Notes :This extremely interesting root deserves a detailed commentary. The standard Dravidian root for ear is *kev- (DEDR 1977, see PGn *kev-). The PK form for 'ear' thus stands isolated, the only obvious decent comparison being Tulu ceraḍu 'ear', which is, moreover, dialectal and partly irregular (palatalisation *k- > c- in Tulu is a rare and uncertain thing). In [Burrow 1944] both roots were compared with the Uralic forms for 'ear', especially Finnish korva, etc., and it was concluded that both date back to PDR *kervi. This, however, seems highly unprobable. Loss of -r- everywhere but in Kui-Kuwi is an extremely strange and unique phenomenon, especially considering that not even a single trace of it was preserved in neither of the SDR languages (where we could at least expect some sort of gemination, like Tamil *cevvi instead of cevi). Much more probable is the connection of the PK form with the Dravidian words for 'ear-wax' (also discussed by T. Burrow in [Burrow 1943]). The word should be reconstructed as *kiru-pīl, 'ear-excrements', and it has undergone radical transformations in most languages, the most archaic form being Tulu kirmbilu. In most other languages the first syllable vowel was assimilated to the inlaut combination of labialized sounds (Tamil kur_umpi, PG *koṛ-ŋel, etc.); in Gondi we have the following development: *kirupīl > *korupīl> *koṛmbīl > *koṛŋīl > *koṛŋel. The general situation is thus as follows. In Early Dravidian both forms (*kir- and *kev-) were co-existent, although the difference between them is unclear. Later the form *kir- was in most dialects completely replaced by the form *kev- and survived as an archaism only in PK (and in Tulu?). The original derivative *kirupīl survived, but in most dialects it was modified beyond recognition, especially since the origins of its derivation became lost.
Notes :The root is almost never used by itself, but always in conjunction with certain parts of body: *kir_l- + 'leg' = ankle, etc. The correspondences are extremely complex, since it is very hard to trace PK -rl-, PPM -eir- and Gondi -r- to a single proto-cluster. However, if we assume this reconstruction, it would be quite possible to compare this form with SDR *kir_- 'small' (DEDR 1594), especially since there is an appropriate parallel in Toda kɨt_ eɫ̣f 'ankle', literally 'small bone'. If we consider *-l- a suffix, the forms are easily explainable.
Notes :Glottal stop in PK almost certainly reflects some cluster; unfortunately, lack of external parallels prevents us from determining its exact character.