478 THE POLYNESIAN WANDERINGS. 



The elaboration of this New Guinea material, the more particularly in 

 consideration of its important bearing upon the Viti Stream in the southern 

 gateway out of Indonesia, has made it advisable to extend and, when there 

 is reason therefor, to amend the notes in several cases which have found 

 their place in Appendix I. 



46. 



Until this Torres Straits material became available the line of partition 

 between kai and kani was quite distinctly the division between Polynesia 

 and Melanesia, for kani found its utmost eastern extension in Viti. Lack- 

 ing support from areas of pure Polynesian, the Viti kani could be regarded 

 only in the light of a Melanesian component of that mixed speech. But 

 on the New Guinea coast, at the very portal of exit from the Malay Archi- 

 pelago, we find the two stems in use side by side. Sariba exhibits the 

 Polynesian kai unaltered ; Suau, Mabuiag, and Dobu show the loss of the 

 initial k, that loss being normal in those languages. The supposition that 

 a medial n has been lost, which would be necessary to the argument of the 

 devolution of kai from kani, has been opposed by the fact that in the four 

 languages not only is n never lost, but it retains its proper value far more 

 consistently than in most of the neighbor tongues. We may, then, safely 

 regard kani as an ancient Polynesian stem which maintained its existence 

 to a point just within Nuclear Polynesia and then was dropped from the 

 synonymy. 



47- 



With this considerable record of two forms appearing side by side in 

 each of the Pacific areas, we need have no hesitation in assuming tahi to 

 be an ancient variant of the tehi radical. This slight phonetic change is 

 satisfactorily explained on the theory of the neutral vowel. The series 

 which we find in New Guinea should offer a satisfactory identification of 

 the Indonesian material, which in the earlier note I denied. Waima halt 

 is the tahi radical under 3214 metathesis. For the mutation of the aspirate 

 to the sibilant in Dobu and Kiriwina tasi there is abundant confirmation. 

 The mutation hd in Motu and Kabadi is normal to those languages, and 

 the kappation in the latter has become very familar in these studies. 



We now pass to a group of forms susceptible of less simple explanation, 

 that of which Misima tari is the type. Nowhere have we found evidence 

 in support of a mutation h-r, except in Sinaugoro with its similar s-/ muta- 

 tion. But we have some evidence in support of a t-r mutation, on the 

 nature of which refer to note 258. If, then, we are precluded from the 

 direct passage from tahi to tari we have no difficulty in finding a Motu 

 bridge which shall give us the series tahi-tadi-tari, each span of which may 

 be safely traversed. Thus we are carried over to the identification of the 

 tari of Misima and Panaieti, and to Sinaugoro tali. Thence to art is a safe 

 step, for t is normally dropped in Galoma, Hula, and Keapara. This shows 

 that the languages which use the forms tari-ari did not acquire the word 

 from the Polynesian wanderers through direct contact, but as a secondary 

 loan from the Motu-Kabadi folk. Thus having found the r in the h place 

 and the loss of stem t, and both within regions of Proto-Samoan influence, 

 we may safely accept the Indonesian identifications. 



