According to McLennan and Luce [1], variability in talker identity affects spoken word recognition when processing is slow and effortful. In the present study, we tested this hypothesis by manipulating the neighbourhood density of target words in a repetition priming experiment. Both for words with few and many phonological neighbours, the amount of priming for repeated words was not affected by a voice change. Such observation supports the claim that abstract representations exist and underlie spoken word recognition.