Eyebrow movements, particularly furrows, were found to signal problems in understanding during face-to-face conversations, according to a two-part study.
In an analysis of 586 other-initiated repair cases from Dutch dyadic conversations, 45.7% of verbal repair initiations co-occurred with eyebrow movements. Furrows and raises appeared in nearly equal proportions—50.4% and 49.6%, respectively—but furrows were significantly more likely to accompany restricted clarification requests, while raises were more common in offers seeking confirmation.
Furrows were more strongly associated with restricted requests than either open requests or restricted offers.
In 491 repair cases with responses, 56.3% of furrow-linked initiations led to clarifications, compared with 37.4% for verbal-only cases and 29.7% for those with eyebrow raises. This association held after adjusting for verbal format.
Eyebrow furrows were more strongly associated with clarifications compared to both eyebrow raises and verbal-only repair initiations.
"The findings demonstrate that eyebrow movements play a communicative role in initiating repair during conversation rather than being merely epiphenomenal," said lead author Paul Hömke of the Max Planck Institute for Psycholinguistics, The Netherlands and colleagues in Royal Society Open Science.
To assess causality, researchers conducted a virtual reality experiment in which avatars responded with nods, furrowed brows, or no feedback as participants answered questions. Eyebrow furrows led participants to give significantly longer responses—by an average of 3.77 seconds—compared to nods alone, indicating speakers perceived the furrow as a cue for more information. This effect persisted even when controlling for hesitations, suggesting speakers added substantive content rather than simply pausing more.
In 11 cases from the corpus study, eyebrow furrows prompted clarification responses without any verbal prompt, suggesting furrows alone may initiate repair. All resulted in clarification, with no instances of confirmation or disconfirmation.
The presence or timing of eyebrow movements did not significantly affect repair latency. Mean response times were similar for verbal-only initiations (241 ms), those with concurrent eyebrow movements (232 ms), and those with early eyebrow movements (248 ms).
The study combined analysis of 3 Dutch corpora totaling 298 minutes of conversation for the first phase, while the experimental phase included 34 participants interacting with virtual avatars. Interestingly, individual differences in empathy and fear of negative evaluation did not significantly modulate the effects. Coders achieved high inter-rater reliability across all measures.
The authors reported no conflicts of interest.