Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2020 Jul 1.
Published in final edited form as: J Abnorm Psychol. 2019 May 6;128(5):423–430. doi: 10.1037/abn0000432

A Working Memory Related Mechanism of Auditory Hallucinations

Christopher Gaudiot 1,2, Xiaoming Du 1, Ann Summerfelt 1, Stephanie M Hare 1, Juan R Bustillo 2, Laura M Rowland 1, L Elliot Hong 1,*
PMCID: PMC6774251  NIHMSID: NIHMS1030023  PMID: 31058523

Abstract

Background

Cognitive mechanisms underlying auditory hallucinations (AH) in schizophrenia have been related to working memory (WM), although the formative mechanism is unknown. The phonological loop refers to subvocal rehearsals of information held online for supporting WM. As WM deficiency is frequent in schizophrenia, we hypothesized that AH and WM deficit share a common dysfunction in phonological loop operation, especially when it is taxed by ambiguous auditory and verbal associations.

Methods

We developed an active phonological priming (APP) paradigm in which participants generated arbitrary verbal associations to pseudo-words with ambiguous meaning. They were later asked to rate their familiarity to each pseudo-word, a task that required subvocal evaluation of ambiguous auditory-verbal information. Factor and mediation analyses were used to test the hypothesis that WM, AH, and APP induced phonological bias towards perceiving ambiguous contents as familiar may share a common underlying mechanism.

Results

In 32 patients with schizophrenia (SZ) and 20 healthy controls (HC), SZ rated ambiguous pseudo-words as significantly more familiar compared with HC (p=0.006), indicating a proneness to APP-induced bias. This increased subjective bias to perceive ambiguous contents as familiar after APP significantly correlated with AH severity (p=0.001), and mediated the relationship between WM and AH. Factor analysis demonstrated a common latent factor among WM, AH and the bias induced by active priming to ambiguous contents.

Conclusions

A heightened phonological loop priming to ambiguous contents appears to be mechanistically linked to WM deficits and AH in schizophrenia. These findings emphasize the importance of jointly addressing WM deficits and AH in clinical practice and research.

Keywords: working memory, hallucination, psychosis, phonological loop, psychosis

General scientific summary

The study applies an active priming paradigm that demands phonological loop operation (a working memory component) while requires subjectively generated associations with pseudo-words that have ambiguous meanings. People with schizophrenia are found to be significantly more prone to bias towards recognizing meaningless pseudo-words as familiar, and the bias mediates the working memory - auditory hallucination relationship. Auditory hallucinations and working memory may share a common underlying phonological loop mechanism, a finding that encourages developing new behavioral or pharmacological interventions that jointly target working memory deficit and auditory hallucinations in schizophrenia.

Introduction

Auditory hallucinations (AH) are experiences in which the boundaries of external perception, internal thought and speech productions are blurred. AH are commonly reported by patients with schizophrenia, but also occur in other psychiatric conditions (Andreasen & Flaum, 1991; Sartorius, Shapiro, & Jablensky, 1974). Cognitive deficits are also a core feature of schizophrenia, and predict functional disability in the illness (Barch & Ceaser, 2012; Goldman-Rakic, 1999). Of the cognitive deficits, working memory (WM) impairment is one of the most robust. AH and WM may be related as supported by their clinical correlations (George & Neufeld, 1985; F. Waters et al., 2012) and common neuroanatomical regions involved in both AH and WM (Hashimoto, Lee, Preus, McCarley, & Wible, 2010; Wible et al., 2009). Nevertheless, AH and WM are still largely studied as separate entities as the field lacks an understanding on their shared mechanism.

WM refers to holding recent information “online” when the brain is performing additional information manipulation. Classic theories consider the “phonological loop” a key component in WM, wherein subvocal rehearsal allows information to be held online and manipulated (Baddeley, 2003, 2012; Schulze & Koelsch, 2012). As AH are thought to be aberrant internal representations of thought/perception information (Allen, Laroi, McGuire, & Aleman, 2008; Ford et al., 2012; Frith & Done, 1988; F. A. Waters, Badcock, Michie, & Maybery, 2006), we theorize that AH are associated with the phonological loop, a component of WM. We hypothesize that the impairments in the phonological loop will be particularly apparent under conditions that actively challenge it such as using never-rehearsed phonological tasks, and in this context, such WM deficits should be associated with AH symptom severity.

To test this hypothesis and actively challenge the phonological loop, we designed a paradigm that imposed active priming to ambiguous pseudo-words that resembled real words phonemically but did not have semantic meaning. Participants were then primed by listening to these pseudo-words and asked to speak out loud a word that could be associated with each pseudo-word. This active phonological priming (APP) approach taxes the auditory-verbal system by requiring the generation of a quasi-meaningful association to an otherwise ambiguous auditory input. The paradigm aimed to elicit the extent of biases when the phonological loop was challenged by the requirement to rehearse new (e.g. never-rehearsed) and ambiguous (e.g. lacking semantic meaning) auditory-verbal information.

