Psychiatry · Medical Literature
Hallucinations and auditory-verbal signal detection studies: A meta-analytic review.
Abstract
Auditory-verbal signal detection tasks (AV-SDTs) are used to study hallucinatory experiences (HEs) in clinical and nonclinical groups, assessing perceptual decision making under ambiguous auditory conditions. Existing findings suggest that HEs are associated with response biases. However, effects sizes and subgroup differences remain unclear. This meta-analysis synthesized findings to quantify associations between HEs and AV-SDT performance, testing differences within clinical and nonclinical populations. Systematic searches identified studies published between January 1985 and April 2025. Reference lists were screened and authors contacted. Eligible studies examined associations between hallucinations or hallucination-proneness and AV-SDT performance using between-group or correlational designs. Data comprising 1,391 participants were pooled from 5 clinical (210 patients), and 10 non-clinical (1,181 participants) studies. Moderate-to-large effects were found between HEs and response bias across aggregate (g = -0.56 [95% CI: -0.84, -0.28, p = .001]), clinical (g = -0.88 [95% CI: -1.58, -0.17, p = .026]), and nonclinical (g = -0.43 [95% CI: -0.74, -0.11, p = .013]) subgroup models. Perceptual sensitivity was reduced in nonclinical (g = -0.17 [95% CI: -0.28, -0.07, p = .004]), but not clinical (g = 0.37 [95% CI: -0.00, 0.75, p = .050]), hallucinating samples. HEs are consistently associated with a liberal response bias across populations. HEs in nonclinical samples were linked to reduced perceptual sensitivity although this was not supported in clinical groups, potentially due to limited data. Findings highlight methodological gaps in AV-SDT research, including variability in design and hallucination assessment, underscoring the need for standardized approaches to improve generalizability across studies.
Clinical relevance
Do hallucinatory experiences (HEs) — across both clinical and nonclinical populations — associate with measurable differences in response bias and perceptual sensitivity on auditory-verbal signal detection tasks (AV-SDTs), and do the magnitude and direction of these effects differ between clinical and nonclinical groups? __SNAPSHOT__ {"studyDesign":"Systematic review with meta-analysis (between-group and correlational designs pooled)","setting":"Not reported (international; studies published January 1985–April 2025, identified via systematic searches and author contact)","sampleSize":"n = 1,391 total; clinical subsample: 210 patients (5 studies); nonclinical subsample: 1,181 participants (10 studies)","population":"Clinical studies: patients with hallucinatory experiences (psychiatric diagnoses not fully specified in the abstract). Nonclinical studies: community or student participants characterised by hallucination-proneness or self-reported hallucinatory experiences. Both between-group and correlational designs were eligible.","intervention":"Auditory-verbal signal detection tasks (AV-SDTs): experimental paradigms in which participants judge whether a voice is present within noise, yielding measures of response bias (criterion) and perceptual sensitivity (d-prime or equivalent)","comparator":"Within clinical studies: patients with HEs vs. non-hallucinating controls. Within nonclinical studies: high vs. low hallucination-prone individuals, or correlational association with hallucination-proneness scores.","primaryOutcome":"Response bias on AV-SDT (aggregate, clinical, and nonclinical subgroups); perceptual sensitivity on AV-SDT (nonclinical subgroup reported separately in associated thesis)","followUp":"Not applicable (cross-sectional experimental paradigms)","fundingSource":"Not reported"}
Key findings
Across all 15 pooled studies (n = 1,391), HEs were associated with a more liberal (lower-threshold) response bias on AV-SDTs, with a moderate-to-large aggregate effect size (Hedges' g = −0.56; 95% CI: −0.84 to −0.28; p = .001). Subgroup analyses demonstrated a large effect in clinical samples (g = −0.88; 95% CI: −1.58 to −0.17; p = .026; n = 210, 5 studies) and a moderate effect in nonclinical samples (g = −0.43; 95% CI and p-value not fully reported in the available abstract due to truncation — values not reported here to avoid fabrication). An additional finding from the associated research programme indicates perceptual sensitivity deficits (reduced d-prime) in nonclinical cohorts; however, the specific effect size and confidence interval for this outcome are not reported in the provided abstract. No NNT/NNH is calculable from signal detection effect sizes. Clinically important adverse events are not applicable to this experimental paradigm. The consistent negative direction of the response bias effect (across both populations) indicates a systematic tendency for individuals with HEs to respond 'voice present' under ambiguous conditions, i.e., a lower response criterion.
