Introduction

Previous cognitive studies in persons with schizophrenia (SZ) suggest a deficit in context processing that may underlie the inability to inhibit contextually inappropriate responses1,2. This deficit has been hypothesized to underlie symptoms of disorganization, thought disorder, and errors in language and learning. In SZ, deficits in maintaining context and inhibiting inappropriate responses have been evident on measures of verbal learning, semantic priming, and sustained and divided attention3,4,5.

When healthy control (HC) subjects are presented with language samples varying in the degree of contextual dependency, recall improves with increasing context. This is described as semantic facilitation. The seminal study was reported by Miller and Selfridge6, who examined verbal learning in college students who were aurally presented word lists with increasing levels of semantic context. Verbal context (or contextual dependency) was defined as the extent to which the choice of a particular word depended on the words that preceded it. Results showed that contextual relationships extending over 5 or 6 words facilitated learning of the word list. Without context, in the short range (e.g., 3–4 words), nonsense material was learned as well as meaningful material. The authors concluded that these familiar dependencies between words facilitated learning.

Since then, investigators have found that SZ subjects profit less from context, although the extent to which this is related to symptom profile, thought disorder, and general verbal ability remains unclear7,8,9. The failure of SZ subjects to benefit from verbal context is thought to be related to a semantic processing deficit and/or to cognitive disinhibition that results in erroneous associations between words.

Semantic facilitation has not been systematically studied in clinically-defined schizotypal personality disorder (SPD). This schizophrenia spectrum disorder is thought to be biologically related to SZ10,11, with similar but less severe symptom profiles and cognitive deficits. Schizotypal personality disorder is important to study from two perspectives: first, SPD lends itself to robust interpretations of neurocognitive disorder because subjects do not carry the confounds associated with chronic illness and treatment that impact studies of SZ; second, SPD is itself considered a prevalent and disabling mental health condition. In the NIH-sponsored epidemiological survey, Pulay and colleagues12 reported a 3.9% lifetime prevalence of SPD, with highly disabling effects on social and occupational functioning.

Our laboratory reported selective deficits in verbal learning, semantic organization and abstract reasoning in SPD, suggesting greater left hemisphere and frontal networks dysfunction in this population13. Further examination of the verbal learning deficit has suggested that a primary disorder of early semantic processing and left temporal lobe structures may be a characteristic of cognitive function in males, but not females, with SPD14,15,16, a finding similar to gender differences evident in verbal processing in SZ17,18. Although verbal skills in general are lower than nonverbal skills in male SPD, specific language deficits have not been evident on clinical measures of naming and repetition13. These findings support the hypothesis that SPD, like SZ, may be characterized by disinhibition of the semantic language network, particularly in males. A deficit in context processing resulting in erroneous associations between words could give rise to SPD symptoms such as magical ideation and odd speech in addition to deficits in verbal learning.

Although semantic facilitation has not been studied in SPD, context processing deficits have been found in this group on continuous performance measures of complex attention19,20,21, suggesting similarities to SZ. One verbal memory study of young first-degree relatives of individuals with SZ has shown context-sensitive deficits compared to healthy controls22. In studies of nonclinical psychosis-prone schizotypal subjects defined by psychometric measures of schizotypy, deficits of cognitive inhibition23 and context processing24 have been suggested. However, these psychosis-prone subjects are generally healthy subjects with high scores on one or more schizotypal symptom rating scales, and the relationship of these groups to clinically-defined SPD and SZ remains unclear.

The purpose of the current study is to examine the effect of context and gender on verbal learning performance in men and women with DSM-5 clinically-defined SPD25. We used 10-word sentences from the original Miller and Selfridge6 semantic facilitation task to measure verbal learning in male and female SPD and HC subjects. We predicted that SPD subjects would benefit less from increasing verbal context than HC. In light of our previous studies showing selective verbal learning deficits in males15,16, we further predicted that this deficit would be particularly pronounced in male SPD.

