POLYNESIAN AND MALAYAN. 137 
do not appear in Polynesia. Indonesia shows but one such mutant, 
f—p, spirant to mute and both surd, in 48, 49, 51, and 52. In Melanesia 
this mutation is found in 13 and 34; but there are other labial mutants, 
f_v, surd spirant to sonant spirant, in 24, 25, 30, and 39; f—-w, spirant to 
semivowel proximate to the labial tract, in 9, 10, 26, 27, 28, and 29; to 
extinction along this channel in 11, 12, and 31. 
In our next group of mutations we find the result in the aspirate. 
At this point I must renew attention upon the fact that speech has 
three aspirates, one proximate to each of the three tracts of speech 
organs. Itisa breathing always, almost formless, not dependent upon 
palate, tongue or lips for its production, therefore not to be set in pala- 
tal, lingual or labial series but in close juxtaposition thereto. We find 
mutation to an aspiration near the labial, f—h, in Polynesia 2, all Tonga- 
fiti languages except Nukuoro, a secondary Samoan, and Aniwa, best 
regarded as secondary to some undetermined language of Nuclear Poly- 
nesia; in 5, one of the islands of the Western Verge, I hope to show that 
this aspiration is not labial. ‘Through this channel we find the extinc- 
tion of the second consonant in 4, both Tongafiti languages. Now let 
us examine 3, the niicho of Sikaiana, an island of the Western Verge, 
and compare with it micht of Bouton in the Celebes subprovince of 
Malaysia. This tch isa lingual, therefore not to be considered a muta- 
tion product from f labial, for such mutation extra seriem is not to be 
considered when another explanation is possible. In Subanu ugzsz we 
have another lingual, and it is at least interesting that the initial conso- 
nant ng also occurs in Nuguria (5), a near neighbor of Sikaiana. It is 
true that nitcho-ngiho differ in the final vowel from ngisz, but that 
amounts to little since in 41 and 42 we have nzfy and uifin, undoubted 
congeners of mifo. We are justified in the conclusion that there were 
two primal stems nzfo(z) and ngisi(o); the fact that we have one in 
Malagasy and Chamorro, the other in Subanu, shows that they were at 
least of equal currency in the earliest period. ‘The forms with s are 
found in 46 and 47 in Indonesia, in 14,15,21,and 35 in Melanesia, andin 
Polynesia are absent. ‘he mutation s—tch is well established in the 
labial series; we find it here in 45, and that should be sufficient to set 
nitcho of Sikaiana as an S—derivative. The most common mutation of 
the sibilant is the weakening to the juxtaposed aspiration, s-h. Begin- 
ning in the Indonesian region, where we have first found the vg7s7z stem, 
we identify s—h mutants in 43 and 44, which are in the neighborhood of 
the Subanu, therefore archetypal in this region. In Melanesia we shall 
find the geography of the s and the h forms instructive. We find ngise 
in Pala, mise in Kabadi and uzsan in Nok6n, all in the gateway through 
the Bismarck Archipelago where the Samoa Stream made exit into the 
Pacific. Next we find uzho (7) in Ulawa, Saa, Bululaha, Buka, all just 
at the portal in the Solomon Islands, and lzho (23) in Buka, Ugi and 
Bougainville, and rzho at Wango, in the same chain of islands along the 
