The voice prefix di- in Malay/Indonesian continues to be a source of vigorous debate, most of which focusses on its function. In this brief paper I want to look at another aspect of this intriguing prefix, namely its origin. Specifically I will deal with the three hypotheses that have been put forward in this regard. These can be summarized as follows:

1. Di- is a reduced form of the free pronoun dia (Van Ophuysen).
2. Di- is a denasalised reflex of an earlier voice marker ni- (De Casparis, Teeuw).
3. Di- is originally a preposition that became grammaticalised as a voice prefix (Adelaar, Ross?).

I will look at the arguments presented for each case, focussing on the last two. Most of the supportive material for the 'grammaticalised preposition' hypothesis has come from closely related languages/isolects where the function of di- is markedly different from that in Standard Indonesian/Malay. I will argue that position 3 is difficult to maintain in the face of typological evidence, and that the divergent uses of di- in those languages/isolects can be accounted for in other ways. The easiest solution is that Malay di- is a direct reflex of the PMP voice marker *ni-/-in-.