Results

Repeated measures ANOVA demonstrated a significant Word List X Group X Sex interaction, F(7,121) = 2.352, p = 0.022, partial eta squared ηp2 = 0.019 (two-tailed; see Fig. 1). This result remained significant when BDI was added as a covariate. Examination of simple effects revealed a significant Word List X Group interaction for males [F(1,75) = 4.118, p = 0.046, partial eta squared ηp2 = 0.052] but not for female subjects [F(7,45) = 0.9, p = 0.515].

Fig. 1: Semantic facilitation task performance in male and female SPD and comparison groups.
Fig. 1: Semantic facilitation task performance in male and female SPD and comparison groups.
Full size image

Mean number of words recalled for each word list as a function of diagnosis and gender. Standard error (SE) bars are indicated. Note that all subjects benefit from context, but that male SPD subjects begin to perform more poorly at the 3rd order sentences, where each word is determined by the context of two preceding words. Significant differences were evident between male SPD and HC groups at the 3rd, 4th, 7th, and Text level of contextual constraint (Tukey HSD, p < 0.05). No differences were evident between female SPD and female or male HC groups.

To control for multiple comparisons, eight post hoc Tukey HSD tests were employed to examine specific differences in performance between the male SPD and HC groups (Table 1). Compared to male control subjects, male SPDs performed normally on non-meaningful sentences, but showed a deficiency in list recall beginning at the 3-order approximation list, when context increased. Post hoc pairwise Tukey tests indicated that male SPDs list recall performance was significantly different from male HCs beginning at the 3-order approximation. Significant differences were evident at the 4-order approximation, 7-order approximation, and text list as context increased. Performance at the 5-order approximation was not significantly different. In contrast, there was no difference in performance of female SPDs, female HCs, and male HCs who all showed similar rates of learning on this task.

Table 1 Results of eight post hoc Tukey HSD tests examining specific differences in list recall performance for male SPD and HC groups.
Full size table

Discussion

This study examined the effect of semantic context facilitation and gender in DSM-5 clinically-defined SPD. As predicted, on a verbal recall task including sentences with increasing semantic context, male SPDs profited less than controls from increasing context. The performance of female SPDs was similar to female controls. These findings suggest that like SZ, male SPDs have a limited ability to profit from verbal context; these results are consistent with hypotheses of deficient context processing in SZ.

These and other neuropsychological test results from our laboratory have been consistent with hypotheses of disinhibition of the semantic language network in schizophrenia-spectrum disorders26 and suggest that similar dysfunction may exist in SPD13,14,27. Such hypotheses purport that the response-bias established by semantic content may be inefficiently selective in schizophrenia, resulting in a state of “cognitive disinhibition.” Within this framework, the disturbance of language processing in SPD could stem from a lack of acquisition of semantic context or associational interdependencies of language. If the associations between words are not firmly established in SPD, this would be expected to make language samples of high contextual dependency less redundant, and thus more difficult to learn.

In addition to semantic processing deficits or disinhibition, alternative mechanisms may impact subjects’ performance when recalling word-lists with varying levels of context. These include differences in fatigue, motivation, and effort, potential cognitive weaknesses in attention, processing speed, and self-regulation, and the impact of retroactive or proactive interference on memory recall28. In SZ, a deficit in primacy serial position recall has been documented and associated with decreased context-dependent recall and with speech errors29, raising the possibility of similar deficits in SPD.

Gender differences in SPD

As in our previous studies, male SPDs showed more severe cognitive impairment than female SPDs, particularly with regard to verbal learning15,16. Female SPDs showed no deficits on this task and performed similarly to controls. As in SZ studies, the reasons for better verbal memory performance in women with the disorder are unclear but may be explained as a function of the normal sexual dimorphism of the brain or the modulating effects of estrogen on dopamine systems26,30.

Other studies have suggested that left hippocampal and planum temporale abnormalities may contribute to language dysfunction in male SZ31, and that inferior parietal-prefrontal cortical connectivity may be related to the relative preservation of verbal memory in female as contrasted with male SZ32. Despite this evidence, studies of cognition in SZ and SPD lack consistent measurement of the effect of gender, and most examinations of context processing have employed mixed groups of males and females. Gender should be considered an important factor in future studies of SPD and SZ, as it appears to modulate cognitive performance.

