When having a conversation in a multitalker setting (‘cocktail party’ listening), listeners are required to selectively attend a to-be-attended talker and to ignore speech from other competing talkers. This talk discusses a research project that investigated which acoustic characteristics of the speech from competing talkers interferes with the perception of attended speech, and which do not. The surrounding acoustic context in which a given target speech sound is heard is known to directly impact perception. For instance, a vowel ambiguous between a short and a long vowel in Dutch is perceived differently depending on the preceding sentence’s speech rate (‘temporal context effect’). Similarly, the perception of a vowel ambiguous between /ɪ/ and /ɛ/ is influenced by the first formant characteristics of the preceding sentence (‘spectral context effect’). Using these context effects, we tested whether listeners can successfully ignore the temporal and spectral characteristics of a competing talker when interpreting ambiguous speech sounds. Outcomes indicate that, while spectral context effects are strongly modulated by selective attention, temporal context effects are immune to selective attention. Thus, this project demonstrates that listeners are very successful at ignoring a competing talker’s spectral voice properties, but have a hard time ignoring their speech rate.