Six Questions to Gauge How Openly We Trust Knowledge: New Short Scale Validated

How willing are we to learn from other people? That deceptively simple question sits at the heart of a growing body of psychological research on what scientists call the epistemic stance: an individual’s characteristic openness to receiving knowledge from the social world. Some people readily take in what others tell them, weigh it, and use it to update their understanding. Others respond with suspicion, dismissing new information as manipulation or lies. A third group accepts almost anything they are told, however dubious the source. A new study published in PLOS Mental Health now offers researchers a remarkably compact tool for measuring these three tendencies, distilling an established sixty-item questionnaire into just six questions.

The instrument, known as the ETMCQ-R-6, is a brief version of the Epistemic Trust, Mistrust and Credulity Questionnaire, a measure developed within the framework of mentalization research led by Peter Fonagy and colleagues at University College London. The original revised questionnaire assesses three distinct dimensions of how people relate to social sources of knowledge: epistemic trust, the willingness to treat communication from others as relevant and trustworthy; epistemic mistrust, a defensive stance in which information is discounted as irrelevant, insincere or hostile; and epistemic credulity, an uncritical receptivity that accepts information without adequate scrutiny. The new validation study, conducted by David Riedl, Henry Delamain, Peter Fonagy, Astrid Lampe, Samuel Eke, Tobias Nolte, Hanna Kampling, Johannes Kruse and Chloe Campbell, tested whether a six-item short form could preserve the psychometric strength of the full measure while being light enough for large-scale surveys and clinical screening.

The research team drew on a quota-balanced community sample of 515 adults from the United Kingdom, a sampling strategy designed to mirror the general population across key demographic strata rather than relying on the convenience samples that often limit psychological measurement studies. From the longer questionnaire, the researchers used a stepwise item reduction procedure, guided by statistical criteria, to select the two items per subscale that best captured each of the three constructs. Confirmatory factor analysis, the standard technique for testing whether a hypothesized structure of latent traits fits observed response data, was then applied to see whether the reduced set still measured three separate but related dimensions.

The results were encouraging for such a radically shortened scale. The three-factor structure with two items per subscale showed acceptable fit to the data. The confirmatory fit index, adjusted robustly for the ordinal nature of the response data, reached 0.978, and the Tucker-Lewis index came in at 0.945, both close to or above conventional benchmarks of 0.95. The standardized root mean square residual, a measure of the average discrepancy between observed and model-implied correlations, was a very low 0.029. The root mean square error of approximation, which penalizes model complexity, was 0.073 with a ninety percent confidence interval spanning 0.042 to 0.107, a value that sits within the acceptable range for a model with so few degrees of freedom. In plain terms, the six items behaved as though they were indeed tapping three coherent dimensions of epistemic stance.

Reliability, the consistency with which the items measure their intended construct, presented a more nuanced picture. The trust subscale achieved a Cronbach’s alpha of 0.75 and the credulity subscale 0.77, both within the conventional acceptable range for short scales. The mistrust subscale, however, fell just below the usual threshold, with alpha, omega and rho all at 0.68. With only two items per scale, such coefficients are inherently constrained, since reliability estimates depend partly on the number of items. The authors interpret this as acceptable for a screening and research instrument, while acknowledging that the mistrust subscale should be used with awareness of its slightly weaker internal consistency.

A critical question for any shortened questionnaire is whether it measures the same thing across different groups of people. The team therefore tested measurement invariance across gender and age groups, a sequence of increasingly stringent statistical constraints. Configural invariance, meaning the same factor structure holds across groups, was supported, as was metric invariance, meaning the items relate to their underlying constructs with equal strength in each group. Scalar invariance, which permits meaningful comparison of mean scores between groups, and structural invariance were also established. Strict invariance, the most demanding level requiring equal item residual variances, was not supported, which is a common finding and generally not fatal for comparing scale scores. The practical upshot is that researchers can meaningfully compare average levels of trust, mistrust and credulity between men and women and across age bands using this short form.

