Speech unfolds rapidly and rarely happens under optimal conditions. Everyday listening requires the brain to extract meaning from incomplete and overlapping acoustic signals. Speech comprehension is thought to benefit from predictive mechanisms that integrate early cues to anticipate upcoming information. In Swedish, prosodic patterns (word accents: Accent 1 and Accent 2) on the word stem signal different suffix continuations, allowing listeners to generate expectations about the upcoming information before the word fully unfolds. This study investigated how background noise affects predictive use of such prosodic information.

Speech-shaped noise was used to mask the speech signal. Electroencephalography (EEG) was recorded while adult native Swedish listeners with normal hearing (n = 22, mean age = 22.5 years, SD =2.46, range = 18–28) heard sentences containing target words with a word accent followed by either an incorrect suffix (mismatch) or a correct suffix (match). Stimuli were presented in quiet and in two different signal-to-noise ratios (SNRs): 0 dB SNR and -5 dB SNR.

Behavioral results showed that predictive speech perception was affected by both listening condition and word accent. Response times increased in noise and mismatch trials. In the EEG results, mismatched stimuli elicited N400–P600 activity in quiet speech but these responses weakened when the speech signal was masked by noise. Crucially, the two prosodic patterns differed in their resistance to noise: accent 1 remained an effective predictive cue across conditions, whereas accent 2 lost its reliability as a cue when the masking increased. This difference suggests that prosodic cues vary in their acoustic informativeness and in how strongly they engage predictive mechanisms under adverse listening conditions.

These findings suggest that noise constrains—yet does not eliminate—the predictive use of prosody during speech comprehension. Prosodic cues continue to guide comprehension when the speech signal is degraded, though with less precision than in clear speech. This gradual weakening of predictive processes highlights how the brain adapts to uncertainty in natural listening and could offer a basis for future research on how such mechanisms operate in populations facing everyday hearing challenges.