MELANESIAN ANNOTATIONS ON THE VOCABULARY. 5) 
The group of consimilars 29-31 is particularly difficult of adjust- 
ment. The three forms occur in two languages north and south of 
the extreme southeastern tip of New Guinea. The only forms with 
which a possible stem gi might be related are 26, found at a consider- 
able distance in the northern Bismarck Archipelago, and 27, still more 
remotely placed at the south of Melanesia. The prefix of this puta- 
tive stem eludes analysis. 
STEM au. 
Here we associate the forms 35-43. In 41-43 we have a distinct 
association restricted to two points at the southeastern tip of New 
Guinea and an island lying not far off that shore., This type suggests 
a kinship with the au stem, modulated by a prefix composition member 
of uncertain value. In the Tavara and Tubetube the difference is 
so slight as to be negligible, the appearance in Awalama of 1 in the 
place of a vowel recurs in several instances in that language and is 
paralleled by the same usage in Galavi, Boniki, and Mukawa (509), 
and probably in Taupota (31), these languages being found not widely 
separated on the same stretch of coast. 
Associated with this au stem we find the same consonantal modu- 
lants as with ai, namely, h, g,k. Motu and Keapara (36-37), closely 
allied languages, most probably derive their hau from the gau form, 
since that alone appears in that neighborhood. ‘The kau type is found 
only in the New Hebrides. Concerning the affix ubu (bu) (37, 39) 
we continue Ray’s note that probably it signifies the trunk, and we 
point out that in the Polynesian employment of this au stem it is 
almost wholly without ability to stand alone, requiring the preface 
of another member, which usually yields the form of the type la-kau. 
STEM asu. 
I present this stem in the nude form solely through inference, for 
in none of the material which we examine does it appear without the 
support of a consonantal prefix. If we admit hasie we have in 44-46 
the same series of consonantal modulants, h, g, k, as in ai and au. 
Codrington’s material from Wango affords no second instance of a 
transformation of final u to ie, but at the same time it is to note that 
his material affords only one instance where such a transformation 
might be possible, namely, that of asu. Yet since it lies on the track 
of migration toward communities in which the kasu stem appears 
unmistakably, we need have no hesitation in admitting this to the 
series, despite the vowel anomaly. ‘The forms 47-49 are less clear. 
They occur at widely separated spots on the south shore of Papua— 
languages so different in vocabulary that in Ray’s collection we find 
but two other words in which the slightest similarity may be discovered. 
Despite this objection there is sufficient resemblance to suggest a 
