The Dravidian tree

Dravidian languages are now concentrated in the south, but they were once more widespread. Brahui in Balochistan and Kurukh and Malto in the eastern hills lie far to the north of the main block, and Dravidian place-name elements and loanwords in Marathi and Gujarati suggest that Indo-Aryan languages spread over areas where Dravidian had been spoken.

Two broad views compete. On the first, Proto-Dravidian was spoken in the northwest of the subcontinent, possibly by some of the people of the Indus civilisation, and its speakers moved south and east as Indo-Aryan spread. Brahui would then be a relic left behind. On the second, the family grew up in the peninsula itself. Franklin Southworth and the archaeobotanist Dorian Fuller compared the words for plants and farming that can be reconstructed for Proto-Dravidian with the crops found on excavated sites. The reconstructed vocabulary includes native plants such as horsegram and mung bean and fits the Southern Neolithic of the Deccan, around 3000–2000 BCE, better than it fits the wheat-and-barley farming of the Indus. On this view, Brahui reached Balochistan by a later migration.

Genetics has not settled the matter. Large studies of ancient and modern DNA (Narasimhan et al. 2019) show that South Asians descend from a mixture of ancient South Asian hunter-gatherers and people related to early farmers of the Iranian plateau, and some of the authors suggest that this mixed population, present in and around the Indus sites, may have spread Dravidian. Genes do not carry languages with them, however, and the suggestion remains a hypothesis.

Dating is equally uncertain. Estimates from shared vocabulary put the break-up of Proto-Dravidian at roughly 2500–2000 BCE, but these methods carry wide margins of error.

Dravidian languages look very different from one another in their scripts and sound quite different in their modern vocabularies, but their grammars share a strong common profile.

In sound, the most distinctive feature is the set of consonants made with the tip of the tongue curled back (retroflex), which contrast with dental ones: Tamil kaṇ 'eye' has a retroflex [ɳ], not an ordinary [n]. Proto-Dravidian had five vowels, each short or long, and did not distinguish voiced from voiceless stops at the start of words; the word-initial voiced stops of modern Telugu and Kannada come mostly from later changes and from loans. Tamil and Malayalam keep the rare approximant ḻ, as in the name Tamiḻ itself.

In grammar, words are built by adding suffixes one after another, never prefixes. The verb comes at the end of the sentence, and relations are marked by postpositions and case endings. There are no relative pronouns: a verb form placed before a noun does the work of an English relative clause, as in Tamil nāṉ paṭitta puttakam 'the book I read', literally 'I read-PAST book'. Nouns are divided by meaning into humans and everything else. Tamil distinguishes avaṉ 'he', avaḷ 'she' and atu 'it', whereas Telugu uses one word, adi, for both 'she' and 'it'. Most languages have two words for 'we', one including the listener and one excluding them: Tamil nām against nāṅkaḷ.

The four large literary languages are written in scripts descended from Brahmi. Each has a large stock of Sanskrit loanwords, and Tamil and Telugu show a marked gap between the formal written language and everyday speech. Most of the smaller languages have no established script of their own; Brahui is written in a form of Arabic script.

The internal shape of the family is broadly agreed, but the details are not. The main uncertainty concerns South-Central Dravidian, the branch of Telugu and Gondi. Krishnamurti grouped it with South Dravidian I on the basis of shared sound changes, but some of the evidence can also be read as contact between neighbours, and others have grouped it with Central Dravidian or treated it as intermediate. Whether Central Dravidian itself is a true subgroup, and whether Brahui belongs in the same branch as Kurukh and Malto or split away separately, are also debated.

The biggest external question is the Indus civilisation. Asko Parpola and Iravatham Mahadevan have long argued that its inscriptions record an early Dravidian language, pointing, for example, to the fish sign, since Dravidian mīn means both 'fish' and 'star'. Other scholars doubt that the sign system could be read at all, and Steve Farmer, Richard Sproat and Michael Witzel (2004) argued that it was not a full writing system. Without a decipherment the question cannot be settled.

Various proposals connect Dravidian to families outside India. The Elamo-Dravidian hypothesis of David McAlpin links it to the extinct Elamite language of southwestern Iran, and Nostratic theorists include it in a macro-family with Indo-European, Uralic and others. Caldwell himself noticed resemblances with Uralic. None of these has won general acceptance, and Susumu Ōno's proposal that Japanese descends from Tamil is widely rejected.

Finally, scholars disagree about how much Dravidian shaped the Indo-Aryan languages. Their retroflex consonants, their placement of verbs at the end of the sentence and their participle constructions are often attributed to Dravidian influence. How many early Sanskrit words are Dravidian loans is still argued; Michael Witzel, for one, assigns many of them to other, unknown languages.