354 THE POLYNESIAN WANDERINGS. 



component of various tree names. It is in Yiti first that we find it in 

 free existence. In Melanesia this form is rare. It occurs as kau in Efate, 

 Sesake, Epi, Nguna, and perhaps may be preserved in Aneityum ; as gau in 

 Marina ; as au in Motu and somewhere in the Solomon Islands. 



The triplicity of the Efate forms suggests a possible transition. Kasu 

 and kas are easy to be correlated, kasu and kau less easy. They might be 

 linked by the assumption of a parent form kahu, from which each might 

 derive. This would appear in modern Samoan as kau; but I have found it 

 the rule that even the mildest aspirate in Proto-Samoan becoming extinct 

 in modern Samoan is yet retained as aspiration in Nuclear Polynesia and 

 as th in Viti, none of which mutations is found on this record. 



With the statement of this contrary argument, let us adopt this as a 

 working hypothesis. With kasu we identify Nggao gazu and Malagasy hazu. 

 With a parent kahu we may associate Malay and Baju kayu. This is as 

 far as this unsupported hypothesis will carry us, although kayu invites 

 enticingly as a bridge to kau. 



We next employ the comparison of Polynesian lakau and Bierian lakai to 

 give us a transformation phase by which we may cross from kau to kai, and 

 in Viti we find kai occurring as a dialectic variant of kau. This secondary 

 stem kai (as kai, cai, gai, gae) is discoverable in the free state or in com- 

 position over a wide Melanesian range, and in Teor in Indonesia. In Vatu- 

 ranga and New Georgia hai we find the transition to Malekula n-ai and to 

 at of Ulawa, Bululaha, and Alite, probably including Duke of York diwai 

 as a composite. 



For the next deterioration phase we have the benefitof a similar transition 

 member. As we were able to link lakau and lakai, so may we link lakai and 

 Lo raga, Lakon rega, and Bierian leke. Thus we obtain a tertiary stem ka, 

 ke. The ka (ga) form occurs only in Lakon and Lo. The ke (ge) form we 

 list from Norbarbar, Vuras, Mosin, Pak, Sasar, Alo Teqel, Malekula ; and 

 the gi from Tanna. 



Let us not neglect to observe that in many of these Melanesian languages 

 kau is able to stand free, as in Viti and Efate ; omitting the combined article 

 n variously vocalized, we list free forms in Nggao, Sesake, Epi, Marina, 

 Nguna, Motu, Aneityum, Solomon Islands, Omba, Arag, Nggela, Bugotu, 

 Murray Island, Tanna, Malekula, Ulawa, Bululaha, Alite, Vaturanga, and 

 New Georgia. 



In the Polynesian area we have noted that the prior composition element 

 is la, variously modified to ra, nga, a. We find this recurring in Melanesia. 

 First as la: Aneityum, ne-lcau-un. Then as e, corresponding to a in Tonga, 

 etc. ; Pak, Sasar, Alo Teqel, further establishing a variant stem nge out of 

 the tertiary ge. We now pass from la to ta, a mutation ill supported ; it 

 may be better to regard it as an independent member of the composite. 

 As ta we note its occurrence in Tangoan Santo, Mota, Merlav; as te, in 

 Volow, Motlav, Norbarbar. It may even be di in Duke of York diwai. 

 The form tenge illuminates Vuras retenge as a double composite, the com- 

 mon la stem prosthetic upon the te-nge compound ; and upon retenge hangs 

 Mosin rekenge. The prior element in Malo wu-cai is not elsewhere found. 

 Maewo geiga eludes analysis, except upon the scheme that it is a duplicate 

 of slightly variant kai derivatives. 



