Abstract: South Asia has for a long time been recognized as a major convergence area, and one of its defining features is retroflexion, i.e. a contrast between dental and retroflex consonant (an exception being the Northeast). The origin of retroflexion is most commonly attributed to unidirectional Dravidian substratum influence on prehistoric Indo-Aryan. I critically examine the arguments in favor of this explanation, against the background of the uneven chronological attestation of South Asian languages, the possibility of chance similarities, and the fact that Indo-Aryan retroflexion can be explained by internal developments. In these developments the presence of a voiceless retroflex sibilant ṣ plays a major role. The fact that Sanskrit/Indo-Aryan has, in fact, a triple sibilant contrast (ś : ṣ : s) points in the direction of a different account. This contrast is widespread in the northwestern transition zone between South and Central Asia, from a very early time. Sanskrit/Indo-Aryan, thus, can be assumed to have had that contrast in the Northwest before its speakers entered the peninsula. Subsequent changes of retroflex sibilant + dental stop introduced retroflex stops. This development is paralleled by prehistoric Dravidian changes of alveolar or retroflex liquids + dental stop yielding alveolar and retroflex stops, a fact that suggests convergence between Dravidian and Indo-Aryan. I conclude by arguing that these changes reflect bi- or multi-directional interactions, rather than unidirectional substratum influence of one language on the other(s). I suggest that the motivating factor is accommodation in the context of bi- or multilingual of speakers communicating with each other, and that this is the usual way ithat speakers interact in situations of extended bi- or multilingualism.