, the mean years of age was 20.14 (SD=2.73) and the mean years of education was 13.3 1 (SD = 1.09). Informed consent was obtained from all control subjects and they were paid for their participation. The two subject groups differed significantly in mean years of age (t=8.81, p