Method

Participants

The study included 31 patients with schizophrenia (SZ) diagnosed using the Structured Clinical Interview for DSM-IV (SCID), and 20 healthy controls (HC). They were between 18–65; native American English speakers; and right-handed according to the Edinburgh Handedness Inventory. Except for 7 medication-free patients, all were on antipsychotics: 17 taking atypical, 2 taking typical, and 5 taking more than one antipsychotic type. The average chlorpromazine equivalent (CPZ) was 441±409 (mean±sd). Subjects had no major medical and neurological illnesses, head injury, and no DSM-IV substance abuse in the last 6 months or substance dependence in lifetime (except nicotine). We made attempts to frequency-match the smoking status between patients and controls (see Table 1). Patients were recruited through outpatient mental health clinics. Patients were stable outpatients, defined as no hospitalization and no change of antipsychotic medications in the past 4 weeks. Other demographic information is in Table 1. Media advertisements were used to recruit HC, who were screened using SCID and without Axis I diagnoses. Participants with schizophrenia were evaluated for their ability to provide informed consent before signing consent documents. All participants gave written informed consent prior to participation in the study. This study was approved by the University of Maryland Baltimore Institutional Review Board.

Table 1.

Demographic and clinical characteristics of participants (mean±sd). SZ: Schizophrenia participants; HC: Healthy control participants. Verbal working memory: digit sequencing task age and sex adjusted T score from the Brief Assessment of Cognition in Schizophrenia. Auditory hallucinations: Psychotic symptom rating scales - auditory hallucination scale total score. BPRS: Brief psychiatric rating scale total score. CA: Caucasian Americans; AA: African Americans.

SZ (n=31) HC (n=20) F or Χ2 p
Age (years) 43.8±14.2 41.9±14.2 0.2 0.64
Sex (female:male) 9:22 6:14 0.01 0.94
Current smoker (%) 31.7% 20.6% 1.3 0.34
Race (%CA:AA:other) 43:52:8 64:27:9 2.4 0.12
Education (year) 12.9±2.2 14.5±2.3 9.6 0.003
Duration of illness 11.7±10.8 n/a n/a n/a
Verbal working memory 37.6±15.0 48.3±9.9 7.8 0.007
Auditory hallucinations 12.4±10.9 n/a n/a n/a
BPRS 43.2+11.6 n/a n/a n/a
Open in a new tab

Clinical and cognitive assessments

AH were assessed with the Psychotic Symptom Rating Scales - Auditory Hallucination Scale (PSYRATS-AHS) (Haddock, McCarron, Tarrier, & Faragher, 1999); its total score has shown good validity in measuring AH (Drake, Haddock, Tarrier, Bentall, & Lewis, 2007; Haddock, et al., 1999; Woodward et al., 2014). The Brief Psychiatric Rating Scale (BPRS) (Overall & Gorham, 1962) provided additional assessment on general psychopathology; its subscale scores in thinking disturbance, depression/anxiety, withdrawal, hostility, psychosis, and activation (Overall & Gorham, 1962; Hedlund & Vieweg, 1980; Velligan et al., 2005) allowed further testing for clinical symptom specificity. Symptoms ratings were carried out by clinicians trained to administer these scales and maintained an interrater reliability at 0.80 or more with the gold standard. The Brief Assessment of Cognition in Schizophrenia (BACS) digit-sequencing task was used to measure WM, where digits of increasing length were orally presented and participants are asked to recall the digits from lowest to highest (Keefe et al., 2008). To test for cognitive specificity, we also assessed processing speed as measured by the digit symbol coding task of the WAIS-3 (Weshsler, 1997).

Active phonological Priming (APP)

In this paradigm, verbal association of the auditory stimuli was primed by requiring a verbally generated association (e.g., when hearing the word “sleep”, one may respond with “bed”). In active priming to real words as in the above example, the “free association” can be almost automatic. In comparison, pseudo-words were not inherently meaningful, and one must create a new phonological association in order to generate a verbal response. APP to ambiguous pseudo-words was then compared with three parametrically designed control conditions: 1) APP to real words; 2) passive priming to ambiguous pseudo-words; and 3) passive priming to real words. The design enabled pseudo-randomized, balanced control for the passive vs. active and the real word vs. the pseudo-word parameters, and aimed to control for potential group differences in understanding of the task, recognition of the stimuli, and effort and/or attention. For instance, the APP to real word condition aimed to control for verbal responses and attention to stimuli by requiring subjects to respond verbally to real word stimuli as in APP to pseudo-words; and the real words were constructed using closely matched vowels, consonants and word structures, to minimize potential confounds induced by differences in attention, behavioral response, or linguistic characteristics.

The APP paradigm and the three control conditions were generated using 200 real words and 200 phonemically matched pseudo-words. The real words were chosen to be single syllable. All possible American English vowels and consonants that occur in nouns (Coltheart, 1981) were represented based on high ratings on recognition and concreteness to ensure that they are familiar to SZ and HC alike. Examples of these words are: cake, toast, foot, tongue, bar, and fog.

To create the 200 pseudo-words, the 200 real words were changed in one or two elements, which could be the vowel, the beginning consonant(s), the ending consonant(s) or a vowel and one consonant. For example, we could change cake (pronounced as “keɪk”) by its ending consonant from “k” to “θ”, resulting in the pseudo-word “kaith” (keɪθ). The participants never saw the written pseudo-words to avoid visual priming. Other examples were “foot (fʊt) - foat (foʊt)” or “fog (fog) - fosh (fof)”. The goal was to create pseudo-words phonemically resembling a real word but engendering ambiguity.

