Language comprehension requires integration of multiple cues, but the underlying mechanisms of how accentuation, as a significant prosodic feature, influences the processing of words with different levels of cloze probability remains unclear. This study exploits event-related potentials (ERPs) to examine the processing of accented and unaccented words with high-, medium-, and low-cloze probabilities embedded in the final position of highly constrained contexts during spoken sentence comprehension. Our results indicate that accentuation and cloze probability interact across the N400 and post-N400 positivity (PNP) time windows. Under the accented condition, N400 amplitudes gradually increased as cloze probability decreased. Conversely, under the unaccented condition, PNP amplitudes gradually increased as cloze probability decreased with a frontal distribution. These results suggest that the effect of predictability is influenced by accentuation, which is likely due to the processing speed and depth of the critical words, modulated by the amount of attentional resources allocated to them.