Accent bias in evaluations despite no detected learning difference

Updated Sep 10, 2026

Sosa-Herrera and colleagues found that students' ratings could differ by an instructor's accent even when a learning test detected no statistically significant difference. Their study also reports mixed results for a brief prompt intended to reduce bias.

What the experiments show

In two laboratory experiments, students were randomly assigned a short lecture narrated with a Mandarin or American English accent. Before evaluating the instructor, one condition received awareness instructions; the control proceeded to the evaluation.

The Mandarin-accented instructor received less favourable ratings in both experiments. There was no statistically significant difference on the multiple-choice learning assessment. This does not establish that students understood everything or that learning was exactly equal.

The prompt reduced differences in overall ratings in Experiment 1. It was less effective in Experiment 2, which used different lecture material and a more diverse student sample. The experiments therefore support a mixed finding, not a reliable remedy for bias in every course.

Reviewing evaluations in practice

Our recommendation is to treat an unexpected rating difference as a question for review rather than a judgement about an individual. Check the teaching context, the question asked, who responded and the other evidence available. A difference between staff groups does not identify its cause on its own.

Comments may help clarify the experience being described, but categories such as communication or clarity do not automatically distinguish bias from a substantive concern. Keep a route for students to raise specific difficulties and give staff an opportunity to provide context. The governance checklist offers questions about access and interpretation when feedback informs sensitive decisions.

FAQ

Should every survey add an awareness prompt?

This study does not establish a universal benefit. A local pilot should define the intended outcome and evaluate whether the prompt changes it.

Does a non-significant learning result mean equivalence?

No. Establishing equivalence requires suitable methods and an explicit margin. Failure to detect a difference is a more limited finding.

Reference

Sosa-Herrera, E., Silaj, K., Keushkerian, M. J., & Paquette-Smith, M. (2026). Evaluating the effectiveness of an awareness-based intervention on reducing accent bias in students' evaluations of instructors. Assessment & Evaluation in Higher Education.

Correction, 7 September 2026: corrected the interpretation of the learning test and retained the mixed intervention result.