Measuring Paranormal Beliefs: Reconceptualization and Empirical Validation of the Paranormal Belief Construct

Matas, H., Munar, E., Ballester, L., & Hernández-Lloreda, M. J. (2024). Measuring Paranormal Beliefs: Reconceptualization and Empirical Validation of the Paranormal Belief Construct. International Journal of Transpersonal Studies (Advance Publication).

AI Assessment

A large psychometric study testing whether a widely used questionnaire, the extended Revised Paranormal Belief Scale, measures a coherent, well-structured construct. With 6,584 Spanish participants, the authors fit a hierarchical model (twelve subscales feeding three factors and one overall paranormal-belief factor), found strong and significant factor loadings (.65 to .91), replicated the structure across two independent subsamples, and showed the scale works equivalently for men and women. This is careful measurement work, and it is important to be clear about what it does and does not test: it validates a tool for measuring how strongly people believe in the paranormal, not whether paranormal phenomena are real. Its main methodological wrinkle is that adequate fit required removing the reverse-worded items, a known and debated issue with this scale that can mask response biases. This audit reports what the study found and how; it concerns belief measurement, not the existence of psi.

Provenance

Publication. International Journal of Transpersonal Studies, 2024 (Advance Publication), hosted on Digital Commons at CIIS.

Study type. A psychometric construct-validation study using confirmatory factor analysis. It measures belief, not paranormal phenomena.

Authors. Hugo Matas, Enric Munar, Lluis Ballester (University of the Balearic Islands), and Maria José Hernández-Lloreda, working in evolution-and-cognition and pedagogy groups.

Data availability. Not stated in the article.

Source basis. Every figure below is taken from the article’s own Abstract and Results.

What the paper reports

The Revised Paranormal Belief Scale (RPBS) is the most widely used instrument for measuring paranormal belief, but its structure has long been debated. Following an extended, hierarchical reconceptualization, the authors test whether that model fits well in a large sample and whether it measures the same construct equivalently across sexes.¹ Their finding is that, after some item adjustments, the extended model fits adequately, its factors are strongly defined, and it is invariant by sex, suggesting the scale could be generalized to other cultures while a precise definition of “paranormal belief” still needs to be agreed.

After reversed items were removed, the model demonstrated an adequate fit, significant factor loadings and invariance between sexes. The results suggest the possibility of generalizing the RPBS to other cultures.

How it was run

• Sample. 6,584 participants, a semi-probabilistic, self-selected sample recruited via the University of the Balearic Islands intranet, completing the extended RPBS on a 7-point Likert scale.
• Model. A second-order confirmatory factor analysis: parcels representing 12 subscales load onto 3 first-order factors (Agents, Signs, and Vital Power), which load onto one second-order Paranormal Beliefs factor.
• Estimation. AMOS version 21 with Maximum Likelihood estimation and 1,000 bootstrap resamples; 244 multivariate outliers were removed using Mahalanobis distance.
• Invariance. Multigroup CFA tested configural, metric, scalar, and residual invariance by sex, and the structure was checked across two independent subsamples.

Results, as reported

Values are reproduced from the article’s Abstract and Results. The scale, in its trimmed form, behaves as a coherent, sex-invariant measure of paranormal belief, with the important caveat that adequate fit depended on dropping the reverse-worded items.

Eleven-dimension audit

Pre-registration

Not preregistered. The analysis is confirmatory in the sense that it tests a pre-existing model, but the decisions to drop three reversed items and one further item to achieve fit are data-driven, and without a registered plan a reader cannot tell how many alternative trimmings were considered.

Randomization

Not applicable: this is a survey with a self-selected sample and no experimental assignment. The relevant sampling concern is that the respondents came from a single university intranet.

Sensory leakage

Not applicable to a self-report questionnaire. The analogous validity threat is response bias, particularly acquiescence (a tendency to agree regardless of content), which is directly relevant to the reversed-item issue discussed below.

Blinding

Not applicable; participants self-report their own beliefs and there are no raters or conditions to blind.

Optional stopping

Not applicable to a fixed survey dataset. The analogous analytic choices are the removal of 244 outliers and of the misfitting items, both of which are defensible but are decisions that shape the reported fit.

Outcome measure

Standard and appropriate psychometric outcomes (model fit indices, factor loadings, invariance tests). The measure is of belief strength and structure, not of any paranormal ability, which is the key framing point for this page.

Effect size

In psychometric terms the indicators are strong: factor loadings from .65 to .91 indicate well-defined factors, and the fit is reported as adequate. These are properties of the measurement model rather than effects about phenomena.

Multiple comparisons

Multiple analyses were run (the CFA, a four-level invariance sequence, and subsample replication), which is standard for validation work. The multiplicity concern lies more in the item-selection decisions than in the confirmatory tests themselves.

Internal replication

A genuine strength: the structure held across two independent subsamples of the large dataset, which is stronger evidence of a stable model than a single fit would be.

External replication

The sample is from a single culture and institution, so external generalization is untested; the authors explicitly frame cross-cultural generalization as future work.

Transparency

The methods are described in enough detail to reproduce (software, estimator, bootstrap, invariance sequence, outlier rule), and the item removals are reported rather than hidden. The transparency limits are the absence of preregistration and of a stated data-availability provision.

The adversarial record

• The reversed-item removal. The most substantive concern is that adequate fit required dropping the reverse-worded items. This is a well-known problem with the RPBS: reverse-worded items often misbehave precisely because they catch acquiescent or inattentive responding, and a model that fits well only after they are removed may be describing a cleaner but partly artefactual construct, with acquiescence bias absorbed rather than controlled.
• Sample. Despite its size, the sample is self-selected from one university’s intranet, so representativeness, not statistical power, is the limiting factor for generalization.
• Researcher choices, unregistered. Removing four items and 244 outliers to reach fit, without preregistration, leaves room for the trimming that best fit the model to have been selected after the fact.
• What it does not show. A reader should not over-read the paper: it validates a way to measure belief in the paranormal, and says nothing about whether those beliefs are true. That is a feature, not a flaw, but it is worth stating plainly on an evidence-focused site.
• What is genuinely strong. This is rigorous measurement work: a very large sample, a proper confirmatory factor analysis with formal invariance testing, replication across subsamples, and transparent reporting of the analytic decisions. As a validation of a belief-measurement instrument it is well executed; its conclusions simply live in the domain of psychometrics rather than of paranormal evidence.