192 THE POLYNESIAN WANDERINGS. 



still remain impracticable in this instance for the distinct reason that we 

 are quite unable to adduce a single word in which all Polynesia has dropped 

 a medial consonant which all Melanesia has retained. 



Viti shows that a or i may be the final vowel. The z-forms are found in 

 Efate, Viti, Marina, Sesake, Belaga, Nggela, Nguna, Duke of York, Buka, 

 Kabadi; to which may be added an e-form in Ambrym. The a-forms 

 occur in Viti, Nggela, Mota. Terminal abrasion is found in Gog, Merlav, 

 Retan, Lo, Maewo, Duke of York, Nguna, Buka, Kabadi, New Britain, 

 Leon, Sasar, Vuras. 



The earlier vowel of the radical is found as a in Efate, Viti, Marina, 

 Sesake, Belaga, Gog, Merlav, Nggela, Mota, Retan, Lo, Maewo, Duke of 

 York, Nguna, Buka, Kabadi, New Britain. An e-form occurs in Ambrym, 

 where it is associated with final e, and in Leon, Sasar and Vuras. If indeed 

 it be the same radical an o-form is seen in Gog. 



The initial consonant k occurs in Efate, Viti, Marina, Sesake. It passes 

 from surd to sonant g in Sesake, Belaga, Gog, Merlav, Nggela, Mota, 

 Retan, Lo, Maewo, Leon, Sasar, Duke of York, Vuras, Ambrym. It goes 

 still higher in the palatal series to ng in Nguna, Ambrym and perhaps Gog. 

 It is abraded in Duke of York, Kabadi and New Britain. The second 

 consonant remains n without change in all these cases, save again the Gog 

 anomaly and the Mekeo. 



We therefore diagram this radical thus: 



I a I | 

 K > hi'e 



I e I 

 ng J J la 



The anomalies are now to be considered. With a vowel not elsewhere 

 found, with a change n-ng which the radical but once elsewhere exhibits, 

 and with the assumption of a final t which elsewhere does not appear, 

 Gog ngongot is not acceptable in this record. The only way by which it can 

 be included is to assume it to be a composite form, ngon-got, and this parting 

 of the palatal nasal would be wholly indefensible. The remaining anom- 

 alies, at the extreme north of the Samoa track, suggest an explanation by 

 a developmental series. Frontal abrasion is sufficiently well established in 

 the Melanesian handling of Polynesian loan words to admit Duke of York 

 ani to the kani radical. This admission will naturally carry with it Kabadi 

 ania (a the common verb isolating objective suffix), and by terminal 

 abrasion New Britain an. This an may be taken to admit New Britain 

 tan, and this in turn carries with it Buka iana. The other Buka forms 

 involve a further leap. Reverting once more to Duke of York ani, if we 

 can regard nanni as in some sort a reduplication (which it must be ac- 

 knowledged is informal) that would carry nan with it and for tuanan we 

 should have to postulate a composite of this nan. This is far too involved 

 to be satisfactory. 



Our Indonesian material is based on a radical m-k-n and the identifica- 

 tion should be overruled. The Malagasy is at variance with Indonesian 

 and Melanesian alike and should be rejected. It is impossible to see that 

 the Semitic has anything to do with the case. 