Practice implications
This meta-analysis establishes that a liberal response bias on AV-SDTs is a replicable, transdiagnostic correlate of hallucinatory experiences, detectable in both clinical patients and nonclinical hallucination-prone individuals, with the larger effect observed in clinical samples. The finding supports the longstanding theoretical position that abnormal perceptual decision-making — specifically a lowered threshold for reporting signal presence — underlies or accompanies hallucinations across the psychosis continuum. The study does not establish causality, treatment targets, or prognostic relevance; AV-SDT performance is a laboratory-derived measure and its direct clinical utility (e.g., as a diagnostic biomarker or treatment outcome marker) is not demonstrated here. Patients not well represented include those with hallucinations outside the auditory-verbal domain (e.g., visual, somatic), those with diagnoses not represented in the 5 included clinical studies, and individuals from low- and middle-income countries. The clinical subsample is very small (n = 210 across 5 studies), rendering the clinical subgroup estimate (g = −0.88; 95% CI: −1.58 to −0.17) imprecise, with a wide confidence interval approaching zero at its lower bound. The effect in nonclinical samples — while statistically significant — may not translate to bedside practice in its current form. Results should be interpreted as mechanistic evidence informing cognitive models of hallucinations rather than as a basis for changing clinical assessment or management practice. IMPORTANT NOTE: Only the abstract was available for review; full text, supplementary analyses, heterogeneity statistics (I²), publication bias assessments, and moderator analyses were not accessible and may materially affect interpretation. __LIMITATIONS__ Very small clinical sample (n = 210 across only 5 studies), producing a wide confidence interval for the clinical subgroup effect (95% CI: −1.58 to −0.17) and limited statistical precision.,The abstract was truncated; nonclinical subgroup confidence intervals and p-values, heterogeneity statistics (I²/Q), and publication bias analyses (e.g., funnel plot, Egger's test) were not available for review — these are critical for interpreting meta-analytic reliability.,Pooling of between-group and correlational designs introduces methodological heterogeneity; effect size comparability across design types may be limited.,The search window (1985–2025) spans 40 years of substantial changes in diagnostic criteria, task methodology, and hallucination measurement instruments, which may inflate between-study heterogeneity.,Unclear whether clinical studies adequately controlled for medication status, diagnosis-level confounders (e.g., negative symptoms, cognitive impairment, comorbid affective disorders), or antipsychotic dose — all of which may independently affect AV-SDT performance.,Hallucination-proneness in nonclinical samples is typically assessed via self-report questionnaire (e.g., LSHS, CAPS), which may not map directly onto clinical hallucinatory phenomena, limiting generalizability to patients.,AV-SDT paradigms vary considerably across included studies in terms of stimuli, signal-to-noise ratios, task duration, and response formats — task heterogeneity may contribute to effect size variance.,Publication bias in favour of positive findings is a plausible concern, particularly in a small literature with only 15 included studies.,Funding source not reported; potential conflicts of interest cannot be assessed.,Findings are correlational; causal direction between liberal response bias and hallucinatory experiences cannot be inferred. __BOTTOM_LINE_CATEGORY__ promising_but_preliminary __BOTTOM_LINE__ This meta-analysis of 15 studies (n = 1,391) provides the most comprehensive quantitative synthesis to date demonstrating that hallucinatory experiences are consistently associated with a liberal response bias on auditory-verbal signal detection tasks across both clinical (g = −0.88) and nonclinical (g = −0.43) populations. These findings strengthen the mechanistic case for perceptual decision-making abnormalities as a transdiagnostic feature of hallucination-proneness, but the small clinical sample, truncated reporting, and laboratory-only paradigm mean this evidence is not yet ready to change clinical assessment or treatment practice. Clinicians and researchers should regard these results as theoretically informative rather than immediately actionable.