These 400 stimuli were individually pronounced by a male native speaker of American English in normal intonation and pace, and recorded at 48 kHz (Tascam DR-05, Tascam, CA) in an acoustic chamber. The 400 recorded stimuli were normalized by equalizing peak amplitudes using Audacity 2.1.2 software. The average stimulus lengths for real-words (0.54±0.13s) were similar to pseudo-words (0.56±0.13s; p>0.05). The 400 stimuli (www.mdbrain.org/real_pseudoword_stimuli.pdf) and audible examples (www.mdbrain.org//real_pseudoword_auditory_example.zip) are available online.

The stimuli were divided into two sets, each with 100/100 of real/pseudo-words, one for APP (Fig. 1, top two rows) and one for passive priming by passive listening (bottom two rows). The assignment was random with the exception that each set included all possible vowels and consonants. In each set, real and pseudo-words were mixed, arranged to avoid three or more real (or pseudo) words in a row. All participants had the same sequence. Stimuli were presented using headphones (Razer Blackshark, Razor, CA) at 70 db SPL. Inter-stimulus interval was between 4.0 to 5.5 seconds. During APP, participants were told that they would hear a series of words; that some might be difficult to understand; and no matter what they heard they were to say the first word or thing that came to mind. Examples of the responses were: cake - “walk” or toast - “bread” to real words and kaith - “case” or mub - “dirty” to pseudo-words. To reinforce responses, a microphone was present and participants were told that their responses were recorded. The contents of the responses were not used besides counting the responses made. The APP lasted for 16 minutes in one run.

Fig. 1.

Fig. 1.

Open in a new tab

An illustration of the study paradigm with actual examples from participants. Top two rows: participants listened and immediately responded with an audible verbal association to a pseudo-randomized set of 100 words and 100 pseudo-words; these stimuli were then repeated and participants were asked to rate their familiarity to the words and the pseudo-words (1=not familiar; 2=somewhat familiar; 3=very familiar). Bottom two rows: participants engaged in a passive control task by listening to another set of pseudo-randomized 100 words and 100 pseudo-words without making any verbal responses; these stimuli were then repeated and participants were again asked to rate their familiarity to the words and the pseudo-words.

After about 2 minutes, Set 1 was repeated (Fig. 1, second row) and participants were instructed to decide how well they recognized the meaning of each stimulus and rated each stimulus as 1=not familiar, 2=somewhat familiar, and 3=very familiar using a response pad. Specifically, participants were given the following verbal instructions: use “1” to indicate “not familiar”, meaning “I am not familiar with this word and I don’t know what it means”; “2” to indicate “somewhat familiar”, meaning “I am somewhat familiar with this word and I may not be totally sure what this word means”; and “3” to indicate “very familiar”, meaning “I am very familiar with this word and I know what it means.” The full instructions were given prior to task initiation and were reinforced during the practice session. Throughout the task subjects were given the following visual cue on the screen as a reminder: “1 = not familiar”,” 2 = somewhat familiar”, and “3 = very familiar”. The recognition rating is the average familiarity score, calculated as the total score divided by the number of responses. The assumption is that, for most real words, the rating should 3 while for pseudo-words, the rating should be 1. Higher ratings for pseudo-words imply a bias towards being more prone to misidentify ambiguous auditory verbal information as familiar. The term ‘bias’ is used to describe that ambiguously-sounding pseudo-words were perceived as familiar after the initial exposure.

In the passive listening task (Fig. 1, third row), Set 2 was presented. Participants were given the same instruction except that they just passively listened. After about 2 minutes, Set 2 was repeated (Fig. 1, fourth row). Participants were instructed to rate how well they recognized each stimulus.

Statistical Analysis

Demographic variables were compared using chi-square or t-tests. Repeated-measures ANOVA with Greenhouse-Geisser corrected statistics were employed, where average familiarity score was the dependent variable and listening mode (APP vs. passive listening) and word type (real vs. pseudo) were repeated measures. Exploratory factor analyses were performed to explore latent constructs among APP to pseudo-words and the other three control conditions, AH and WM in the patients using oblique rotations. Factor solutions were evaluated by eigenvalues >1.0, loadings >0.40, percent variance explained, and sampling adequacy. Mediation analysis was conducted using PROCESS in SPSS where AH score was the outcome variable, WM was the independent variable, and familiarity score of pseudo-word after APP was the mediator (Preacher & Hayes, 2008). The indirect effect and 95% confidence interval (CI) were estimated using a bootstrap with 5,000 resamples. Pearson’s correlations were used to examine the relationship between APP measures and AH and WM. All tests were two-tailed at p<0.05.

Results

1. Clinical characteristics

Compared to NC, SZ were not significantly different in age (p=0.64) or sex (p=0.94), but had significantly less years of education (p=0.003) (Table 1). SZ in this study had an average of 11.7 years of duration of illness, and had significantly reduced verbal working memory performance (p=0.007) as compared to NC. Their AH severity as rated by the PSYRATS-AHS had a mean score of 12.4 (range 0 to 33), which is consistent with reports of AH severity using this scale in other outpatient SZ samples (e.g., Steel et al., 2007). Additional clinical and demographic descriptions are in Table 1.

2. APP Effects

Both groups made verbal responses during APP (real or pseudo-word) to ≥97.5% of the stimuli (word type response main effect p=0.61; word type x group interaction p=0.53). Similarly, both groups responded to recognition ratings in ≥98.5% of the trials (recognition rating response main effect p=0.27; recognition rating response × group interaction p=0.76). There were also no significant word type response x recognition rating interaction (p=0.17) or word type response x recognition rating x group interaction (p=0.52). Therefore, performance was near 100%, suggesting both groups were able to similarly complete the tasks.