Neuropsychology of SPD

The current results support hypotheses that, like SZ, the neuropsychology of SPD involves deficits in an interconnected neural system involving prefrontal and temporal lobe structures11, pathways that appear crucial for context processing in this population20,21. In this context, cognitive deficits could reflect dysfunction at any one or more points within the system and vary accordingly. The results again underscore the cognitive deficits and treatment needs of SPD individuals, and suggest that these may differ according to gender. SPD is a significant but understudied mental health problem, and important from a research prospective. The NIH-sponsored National Epidemiologic survey targeted a nationally representative sample of 34,653 and found the lifetime prevalence of SPD to be 3.9% based on DSM-IV diagnostic criteria12,33. SPD was reported to be highly disabling with specific impairments in social functioning and reduction in quality of life. Our results suggest that difficulties in social and occupational performance in this group12,34 may stem from both characterological features and cognitive deficits.

SPD and clinical high risk (CHR) populations

Similar semantic processing deficits have been found in individuals at clinical high risk for psychosis35,36,37, implicating similar neural systems. The field of clinical high risk (CHR) research is rapidly expanding to address risk and prediction of transition to psychosis, initiatives in clinical care for CHR youth, and preventive interventions that may benefit this vulnerable population. The relationship of clinically-defined SPD and CHR populations is not entirely clear, although 1) a diagnosis of SPD may be an element among factors determining transition to psychosis among CHR youth38, and 2) a sizeable number of individuals who meet clinical criteria for SPD would be expected to fulfill the Criteria of Psychosis-Risk Syndromes (COPS39) definition of Attenuated Psychotic Symptoms Syndrome (APSS), one of four possible psychosis-risk syndromes required to meet the definition of CHR40. When studied prospectively, individuals diagnosed with SPD have shown an elevated rate of SZ and other psychoses at follow-up41, further supporting the SPD diagnosis as a risk factor for transition to psychosis. The DSM 5 TR42 includes proposed criteria for attenuated psychosis syndrome (APS), including at least one of the following symptoms: attenuated delusions, hallucinations, or disorganized speech, that are occurring at least once per week, have begun or worsened in the past year, are sufficient to warrant clinical attention, are not better explained by another mental disorder, substance or medical condition, and the criteria for a psychotic disorder have never been met. The subjects that meet criteria for SPD in ours and other studies may meet criteria for APS and thus make up a significant proportion of those individuals included in CHR studies.

Although early intervention to prevent severe outcomes and improved understanding of disease progression are positive goals for CHR research, there are critical views of the current definitions used in this model43,44,45. Concerns have been raised that the bulk of CHR groups have common psychiatric symptoms of depression and anxiety and, although they may have some psychotic symptoms associated with these disorders, they will never transition to psychosis or SZ. Designation of these patients as high risk for psychosis may unfairly label them and does not appear to promote appropriate treatment for their symptoms. Additional concerns include potential stigma and discrimination46, inefficiency and inaccuracy in diagnosis, challenges in accurately identifying true risks for psychosis, and the efficacy and safety of early antipsychotic use.

Study limitations

One possible limitation of the current study is the unequal numbers of male and female subjects, and the resultant impact on effect size, i.e., no significant deficits may have been evident in female SPD simply because of the smaller sample size and resultant reduction in power. For the purpose of comparison, in Table 2, effect sizes for male and female SPD subjects are listed for each learning trial. Beginning at the second order of approximation, effect sizes for male SPD are consistently larger than for female SPD, strengthening the argument for a selective verbal learning deficit in male SPD. The current results are also consistent with previous studies from our laboratory showing more MRI47,48 and evoked potential49 abnormalities in males than females with SPD. Another limitation is the lack of detailed family history available for our study population. We have consistently found our SPD subjects to 1) have limited information about their family history of mental illness and 2) to be guarded in allowing us to contact a relative to obtain more detailed information. They seem more comfortable answering a general family medical history questionnaire administered by the examiner. From this type of general questioning, no SPD subjects in the current study were aware of a confirmed diagnosis of SZ in the family. However, two male SPD subjects endorsed probable psychosis and seven endorsed depression in a first-degree family member. Three female SPD subjects endorsed probable psychosis and two endorsed depression in a first-degree relative; one endorsed bipolar illness in a second degree relative.

