Notes :The two "variants" - *aṇai and *anḍai - in reality cannot be joined together as easily as Burrow & Emeneau propose it in DEDR. In order to trace them back to the same source we would have to suppose that both go back to an early, and presumably, verbal root like *aṇ-, and even then the two lines of derivation would be rather vague. I would deem it more reasonable to reconstruct a PSDR nominal stem *aṇai "embankment". Tamil and Malayalam aṇṭa(i), with an obviously narrower meaning ('field bund', but not 'dam' in general) are a different, more local, formation, maybe contaminated with *anḍ- ' to approach, join' (?). As for Kodagu aṇḍa, in this case the retroflex cluster may actually go back to a simple retroflex nasal, i. e. reflect the earlier *aṇai; cf. a similar case in PSDR *pUṇai 'bamboo'. (It is possible that *-ṇ- > -ṇḍ- is a regular development in Kodagu nominal stems in the position before a strong suffix; unfortunately, the data are very scarce).