The full-model repeated measure ANOVA showed a significant listening mode × word type × diagnosis interaction on the familiarity rating (F=3.4, p=0.048) and a significant main effect of diagnosis (F=4.9, p=0.035). Patients with schizophrenia had a significant listening mode × word type interaction (p=0.005) where they reported essentially the same recognition after APP vs. passive listening to real words (p=0.96; Fig. 2B), but endorsed significantly more familiarity to the pseudo-words after APP as compared with after passive listening (p=0.006; effect size in Cohen’s d=0.64; Fig. 2D). Therefore, active vs. passive priming itself did not significantly altered patients’ judgment on recognizing real words. The bias towards an increased sense of familiarity became significant only after actively priming with pseudo-words.

Fig. 2.

Fig. 2.

Open in a new tab

Average familiarity scores from recognition rating on both real words (A and B) and pseudo-words (C and D) after passive listening (control) and active phonological priming tasks. d is Cohen’s effect size for paired sample comparison.

In comparison, controls showed no significant listening mode × word type interaction (p=0.35) but a significant main effect of listening mode (p=0.028). They rated real words as significantly more familiar after APP compared to passive listening (p=0.01; d=0.63; Fig. 2A). Controls did not rate pseudo-words as significantly more familiar after APP compared to passive listening (p=0.10) although the effect size was at d=0.39 (Fig. 2C).

3. APP and AH

The familiarity rating induced by APP to pseudo-words was positively correlated with AH score (r=0.58, p=0.001; Fig. 3D). In comparison, familiarity scores after passive listening to either real or pseudo-words showed no significant correlations to AH (r=0.03, p=0.86; r=0.15, p=0.43; Fig. 3A and 3B). There was also no significant relationship between AH and APP to real words (r=0.05, p=0.79, Fig. 3C). Correlations between familiarity rating induced by APP to pseudo-words and other symptoms of schizophrenia such as delusions as measured by the PSYRATS – Delusion Scale (r=0.08, p=0.66) and by the BPRS total score as a measure of general psychopathology (r=−0.08, p=0.72) were not significant. To further test for clinical specificity, we also explored the subscale scores in BPRS for thinking disturbance, depression/anxiety, withdrawal, hostility, psychosis, and activation subscale scores, and found that none of the symptoms were significantly correlated with APP to pseudo-words (r=−0.15 to 0.025, p=0.46 to 0.98).

Fig. 3.

Fig. 3.

Open in a new tab

Relationships between auditory hallucinations and average familiarity scores after passive listening (A and B) and active phonological priming (C and D) tasks in SZ. E: Verbal working memory (WM) correlations with average familiarity scores (recognition rating). Poorer WM was significantly associated with reporting stronger recognition of pseudo-words as familiar after active phonological priming (APP) in HC (r=−0.51, p=0.022) and SZ (r=−0.46, p=0.009), but not in any of the three control conditions. Y-axis plots the correlation coefficient r values between WM and average familiarity score during recognition rating (y-axis is flipped from negative to positive). F: Scatter plots between working memory and average familiarity scores after active phonological priming to pseudo-words in patients and controls. G: Mediation analysis. All values are t values. The total effect of WM on AH was significant (path C, t=−2.38). However, the direct effect of WM on AH was not significant when controlling for the mediator (path C’, t=−1.02), suggesting the effect of WM on AH was largely mediated by the familiarity level to pseudo-words after active phonological priming. * Statistically significant.

4. APP and Verbal WM

WM was impaired in patients (Table 1). Patients reporting stronger APP effect to pseudo-words were significantly more impaired in WM (r=−0.46, p=0.009); this was not found in any of the three control conditions (Fig. 3E). The pattern was similar in controls: only APP effect to pseudo-words was significantly correlated with WM (r=−0.51, p=0.022) (Fig. 3E). Therefore, APP effect to pseudo-words was related to WM, replicable across groups. In comparison, processing speed was not significantly correlated with APP to pseudo-words (r=−0.16, p=0.43).

5. Factor Analysis

The analysis identified three factors explaining 82.6% of the variance (Table 2). Factor 1 was loaded with rating after APP to pseudo-words, AH, and WM, and is named the “phonological loop” factor. Factor 2 was loaded with APP and passive listening to real words, and is named the “real meaning” factor. Factor 3 was a single measure on passive listening to pseudo-words, and thus named the “no meaning” factor. The Kaiser-Meyer-Olkin test for sampling adequacy was 0.54 (>0.50 is considered adequate for factor analysis)(Hair Jr, Anderson, Tatham, & William, 1995; Tabachnick & Fidell, 2007) and Bartlett’s test was χ2=111.9, p<0.001 (p<0.01 implies adequate) (Hair Jr, et al., 1995; Tabachnick & Fidell, 2007). Therefore, the overall solution was interpretable, and Factor 1 was consistent with the hypothesis that APP to ambiguous meanings may tap into a common pathway in AH and WM.

Table 2.

Factor analysis results. Three factors were identified, each with eigenvalues >1.0. Factor 1 was consistent with the main hypothesis as rating after APP to pseudo-words, auditory hallucinations and verbal working memory were clustered together and this factor explained 35.5% of the variance.

