316 THE POLYNESIAN WANDERINGS. 



(b) ui, uui (uwi), the yam. 



Samoa, Tonga, Futuna, Niue, Uvea, Aniwa, Sikayana : ufi. Vat6 : 

 uf. Maori: uwhi, uhi. Nukuoro, Moiki, Tahiti, Hawaii, 

 Mangareva, Rapanui, Paumotu: uhi. Marquesas: buauhi. 

 Mangaia, Rarotonga: ui. Fotuna: aufi. 



Viti: uvi. Rotuma: uk. 



British New Guinea, Murua, Kiriwina, Galavi, Boniki: kuvi, yam. 

 Nggela: kuikuvi, a cover, to cover; kuvihia, to cover; kuikuvi, 

 a leaf used as an umbrella. Mukawa: kubi, yam. Nada: 

 kuva, id. Fagani: uvi, id. Pokau: veu, id. Wango, Saa: 

 uhi, id. Motu: uhe, the end of the yam, which is kept for 

 planting. Aneityum: n-uh, yam. Tanna: n-uk, id. New 

 Caledonia: ubi, id. Vanikoro: upie, id. Baki: yubi, id. 

 Eromanga: n-up, id. Baravon: up, id. Makura: na-u, id. 

 New Ireland (Carteret Harbor): u, id. Pala: uh, id. 



Malay: ubi, yam. Kayan: uvi, id. Java, Kisa: uwi, id. 



Malagasy: ubi, ovy, id. 



Arabic: "ayab, roots (so called because buried in the ground). 



The cause of this collocation is the satisfaction of Macdonald's Semitic 

 theory in his naif interpretation that roots are things buried in the ground. 



A. 



It is highly problematical whether there is any identity between afa ki and 

 ufi. The Polynesian and the Viti make no reference to burial ; the dominant 

 sense is to cover by setting the covering agent upon the object to be con- 

 cealed. It is not until we reach the Malagasy, and a secondary form at 

 that, that we find any recognition of the burial sense. In general I am 

 opposed to the acceptance of any Malagasy identification where there is 

 any but the closest resemblance of form and meaning, except where we find 

 it upheld by a satisfactory suite of transition forms in Indonesia. The 

 Arabic is quite as imperfectly correlated with Polynesian ufi as is Efat6 

 afa-ki, in fact more so, for we should need evidence as to the abrasion of 

 so strong a palatal as ". Furthermore the Proto-Samoan stem is uf it. Its 

 only Melanesian appearance is in Nggela kuvihia, which bespeaks a kuvis 

 stem, not wholly to be accepted as to ^-accretion and t-s mutation in the 

 absence of transition forms, although we have some evidence tending to 

 establish this mutation. 



B. 



The absence of the inner labial spirant in the Efate forms finds a true 

 Polynesian parallel in Mangaia and Rarotonga; in Melanesia it is found 

 only in the most degenerate forms, in which naught survives of the word 

 but its initial vowel. 



The Polynesian life-history of the word is plain ; it is written about the 

 successive mutations of the central consonant, /-z>-/*-extinction. 



In Melanesia we shall examine the identifications by the same criterion. 

 The form kuvi is anomalous in the frontal fe-accretion and it is loosely 

 accredited to a territory too wide and heterogeneous to admit of more pre- 

 cise identification. The ufi form is absent from this area. The first stage 



