Workshop

<< Back to the main conference page Neurolinguistics in Sweden, NLS2026

Stable perception of spectral information across contexts: From normalization to adaptive category representations

One of the striking features of human speech perception is its stability: despite substantial between-talker differences arising from variation in vocal tract physiology, language background, and social factors, listeners typically understand speech with relative ease. This workshop focuses on the auditory mechanisms thought to contribute to such stable perception by normalizing the speech signal for differences in vocal tract size and/or shape.

More specifically, the workshop aims to connect research on the early normalization of formant and other spectral information with work on downstream adaptive mechanisms beyond formants. Support for the existence of formant/spectral normalization has come from behavioral experiments on speech perception, brain imaging studies, brain stem recordings, and cross-species comparisons. However, important questions remain concerning the computations underlying these rapid and seemingly automatic mechanisms, as well as the extent to which they interact with top-down information from higher-level representations of linguistic categories and contexts.

By bringing together researchers with complementary perspectives from phonetics, cognitive science, neuroscience, and speech technology, the workshop aims to foster focused discussions of open questions concerning speech perception, adaptation, and formant/spectral normalization.

Register for the workshop

Thanks to generous support from Riksbankens Jubileumsfond, the workshop is open to all and free of charge.

Please complete this free registration if you intend to attend the workshop but have not registered for NLS 2026 (so that we can order adequate amounts of coffee/fika).

Registration

Workshop programme

08.30 Welcome address and overview (Anna Persson).

08.45–09.25 Ediz Sohoglu (University of Sussex): Perceptual learning of modulation filtered speech.

09.30–10.10 Kasia Hitczenko (University of Delaware): Speech category imbalances hinder normalization in naturalistic data.

10.10–10.30 Coffee break.

10.30–11.10 Santiago Barreda (University of California, Davis), T Florian Jaeger (University of Rochester) & Anna Persson (University of Oslo): A one-shot model for joint inferences of talker physiology and vowel recognition.

11.15–11.55 Ondrej Šuch (Slovak Academy of Sciences): Phonetic explanation of speaker identification systems.

11.55–12.00 Wrapping up.