Perhaps the most reassuring finding concerns the relationship between the short and long forms. Correlations between corresponding subscales of the ETMCQ-R and the six-item version ranged from 0.83 to 0.87, and the latent correlations between the constructs measured by the two versions, expressed as phi coefficients, ranged from 0.95 to 1.00. These are extraordinarily high values, indicating that the six items recover almost all of the information carried by the full-length instrument at the level of the underlying constructs. For epidemiologists and survey researchers, this means the short form can substitute for the long form in contexts where questionnaire space and respondent burden are at a premium, without sacrificing the core measurement.

But a scale can be reliable and structurally sound and still be useless if it does not predict anything of consequence. The team therefore examined incremental predictive validity using hierarchical regression analyses, asking whether the six-item subscales added explanatory power for mental health outcomes beyond what other measures already provided. The results showed significant incremental validity for the epistemic mistrust and credulity subscales across psychopathology outcomes, meaning that these two brief dimensions carried unique information about psychological symptoms. The trust subscale showed a different, more specific pattern: it demonstrated domain-specific incremental validity for attachment avoidance, the tendency to suppress closeness and dependence on others. This pattern fits the theoretical model from which the questionnaire emerged, in which secure attachment fosters openness to social knowledge, while avoidant strategies and suspicious or uncritical stances each disrupt the transmission of knowledge in characteristic ways.

The theoretical stakes of this work extend well beyond psychometrics. Epistemic trust has been proposed as a key mechanism in both the origins of psychopathology and its treatment. According to this account, children who cannot trust social communication, whether because of mistrust born of adverse experiences or because of indiscriminate credulity, fail to acquire the culturally transmitted knowledge that supports resilient mental health. Conversely, effective psychotherapy may work in part by reopening a patient’s epistemic channel, making them once again receptive to learning from others. A six-item measure that reliably captures these three stances gives researchers a practical instrument for testing such claims in large samples, in longitudinal designs, and in settings where a sixty-item battery would be impractical, from national mental health surveys to routine clinical intake.

The authors position the ETMCQ-R-6 as a brief, psychometrically stable assessment of epistemic stance suitable for epidemiological research, and the evidence from this U.K. general population sample supports that characterization. As with any short form, there are trade-offs: the two-item mistrust subscale sits just below conventional reliability thresholds, and strict measurement invariance was not established, so fine-grained comparisons of item-level error structure across groups should be made cautiously. Yet the combination of acceptable fit, strong correlations with the parent instrument, invariance across gender and age, and demonstrated incremental validity for psychopathology and attachment outcomes makes a compelling case that six carefully chosen questions can meaningfully measure how we trust, doubt and swallow the words of others. In an era when misinformation, conspiracy beliefs and institutional distrust dominate public debate, a validated micro-scale for the psychology of social knowledge may prove to be a small instrument with outsized reach.

Subject of Research: Validation of a six-item short form of the Epistemic Trust, Mistrust and Credulity Questionnaire in a U.K. general population sample

Article Title: Can we trust this short scale? Validation of a Brief Self-Report Measure of Epistemic Trust, Mistrust and Credulity (ETMCQ-R-6) in a sample of the U.K. general population

Article References: Riedl, D., Delamain, H., Fonagy, P., Lampe, A., Eke, S., Nolte, T., Kampling, H., Kruse, J., & Campbell, C. (2026). Can we trust this short scale? Validation of a Brief Self-Report Measure of Epistemic Trust, Mistrust and Credulity (ETMCQ-R-6) in a sample of the U.K. general population. PLOS Mental Health, 3(8), e0000591. https://doi.org/10.1371/journal.pmen.0000591

Image Credits: AI Generated

DOI: 10.1371/journal.pmen.0000591

Keywords: epistemic trust, epistemic mistrust, credulity, questionnaire validation, psychometrics, confirmatory factor analysis, measurement invariance, mental health, attachment avoidance, social cognition, mentalization, PLOS Mental Health