Speech is almost never delivered in ideal quiet conditions. On the contrary, the acoustic signal reaching a listener's ears is degraded by background noise and reverberations. The current study investigates the perception of the voicing contrast of initial stops in English by Polish non-native listeners. Previous research showed that Polish learners do not match native speakers of English in production and perception of English voiced and voiceless stops, which results from different phonetic implementations of voicing in the two languages. In the current study, two groups of Polish listeners recognised voicing of English initial stops in one-syllable words both in quiet and in six-talker babble. The results revealed different patterns of recognition for the two conditions. The place of articulation interacted significantly with voicing both in quiet and in noise, however results obtained suggest that performance in noise did not simply reflect the performance in quiet.