Latent Factor Factor 1 Factor 2 Factor 3
“Phonological Loop” “Real Meaning” “No Meaning”
Passive listening to real word −0.06 0.94 0.17
APP to real word −0.02 0.96 −0.05
Passive listening to pseudo-word −0.07 0.07 0.97
APP to pseudo-word 0.85 0.03 −0.18
Auditory hallucination 0.84 0.08 0.30
Verbal working memory −0.73 0.22 0.18
Percentage variance 35.5% 29.6% 17.5%
Eigenvalue 2.13 1.78 1.05
Open in a new tab

6. Mediation Analysis

The model was based on that WM was related to the familiarity to pseudo-words after APP (A: t=−2.79, p=0.009); the latter was related to AH (B: t=2.97, p=0.006); and WM was associated with AH (C: t=−2.38, p=0.02) (Fig. 3F). When the familiarity of pseudo-words was added as a mediator, the direct effect from WM to AH was no longer significant (t=−1.02, p=0.32; Path C’) and the indirect path was significant (b=−0.17, 95% CI=−0.45 to −0.04). Therefore, the WM-AH relationship in schizophrenia was largely mediated by a bias towards reporting more familiarity with ambiguous pseudo-words after APP (Fig. 3F).

7. Other Clinical Correlates

APP to pseudo-words was not associated with age (ps>0.90) or sex (ps>0.26) in SZ or HC. Antipsychotics dose (CPZ) was not significantly correlated with APP to pseudo-words (r=−0.15, p=0.41), WM (r=−0.16, p=0.42), or AH scores (r=−0.08, p=0.70).

Discussion

Identifying ambiguous auditory stimuli as familiar (or not) after one makes arbitrary auditory-verbal associations likely involves a concise phonological loop operation, as supported by its significant associations with WM in controls and patients alike. Individuals with lower verbal WM performance tended to be more susceptible to misidentifying ambiguous pseudo-words as familiar after APP. We found that SZ were significantly more vulnerable in this type of misidentification. This vulnerability was associated with more severe AH. The WM-AH relationship appeared largely mediated by how strong the tendency was to acquire such misidentification.

AH are present in about 60 to 80 % patients with schizophrenia at some point in the course of their illness (Aleman & Larøi, 2008; Lecrubier, Perry, Milligan, Leeuwenkamp, & Morlock, 2007). They can have devastating consequences, and have garnered extensive efforts to understand the mechanism (Bentall & Slade, 1985; Bullen & Hemsley, 1987; Dodgson & Gordon, 2009; F. Waters, et al., 2012). Commonly cited theories include the source-monitoring theory that suggests abnormal volitional assignment of self-generated speech leads to external misattribution (Frith & Done, 1988); the corollary discharge theory that suggests that intended thoughts normally prepare the auditory cortex for perceiving internally-generated stimuli, but abnormal corollary discharge leads to inappropriate attribution of auditory perceptions to external sources (Ford, et al., 2012; Heinks-Maldonado et al., 2007); the faulty episodic memory theory that suggests that AH stem from intrusions from stored memories (Jones, 2010; F. A. Waters, et al., 2006); and the hyperexcitation theory where AH is linked to an over-perceived state (Allen, et al., 2008). A metrical stress paradigm in fMRI found that phonological loop performance related functional circuitry was associated with AH (Curcic-Blake, et al., 2013); and others have shown that neural correlates of tasks for monitoring and processing of inner speech, silent thoughts, or perceived speech were impaired in patients who hallucinate (McGuire et al., 1995; Shergill, et al, 2000); although none of these task designs provided task performance readouts that support their behavioral correlations with AH. A common theme among these proposals implies erroneous source monitoring of auditory input, internal thoughts or stored memory. As proneness to misidentifying meaningless “words” as familiar can be construed as erroneous source monitoring, our finding bears resemblance to these theories for AH formation.

The current study showed that this error maybe WM-related and can be quantified in the laboratory. Only APP to ambiguous contents elicited a recognition bias that is significantly related to WM even in HC. Therefore, this effect cannot be just due to antipsychotic medications or psychosis in SZ, but supports that APP to pseudo-words taxed the WM system. The priming to new and ambiguous contents and then being asked to recall them makes it difficult to use automated semantic associations. Rather, recall or rehearsal of the contents of the subjectively generated new associations is needed during the effort to “recognize”. This phonological rehearsal and differentiation operation is a basic element of the phonological loop contribution to WM (S. E. Gathercole & Baddeley, 1989). This demand on WM explains why only APP to ambiguous content but not the other control conditions are related to WM.

The parametrically controlled paradigm provides strong inferences by excluding priming or semantic associations themselves as the main culprits (Fig. 3). Semantic and language-related priming are sometimes found to be abnormal in SZ (Kiang, Kutas, Light, & Braff, 2008; Lerner, Bentin, & Shriki, 2012). Semantic priming tasks are typically used to show a quicker response to a target word after priming (Heyman, Van Rensbergen, Storms, Hutchison, & De Deyne, 2015), thought to engage language/learning mechanisms (Susan E Gathercole & Adams, 1994; S. E. Gathercole & Baddeley, 1989). Active priming was indeed associated with higher ratings on familiarity to the real words in HC (p=0.01) but not in patients (Fig. 2B), which may be consistent with the learning mechanism in priming and its deficit in SZ, although this was not related to AH (Fig. 3).

