Modality Matters for Extraversion: A Big Five Meta-Regression in Online Learning Environments

Abstract

The author's parent meta-analysis (Research Square preprint, DOI 10.21203/rs.3.rs-9513298/v1) pooled k = 9–10 primary studies per Big Five trait across online-learning settings; only Conscientiousness retained a robust positive correlation with achievement (r = 0.167 [0.089, 0.243]). Modality was pre-registered as a moderator but not executed because no cell met the k ≥ 10 rule. The present paper completes that registered-but-unrun moderator. Modality codes were re-checked against original PDFs for four primary-pool studies whose preprint extraction was blank (A-15 Elvers; A-23 Rodrigues; A-26 Wang; A-30 Kaspar). REML with HKSJ-adjusted CIs were used; a long-format weighted-OLS interaction model contrasted asynchronous (A) and mixed-online (M) instruction. Modality moderates the trait–achievement correlation in a trait-specific way: Q_between is highly significant for Extraversion (Q = 17.60, df = 1, p < .001), trend-level for Agreeableness (Q = 3.28, p = .070), and not significant for Conscientiousness, Neuroticism, or Openness. Asynchronous Extraversion shows r = −0.121 [−0.246, 0.007], reversing in mixed-online to r = +0.059 [−0.027, 0.145]; the Extraversion × Mixed interaction is +0.385 Fisher z, p = .009. The joint Wald test on all four interaction terms is χ²(4) = 13.64, p = .0085. Modality does not change the bottom line for Conscientiousness (r = 0.180–0.190 across A and M) but materially changes Extraversion, a trait the trait-only pool reported as null. For online course designers the message is asymmetric: support self-regulation regardless of synchrony, but expect Extraversion to be a silent disadvantage in async-only programmes.