Table 2 Comparison of effect sizes for male and female SPD on each word list of the semantic facilitation task.
Full size table

An additional limitation includes our use of a binary approach to measuring gender, and lack of a more precise method for measuring birth sex and self-reported gender identity50. Social and cultural views and the understanding of sex and gender are changing, and a binary approach to gender measurement may misrepresent the nature of gender diversity and lead to the misclassification of research participants51. Given the apparent importance of this variable in modulating cognitive performance in SZ and SPD, precise measurement and use of inclusive options for self-report are necessary to elucidate the true relationship of gender and neural processes in these patient populations52,53.

Future studies

Further study of the behavioral, functional, and structural correlates of brain function in SPD populations is warranted to examine the cognitive and clinical features of this disorder, as well as to help shape our understanding of the pathophysiology of schizotypy and schizophrenia. The current study demonstrated that, like SZ, male SPD subjects profit less from semantic context in verbal learning than controls. As in SZ, the extent to which this is related to symptom profile, thought disorder, and general verbal ability remains unclear. Unfortunately, the current subject sample did not have a full array of neuropsychological tests for a comparison with performance on the semantic facilitation task. Future studies that examine the relationship of semantic context processing to performance on other neuropsychological measures, including effort, attention, processing speed, speech, and self-regulation, as well as serial position and interference effects in learning and memory recall, will be important for understanding the impact of such deficits on cognitive, social, and occupational functioning. As above, precise measurement of sex and gender identity will be necessary to clarify the specificity of cognitive deficits in SPD. Additional factors for future study include the effects of symptom profile and family history of mental illness on cognition. In the current study, no differences were found between the male and female SPD groups in the number or type of DSM-5 criteria met, but additional information on schizotypal dimensions and symptom severity would be informative. Finally, careful examination of family history of mental illness that respects subjects’ privacy may yield important information about genetic influences on the presentation of SPD.

Methods

Subjects

Subjects were 49 men and 20 women who met DSM-5 criteria for SPD, compared with 23 male and 33 female comparison subjects, respectively, who were comparable on age, education, general intellectual ability, ethnicity, and parental socioeconomic status (SES, see Table 3).

Table 3 Subject characteristics for male and female SPD and HC subjects.
Full size table

Demographic measures

Measures of depression, intellectual ability, and socioeconomic status were administered to all subjects to control for any differences between the SPD and HC groups. To assess current symptoms of depression, subjects were administered the Beck Depression Inventory (BDI54), a 21-item self-report measure of mood in the past week. The Vocabulary and Block Design subtests from the Wechsler Adult Intelligence Scale (WAIS55 or 56) were administered as measures of verbal and nonverbal intellectual ability, respectively. The Vocabulary subtest is an expressive vocabulary definition measure that is highly correlated with overall IQ score and is considered a strong measure of general intelligence. Because verbal skills might be generally reduced in the SPD population, the Block Design subtest was administered as a nonverbal measure of general intellectual ability. This is a timed measure that assesses visuospatial abilities by having subjects replicate a design using two-colored blocks, and is considered a good measure of fluid intelligence, visuospatial construction, and speeded problem-solving. As a measure of personal and parental socioeconomic position, the Hollingshead Four-Factor Index of Social Status57 was calculated based on subjects’ report of their own and their parents’ education, occupation, sex, and marital status.

Two X Two [Group X Gender] ANOVAs were conducted to assess differences between SPD and gender-matched HC subjects on demographic variables. No significant differences were found between the groups on age, education, and performance on the WAIS Vocabulary and Block Design subtests. Although in general SPD subjects had significantly higher levels of depression on the BDI than HCs, these scores generally fell within a subclinical range. No differences in personal or parental SES were evident between the groups.

Recruitment and diagnostic procedures

were the same as in our previous studies13,14,15,16. Sixty-nine SPD subjects were recruited from the general population and answered advertisements in local newspapers asking for men and women who a) “believe they have ESP, clairvoyance, telepathy, or a sixth sense; sense the presence of others when alone; think others can feel your emotions”58, and b) “are shy or uncomfortable around unfamiliar people or in close relationships.” Three subjects were referred by their mental health clinicians, and one subject was recruited at a local “psychic fair.” All subjects were between 18 and 55 years of age and all were right-handed.