Priming experiments often involve visual displays and passive priming. Active priming restricted to auditory and verbal association with pseudo-words is novel. The rationale was not to study priming effects, but to create new associations to otherwise ambiguous inputs. Importantly, controls did not report significantly more familiarity to pseudo-words by APP vs. passive listening (Fig. 2C), suggesting that APP to ambiguity did not significantly alter their judgment. We also found that patients and controls had identical levels of familiarity ratings to pseudo-words after passive listening (compare Fig. 2C vs. D), supporting that there was no prior exposure bias. This design was also informative because active or passive priming to real words did not result in a recognition bias that was associated with WM (Fig. 3E).

The findings suggest that the inability to differentiate from the ambiguous stimulus-response association in the context of impaired WM may underlie AH. The error generated by this manipulation was significantly associated with AH severity, and importantly, mediated the WM-AH relationship. The idea that AH may be related to WM is not new. In past studies, WM has shown the strongest association with AH compared to other cognitive domains (Gisselgard et al., 2014; Jenkins, Bodapati, Sharma, & Rosen, 2018), supporting the specificity. Previous studies showed that hallucinators perform worse than non-hallucinators in tracking masked speech, interpreted as disrupted speech perception in addition to impaired verbal WM in hallucinators (Hoffman, Rapaport, Mazure, & Quinlan, 1999). However, the underlying mechanism linking AH with WM was previously missing. By stressing the phonological loop component of WM (A. D. Baddeley, 1986), the current paradigm may have elicited a deficit in phonological loop that might mechanistically explain the WM to AH association.

An important limitation of the current study is that we only examined one aspect of WM, the phonological loop. Therefore, we do not know whether the phonological loop is the only component of WM that is associated with AH. Other limitations in this study included potential antipsychotics confounds, although this would not explain similar response rates between patients and controls in all conditions, and that both groups showed similar in response after passive listening to pseudo-words. Nevertheless, a specific medication effect only impacting APP to pseudo-words cannot be fully ruled out. This study is limited without anatomical or circuitry correlates; further research with neuroimaging techniques on APP to ambiguous meaning and its link to WM and AH is warranted.

Cognitive deficits and AH are critically relevant aspects in schizophrenia but are often treated as segregated problems in clinical practice and in research. The results here support the theory that impaired WM-related phonological loop operation may underlie AH vulnerability. Our findings encourage future studies to identify the neurobiological mechanism for the AH - cognition connection. Identifying mechanisms that link cognitive impairments to AH may have direct treatment implications, because AH treatment maybe more effective if new behavioral, pharmacological, or other therapies are developed to jointly target WM deficits and AH.

Disclosures and acknowledgements

LEH has received or plans to receive research funding or consulting fee on research projects from Mitsubishi, Your Energy Systems LLC, Neuralstem, Taisho, Heptares, Pfizer, Sound Pharma, Takeda, and Regeneron. All other authors declare no conflict of interest. Support was received from NIH grants MH112180, MH103222, MH108148, MH067533, a State of Maryland contract (M00B6400091), and a generous private philanthropic donation from the Clare E. Forbes Trust.

Disclosure: LEH has received or plans to receive research funding or consulting fee on research projects from Mitsubishi, Your Energy Systems LLC, Neuralstem, Taisho, Heptares, Pfizer, Sound Pharma, Takeda, and Regeneron. All other authors declare no conflict of interest. Support was received from NIH grants MH112180, MH116948, MH103222, MH108148, a State of Maryland contract (M00B6400091), and a generous private philanthropic donation from the Clare E. Forbes Trust.

The data reported in this manuscript originated from the research protocol entitled, “Electrophysiology and Imaging Study of Auditory Hallucination” (HP-00060918), which was approved by the University of Maryland, Baltimore (UMB) Institutional Review Board (IRB).

