A phoneme-detection task shows that listeners of Sepedi use semantic information in processing but not prosody (Experiment 1). Sepedi is a language with no grammaticalised prosodic expression of focus. Sepedi listeners detected phoneme targets faster when the phoneme-bearing words were focussed (as opposed to unfocussed) but not when occurring in a context conducive to prosodic emphasis (as opposed to non-conducive). Experiment 2 tested the role of semantic focus and prosody in processing by Sepedi L1/English L2 listeners (English being a language with systematic focus-to-accent mapping). Non-native listeners detected phoneme-bearing words faster in focussed condition (as opposed to unfocussed) and in accented condition (as opposed to deaccented). The results suggest that the L2 prosodic structure is exploited by Black South African English listeners even if this feature is not present in their L1. Our experiments replicate the pattern of results found in Akker and Cutler's (Biling Lang Cogn 6:81-96, 2003) experiment for Dutch L1/ English L2 listeners, even with listeners whose L1 does not use prosody the way English does.prosody. If we compare the mean values of early/late targets with/without prosody reported in Raborife et al. (2016) with the mean values reported in the present paper, the direction of the results is similar. In the current model, however, this interaction is not significant. This is due to the fact that the outlier diagnostic of the two models is different: unlike here, Raborife et al.'s model included data points above 1500 ms and RTs values higher than this threshold (15 data points) were mainly found for early targets with prosody, which may explain the significant interaction. In the current model, removing data points above such a threshold improved the fit of the model. 9 We are grateful to Anne Cutler for providing us with the original English audio material used in Akker & Cutler (2003).