Authors: Jorge Mejia¹; Leanne Pul²; Jan-Willem Wasmann²; Noortje van der Mast²; Cris Lanting²; Marc van Wanrooij³
¹National Acoustic Laboratories (NAL); ²Radboud Universitair Medisch Centrum; ³Radboud University
View presentation by clicking here. Want to do the test yourself? Feel free to join our ULIXES Quest, a demonstration that allows curious people to complete the ULIXES Quest and compare their results with others.
Background: Standard speech-in-noise tests do not measure how listeners combine auditory and non-auditory cues in real-world communication. People with hearing loss may rely more on visual or contextual cues, but this cue weighting is not routinely assessed. ULIXES, built on NAL’s ULI-test, is a home-based, tablet-compatible speech assessment designed to quantify textual visual capture. Congruent written cues may support perception, while conflicting cues may bias responses toward the distractor. We tested whether capture increases at poorer SNRs and decreases as speech becomes clearer, consistent with reliability-weighted cue use. A capture transition SNR (CT-SNR) was derived as a candidate clinical index.
Method: In a pilot study (N = 4; 372 trials), participants listened to CVC non-words in noise while viewing written options. The written leading cue either matched the target (congruent; 250 trials) or conflicted with it (conflict; 122 trials). SNR varied adaptively from +12 to -15 dB. Outcomes were accuracy, conflict-trial visual distractor capture, and response time. SNR bands were Low (≤0 dB), Mid (+1 to +4 dB), and High (>+4 dB). Mixed-effects models tested SNR and trial-type effects. CT-SNR was defined as the SNR where capture reached 50%.
Results: Visual capture showed a clear SNR-dependent pattern. On conflict trials, capture was highest at low SNR (58%) and declined at high SNR (8%; p ≈ 10⁻⁵). Conflict-trial accuracy was lower than congruent-trial accuracy in noise (30% vs 52%) but converged at high SNR (≈82% for both). Models are exploratory.
Conclusion: These pilot data show that textual cues can systematically bias speech reports in noise. Conflict-trial distractor capture was the clearest marker of cue reliance across SNRs. The capture-SNR function and CT-SNR may provide useful measures of multisensory cue weighting beyond accuracy alone. The language-independent CVC non-word format supports broad clinical use.
