Indirect (implicit) and direct (explicit) self-esteem measures are virtually unrelated: A meta-analysis of the initial preference task

Jakob Pietschnig, Georg Gittler, Stefan Stieger, Michael Forster, Natalia Gadek, Andreas Gartus, Krisztina Kocsis-Bogar, Bettina Kubicek, Marko Lüftenegger, Jerome Olsen, Roman Prem, Nina Ruiz, Benjamin G Serfas, Martin Voracek

Research output: Journal article (peer-reviewed)Journal article

6 Citations (Scopus)


BACKGROUND: The initial preference task (IPT) is an implicit measure that has featured prominently in the literature and enjoys high popularity because it offers to provide an unobtrusive and objective assessment of self-esteem that is easy to administer. However, its use for self-esteem assessment may be limited because of weak associations with direct personality measures. Moreover, moderator effects of sample- and study-related variables need investigation to determine the value of IPT-based assessments of self-esteem.

METHODS: Conventional and grey-literature database searches, as well as screening of reference lists of obtained articles, yielded a total of 105 independent healthy adult samples (N = 17,777) originating from 60 studies. Summary effect estimates and subgroup analyses for potential effect moderators (e.g., administration order, algorithm, rating type) were calculated by means of meta-analytic random- and mixed-effects models. Moreover, we accounted for potential influences of publication year, publication status (published vs. not), and participant sex in a weighted stepwise hierarchical multiple meta-regression. We tested for dissemination bias through six methods.

RESULTS: There was no noteworthy correlation between IPT-based implicit and explicit self-esteem (r = .102), indicating conceptual independence of these two constructs. Effects were stronger when the B-algorithm was used for calculation of IPT-scores and the IPT was administered only once, whilst all other moderators did not show significant influences. Regression analyses revealed a somewhat stronger (albeit non-significant) effect for men. Moreover, there was no evidence for dissemination bias or a decline effect, although effects from published studies were numerically somewhat stronger than unpublished effects.

DISCUSSION: We show that there is no noteworthy association between IPT-based implicit and explicit self-esteem, which is broadly consistent with dual-process models of implicit and explicit evaluations on the one hand, but also casts doubt on the suitability of the IPT for the assessment of implicit self-esteem on the other hand.

Original languageEnglish
Article numbere0202873
Pages (from-to)e0202873
JournalPLoS ONE
Issue number9
Publication statusPublished - Sept 2018


  • Humans
  • Psychological Tests
  • Self Concept


Dive into the research topics of 'Indirect (implicit) and direct (explicit) self-esteem measures are virtually unrelated: A meta-analysis of the initial preference task'. Together they form a unique fingerprint.

Cite this