Vedic Sanskrit has a number of linguistic features which are alien to most other Indo-European languages. Prominent examples include: phonologically, the introduction of retroflexes, which alternate with dentals, and morphologically, the formation of gerunds. Some philologists attribute such features, as well as the presence of non-Indo-European vocabulary, to a local substratum of languages encountered by Indo-Aryan peoples in Central Asia (Bactria-Marghiana) and within the Indian subcontinent during Indo-Aryan migrations, including the Dravidian languages.
Scholars have claimed to identify a substantial body of loanwords in the earliest Indian texts, including evidence of Non-Indo-Aryan elements (such as -s- following -u- in Rigvedic busa). While some postulated loanwords are from Dravidian, and other forms are traceable to Munda or Proto-Burushaski, the bulk have no proven basis in any of the known families, suggesting a source in one or more lost languages. The discovery that some words taken to be loans from one of these lost sources had also been preserved in the earliest Old Avestan texts, and also in Tocharian, convinced Michael Witzel and Alexander Lubotsky that the source lay in Central Asia and could be associated with the Bactria–Margiana Archaeological Complex (BMAC). Another lost language is that of the Indus Valley civilization, which Witzel initially labelled Para-Munda, but later the Kubhā-Vipāś substrate.
