Creativity is a hot topic for decades now. One of the most popular TED videos and of my most popular blog posts ever is also about this skill (although is it truly a skill?). In my post – also in our first Urban Myth book – we already mentioned one of the critical issues: there are many definitions. A new registered report by Warne, Golightly, and Black faces the same issue as they try to find a link between intelligence and creativity.

The study seemed to be very straightforward:

We administered two divergent thinking tests (the verbal and figural versions of the Torrance Tests of Creative Thinking; TTCT) with an intelligence test (the International Cognitive Ability Resource test; ICAR). We then subjected the subscores from these tests to confirmatory factor analysis to examine which of nine theoretically plausible models best fits the data.

But… it took a different turn:

Though our results were unable to answer the original research question of whether creativity and intelligence tests measure the same construct, we do believe that our results raise serious questions about the construct validity evidence of the TTCT-F and, to a lesser extent, the TTCT-V. This finding adds to earlier evidence showing that variance of TTCT scores may be strongly influenced by item format, content, and demand characteristics of specific TTCT tasks. Additionally, the poor fit of each TTCT test’s subscores to a congeneric measurement model seriously undermines the support for interpreting and using a global divergent thinking score. Any apparent, strong positive correlations among raw scores on the TTCT are likely a consequence of the confounding effect of fluency scores on the other scores. Our research presents a serious challenge to the validity of interpreting TTCT-V and TTCT-F scores as measuring a coherent latent construct. We recommend against using overall TTCT-F scores for non-research purposes and believe that creativity scholars should be skeptical about the ability of the TTCT-F to measure a coherent underlying construct. We also recommend caution in using TTCT-V scores for making decisions about examinees. Until there is a coherent measurement model, a scoring system, and interpretation that aligns with that model and with theory, the TTCT-F (and possibly the TTCT-V) should be confined to research settings only.

Abstract of the registered report in PlosOne:

Psychologists have investigated creativity for 70 years, and it is now seen as being an important construct, both scientifically and because of its practical value to society. However, several fundamental unresolved problems persist, including a suitable definition of creativity and the ability of psychometric tests to measure divergent thinking—an important component of creativity—in a way that aligns with theory. It is this latter point that this registered report is designed to address. We administered two divergent thinking tests (the verbal and figural versions of the Torrance Tests of Creative Thinking; TTCT) with an intelligence test (the International Cognitive Ability Resource test; ICAR). We then subjected the subscores from these tests to confirmatory factor analysis to examine which of nine theoretically plausible models best fits the data. Results show that none of the pre-registered models fit the data well, an ambiguous result that leaves unanswered the question of whether intelligence and divergent thinking tests measure the same construct. Exploratory (i.e., not pre-registered) measurement models of each test separately shows that the TTCT-F may not measure a coherent, unitary construct—leading to model misspecification when TTCT-F subtests were included in larger models. This study was conducted in accordance with all open science practices, including pre-registration, open data and syntax, and open materials (with the exception of copyrighted and confidential test stimuli). Materials are available at https://osf.io/8rpfz/.