SAWAIORI MATERIAL IN INDONESIA. 165 



L. 



Grouping 1 and r because of the frequently mentioned impracti- 

 cability of establishing a stem independence in our present data, 

 we note that in Polynesia the most frequent mutation is the extinc- 

 tion. This becomes most strongly marked in the Marquesas. It 

 is infrequent in Melanesia. In Indonesia it appears in Sulu and 

 Tagalog in the same word, talinga (350) ear, and in the Togean 

 Islands in pula (284) to shine, at least the \-y mutation noted in 

 that instance is a close approximation to extinction. 



The \-n mutation in Polynesia seems more a word quality, for it 

 appears in two languages each for data 100, 154, 312, 313; these 

 removed from the reckoning, we find it only five times, all in Nukuoro. 

 In Melanesia it is rare. In Indonesia it is found in pula (284) to 

 shine Timor junan the moon, twice in Wahai; these three words 

 are included in the list of those similarly affected in Polynesia. 



The long leap from lingual semivowel to mute of the same series 

 is found but once in Polynesia (\-t Mangareva 287) and once in Viti 

 (\-nd 287). In Melanesia it is scarcely more frequent. In Indo- 

 nesia \-t does not occur; \-d appears sporadically in several lan- 

 guages, one instance (350) containing the greater number of its 

 Melanesian occurrences as well. This mutation is most marked in 

 Malay (seven times) and Malagasy (five times). Of the Malay 

 instances but two (260, 334) are open to any doubt, of the Malagasy 

 but two (348, 350) are really convincing. The \-j mutation, really 

 a t- variation, is found but once in Indonesia, three times in two lan- 

 guages of the New Hebrides, and not at all in Polynesia. \-s is noted 

 in a single instance in Malay and very unsatisfactory. \-kl, noted 

 once in Malay and Java, is rather an accretion than a mutation. 



N. 



The variants of the three nasals follow two cleavage planes, the 

 vertical in series, the horizontal across all series. In the case of 

 ng we have observed as well ng-n and ng-k. In Polynesia n shows 

 this horizontal movement only backward, n-ng, and this only in 

 Moriori and once in Viti. In Melanesia n-ng is quite widely spread, 

 affecting several vocables and most of them repeated in several lan- 

 guages; here is found in one instance the forward n-m mutation. 

 In Indonesia the n-ng mutation occurs once each in Malay, Kisa, 

 and Tagalog, the Kisa instance doubtful. 



In Polynesia the only vertical mutation is upward in the series 

 to the semivowel, to / in a single word in two languages, to r in 

 another word in two other languages. This mutation in Melanesia 

 has a wide extent in the case of nifo (259) tooth, and is less widely 

 found in three other of our data. In Indonesia it is found once each 

 in Macassar, Sassac, and Wahai, the latter affecting the same vocable 

 as in Alite. 