References

  1. Aleman A, & Larøi F (2008). Hallucinations: The science of idiosyncratic perception: American Psychological Association. [Google Scholar]
  2. Allen P, Laroi F, McGuire PK, & Aleman A (2008). The hallucinating brain: a review of structural and functional neuroimaging studies of hallucinations. Neuroscience and Biobehavioral Reviews, 32, 175–191. doi: 10.1016/j.neubiorev.2007.07.012 [DOI] [PubMed] [Google Scholar]
  3. Andreasen NC, & Flaum M (1991). Schizophrenia: the characteristic symptoms. Schizophrenia Bulletin, 17, 27–49. [DOI] [PubMed] [Google Scholar]
  4. Baddeley A (2003). Working memory and language: an overview. Journal of Communication Disorders, 36, 189–208. [DOI] [PubMed] [Google Scholar]
  5. Baddeley A (2012). Working memory: theories, models, and controversies. Annual Review of Psychology, 63, 1–29. doi: 10.1146/annurev-psych-120710-100422 [DOI] [PubMed] [Google Scholar]
  6. Baddeley AD (1986). Working memory Oxford. England: Oxford Uni. [Google Scholar]
  7. Barch DM, & Ceaser A (2012). Cognition in schizophrenia: core psychological and neural mechanisms. Trends in Cognitive Sciences, 16, 27–34. doi: 10.1016/j.tics.2011.11.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bentall RP, & Slade PD (1985). Reality testing and auditory hallucinations: a signal detection analysis. British Journal of Clinical Psychology, 24 ( Pt 3), 159–169. [DOI] [PubMed] [Google Scholar]
  9. Bullen JG, & Hemsley DR (1987). Schizophrenia: a failure to control the contents of consciousness? British Journal of Clinical Psychology, 26 ( Pt 1), 25–33. [DOI] [PubMed] [Google Scholar]
  10. Coltheart M (1981). The Mrc Psycholinguistic Database. Quarterly Journal of Experimental Psychology Section a-Human Experimental Psychology, 33, 497–505. doi: Doi 10.1080/14640748108400805 [DOI] [Google Scholar]
  11. Curcic-Blake B, Liemburg E, Vercammen A, Swart M, Knegtering H, Bruggeman R, & Aleman A (2013). When Broca goes uninformed: reduced information flow to Broca’s area in schizophrenia patients with auditory hallucinations. Schizophrenia Bulletin, 39, 1087–1095. doi: 10.1093/schbul/sbs107 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Dodgson G, & Gordon S (2009). Avoiding false negatives: are some auditory hallucinations an evolved design flaw? Behavioural and Cognitive Psychotherapy, 37, 325–334. [DOI] [PubMed] [Google Scholar]
  13. Drake R, Haddock G, Tarrier N, Bentall R, & Lewis S (2007). The Psychotic Symptom Rating Scales (PSYRATS): Their usefulness and properties in first episode psychosis. Schizophrenia Research, 89, 119–122. doi: 10.1016/j.schres.2006.04.024 [DOI] [PubMed] [Google Scholar]
  14. Ford JM, Dierks T, Fisher DJ, Herrmann CS, Hubl D, Kindler J, … van Lutterveld R (2012). Neurophysiological studies of auditory verbal hallucinations. Schizophrenia Bulletin, 38, 715–723. doi: 10.1093/schbul/sbs009 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Frith CD, & Done DJ (1988). Towards a neuropsychology of schizophrenia. British Journal of Psychiatry, 153, 437–443. [DOI] [PubMed] [Google Scholar]
  16. Gathercole SE, & Adams A-M (1994). Children′ s phonological working memory: Contributions of long-term knowledge and rehearsal. Journal of memory and language, 33, 672–688. [Google Scholar]
  17. Gathercole SE, & Baddeley AD (1989). Evaluation of the Role of Phonological Stm in the Development of Vocabulary in Children - a Longitudinal-Study. Journal of Memory and Language, 28, 200–213. doi: Doi 10.1016/0749-596x(89)90044-2 [DOI] [Google Scholar]
  18. George L, & Neufeld RW (1985). Cognition and symptomatology in schizophrenia. Schizophrenia Bulletin, 11, 264–285. [DOI] [PubMed] [Google Scholar]
  19. Gisselgard J, Anda LG, Bronnick K, Langeveld J, Hegelstad WT, Joa I, … Larsen TK (2014). Verbal working memory deficits predict levels of auditory hallucination in first-episode psychosis. Schizophrenia Research, 153, 38–41. doi: 10.1016/j.schres.2013.12.018 [DOI] [PubMed] [Google Scholar]
  20. Goldman-Rakic PS (1999). The physiological approach: functional architecture of working memory and disordered cognition in schizophrenia. Biological Psychiatry, 46, 650–661. [DOI] [PubMed] [Google Scholar]
  21. Haddock G, McCarron J, Tarrier N, & Faragher EB (1999). Scales to measure dimensions of hallucinations and delusions: the psychotic symptom rating scales (PSYRATS). Psychological Medicine, 29, 879–889. [DOI] [PubMed] [Google Scholar]
  22. Hair JF Jr, Anderson RE, Tatham RL, & William C (1995). Multivariate data analysis with readings. New Jersy: Prentice Hall. [Google Scholar]
  23. Hashimoto R, Lee K, Preus A, McCarley RW, & Wible CG (2010). An fMRI study of functional abnormalities in the verbal working memory system and the relationship to clinical symptoms in chronic schizophrenia. Cerebral Cortex, 20, 46–60. doi: 10.1093/cercor/bhp079 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Hedlund, & Vieweg. (1980). The brief psychiatric rating scale (BPRS): a comprehensive review. Journal of operational psychiatry, 11, 48–65. [Google Scholar]
  25. Heinks-Maldonado TH, Mathalon DH, Houde JF, Gray M, Faustman WO, & Ford JM (2007). Relationship of imprecise corollary discharge in schizophrenia to auditory hallucinations. Archives of General Psychiatry, 64, 286–296. doi: 10.1001/archpsyc.64.3.286 [DOI] [PubMed] [Google Scholar]