Sex and gender

The importance of differentiating the terms sex and gender in psychological research and in understanding the human experience has become increasingly apparent in recent years50,51. Sex typically refers to biological characteristics such as reproductive organs and chromosomes; gender is a social and cultural construct that encompasses roles, behaviors, and identities that societies attribute to individuals. In the current study, the term gender is considered most appropriate and is used consistently. Initially, only men were recruited through written advertisements; subjects interested in the study called in as men (males) at their discretion. As the study progressed and expanded over time, only women were recruited for a period of time, and then both men and women were recruited. As part of the recruitment intake process, subjects were asked to identify their gender, male or female, in addition to other demographic variables. Because it was at the subject’s discretion to let us know if they were male or female (calling in based on ads recruiting men or women), and/or for the intake research assistant to determine gender based on subjects’ identification or by other social cues (e.g.,name, vocal tone, etc.), the term gender is used throughout the manuscript. No detailed information was collected regarding sex at birth, sex organs, or chromosomes. Although the possibility exists that a subject’s sex at birth differed from their gender identity, subjects were not asked and no subject offered this information. This raises a potential study limitation, as more precise methods have been developed to measure these variables52,53.

SPD diagnostic criteria

SPD subjects met full clinical diagnostic criteria based on structured DSM-IV clinical interviews (SCID59), administered by a licensed clinical psychologist (MMV) or psychiatrist (CCD). The fourth edition of the Diagnostic and Statistical Manual for Mental Disorders (DSM33) was in use at the time subjects entered into the study. No changes were made to SPD clinical diagnostic criteria with the publication of DSM-5 in 201325 and DSM-5-TR in 202242: the nine core criteria have remained consistent to date (ideas of reference, odd beliefs, unusual perceptions, odd speech, suspiciousness, inappropriate affect, lack of close friends, excessive social anxiety), requiring a minimum of five of these for a diagnosis of SPD. The terms DSM-IV and DSM-5 may thus be used interchangeably when referring to SPD clinical diagnostic criteria. For this manuscript, the term DSM-5 is used in the text to inform the reader that the most recent clinical criteria have been used for diagnosing SPD.

SPD subjects had no lifetime history of Axis I bipolar or psychotic disorder, and no current depressive disorder. They were taking no psychiatric medications and had never been treated with antipsychotics. Control subjects answered newspaper advertisements recruiting healthy right-handed individuals for participation in cognitive research. Healthy control subjects had no lifetime history of Axis I or Axis II disorder based on structured DSM-IV clinical interview. After complete description of the study to the subjects, written informed consent was obtained, consistent with local institutional review board (IRB) guidelines. All subjects were paid for their participation in the study.

The IRB at Harvard Medical School, Department of Psychiatry, MMHC division of BIDMC, granted approval for the use of human subjects.

Procedure

Subjects were aurally presented a semantic facilitation task identical to the original Miller and Selfridge6 experiment. Instructions were read to the subject, followed by a practice recall exercise. Ten-word lists with increasing context (see Table 4) were presented individually via tape recorder at a comfortable audible level of 70 dB SPL. The order of the lists was not randomized. As in the original task, all eight 10-word lists were presented in order to each subject, proceeding from the least to the greatest contextual constraint. Subjects were asked to listen to each 10-word list. After hearing each list, they were cued by the examiner to write as many words as they could recall from the list. Content was scored for each word correctly recalled, regardless of the order of recall. Sentences with higher-order context were expected to result in more words recalled.

Table 4 Ten-word lists of increasing context used in semantic recall task.
Full size table

Data analysis

The number of correct responses was computed for each of eight wordlists. An 8 × 2 × 2 repeated measures ANOVA was employed, with Word List (0-Order to Text), Group (SPD/HC) and Gender (Male/Female) as the independent variables and Words Correct as the dependent variable. BDI scores were included as a covariate in a secondary analysis to correct for the effects of symptoms of depression. Simple effects were examined to test the interactive effects of Gender X Word List separately for the SPD and control groups. Post hoc Tukey tests were employed to examine specific differences in performance between groups.