Performance-based metacognitive tests versus self-report: what does prediction tell us?

Background: The measurements of metacognition through performance-based tasks are better predictors of academic performance than those based on self-report tests, but evidence on the prediction of academic performance by standardized performance-based metacognition tests is scarce. The reason is that there are few tests of this nature with psychometric evidence of validity and reliability. Only a single study with Honduran university students compared the prediction of academic performance by a standardized performance-based test, and a self-report test in which both measure cognition regulation, a metacognitive construct. The results indicated that only the standardized performance-based test predicts academic performance, and the measures of these tests are not correlated.

Objective: Two hypotheses are investigated in this article: (1) performance-based metacognitive tests predict academic performance better than self-report metacognitive tests; (2) there is a null correlation between measures of cognition regulation from performance-based standardized tests and self-report tests.

Method: A sample of 264 university students and graduates from Brazil, with an average age of 21.1 years, is used in the study. The majority are female, from private institutions, and enrolled in humanities and social sciences courses. The Meta-Text was used as the standardized performance-based test, and the self-report test was the Metacognitive Self-Regulation Scale of the Motivated Strategies for Learning Questionnaire (MSLQ). The predictors were cognition regulation, measured by both tests, and judgment, measured by the Meta-Text. The outcome was the overall score on the National High School Exam, a large-scale educational assessment for university admission.

Results: Only the regulation of cognition measured by Meta-Text predicts academic performance (β = 0.47, CI 95% [0.36, 0.58]). The correlations between the test measures were null (r = .002, p = .974).

Conclusion: The evidence corroborated both hypotheses and raises doubts about the quality of self-report tests for measuring cognition regulation. It also indicates that standardized performance-based tests have a similar predictive capacity to tasks that require performance. This result is promising because standardized tests are easy to apply and correct, allowing studies to be carried out on large samples, while performance-based tasks require a complex process, only feasible in studies on small samples.