  26. Heyman T, Van Rensbergen B, Storms G, Hutchison KA, & De Deyne S (2015). The Influence of Working Memory Load on Semantic Priming. Journal of Experimental Psychology-Learning Memory and Cognition, 41, 911–920. doi: 10.1037/xlm0000050 [DOI] [PubMed] [Google Scholar]
  27. Hoffman RE, Rapaport J, Mazure CM, & Quinlan DM (1999). Selective speech perception alterations in schizophrenic patients reporting hallucinated “voices”. American Journal of Psychiatry, 156, 393–399. doi: 10.1176/ajp.156.3.393 [DOI] [PubMed] [Google Scholar]
  28. Jenkins LM, Bodapati AS, Sharma RP, & Rosen C (2018). Working memory predicts presence of auditory verbal hallucinations in schizophrenia and bipolar disorder with psychosis. Journal of Clinical and Experimental Neuropsychology, 40, 84–94. doi: 10.1080/13803395.2017.1321106 [DOI] [PubMed] [Google Scholar]
  29. Jones SR (2010). Do We Need Multiple Models of Auditory Verbal Hallucinations? Examining the Phenomenological Fit of Cognitive and Neurological Models. Schizophrenia Bulletin, 36, 566–575. doi: 10.1093/schbul/sbn129 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Keefe RS, Harvey PD, Goldberg TE, Gold JM, Walker TM, Kennel C, & Hawkins K (2008). Norms and standardization of the Brief Assessment of Cognition in Schizophrenia (BACS). Schizophrenia Research, 102, 108–115. doi: 10.1016/j.schres.2008.03.024 [DOI] [PubMed] [Google Scholar]
  31. Kiang M, Kutas M, Light GA, & Braff DL (2008). An event-related brain potential study of direct and indirect semantic priming in schizophrenia. American Journal of Psychiatry, 165, 74–81. doi: 10.1176/appi.ajp.2007.07050763 [DOI] [PubMed] [Google Scholar]
  32. Lecrubier Y, Perry R, Milligan G, Leeuwenkamp O, & Morlock R (2007). Physician observations and perceptions of positive and negative symptoms of schizophrenia: a multinational, cross-sectional survey. European Psychiatry, 22, 371–379. doi: 10.1016/j.eurpsy.2007.03.003 [DOI] [PubMed] [Google Scholar]
  33. Lerner I, Bentin S, & Shriki O (2012). Excessive attractor instability accounts for semantic priming in schizophrenia. PloS One, 7, e40663. doi: 10.1371/journal.pone.0040663 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. McGuire PK, Silbersweig DA, Wright I, Murray RM, David AS, Frackowiak RS, & Frith CD (1995). Abnormal monitoring of inner speech: a physiological basis for auditory hallucinations. Lancet, 346, 596–600. [DOI] [PubMed] [Google Scholar]
  35. Overall JE, & Gorham DR (1962). The brief psychiatric rating scale. Psychological Reports, 10, 799–812. [Google Scholar]
  36. Preacher KJ, & Hayes AF (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40, 879–891. [DOI] [PubMed] [Google Scholar]
  37. Sartorius N, Shapiro R, & Jablensky A (1974). The international pilot study of schizophrenia. Schizophrenia Bulletin, 1, 21. [DOI] [PubMed] [Google Scholar]
  38. Schulze K, & Koelsch S (2012). Working memory for speech and music. Annals of the New York Academy of Sciences, 1252, 229–236. doi: 10.1111/j.1749-6632.2012.06447.x [DOI] [PubMed] [Google Scholar]
  39. Shergill SS, Bullmore E, Simmons A, Murray R, & McGuire P (2000). Functional anatomy of auditory verbal imagery in schizophrenic patients with auditory hallucinations. American Journal of Psychiatry, 157, 1691–1693. doi: 10.1176/appi.ajp.157.10.1691 [DOI] [PubMed] [Google Scholar]
  40. Steel C, Garety PA, Freeman D, Craig E, Kuipers E, Bebbington P, … Dunn G (2007). The multidimensional measurement of the positive symptoms of psychosis. International Journal of Methods in Psychiatric Research, 16, 88–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Tabachnick BG, & Fidell LS (2007). Using multivariate statistics: Allyn & Bacon/Pearson Education. [Google Scholar]
  42. Velligan D, Prihoda T, Dennehy E, Biggs M, Shores-Wilson K, Crismon ML, … Trivedi M (2005). Brief psychiatric rating scale expanded version: how do new items affect factor structure? Psychiatry Research, 135, 217–228. [DOI] [PubMed] [Google Scholar]
  43. Waters F, Allen P, Aleman A, Fernyhough C, Woodward TS, Badcock JC, … Laroi F (2012). Auditory hallucinations in schizophrenia and nonschizophrenia populations: a review and integrated model of cognitive mechanisms. Schizophrenia Bulletin, 38, 683–693. doi: 10.1093/schbul/sbs045 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Waters FA, Badcock JC, Michie PT, & Maybery MT (2006). Auditory hallucinations in schizophrenia: intrusive thoughts and forgotten memories. Cognitive Neuropsychiatry, 11, 65–83. doi: 10.1080/13546800444000191 [DOI] [PubMed] [Google Scholar]
  45. Weshsler D (1997). Wechsler Adult Intelligence Scale (3rd ed.) San Antonio, TX: Psychological Corporation. [Google Scholar]
  46. Wible CG, Lee K, Molina I, Hashimoto R, Preus AP, Roach BJ, … Fbirn. (2009). fMRI activity correlated with auditory hallucinations during performance of a working memory task: data from the FBIRN consortium study. Schizophrenia Bulletin, 35, 47–57. doi: 10.1093/schbul/sbn142 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Woodward TS, Jung K, Hwang H, Yin J, Taylor L, Menon M, … Erickson D (2014). Symptom dimensions of the psychotic symptom rating scales in psychosis: a multisite study. Schizophrenia Bulletin, 40 Suppl 4, S265–274. doi: 10.1093/schbul/sbu014 [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES