62 SISSANO. 
find advantage in discussing the variants of the modifying element u 
or other.) The most frequent added member is u. In 9-11 we find 
from widely dissociated languages a preface to this u, semivocalic in 10 
ni-yu, vocalic of the same series in 9 ne-iu, aspirated in 11 ni-hu; 
this forms a readily comprehensible series of coefficient-functioning 
modulants of secondary stem u; in 13 ni-jog we may have an equiva- 
lent of the yu of ni-yu, but in the lack of certainty of the phonetic 
system employed in the writing of Jibu in roman characters we are 
doubtful, but if j be used in its English double consonantal value we 
have an inexplicable anomaly; in 12 ni-wi we find no difficulty in 
tracing the modification of u into its proximate semivowel w before 
an additional assumption of i. This assumed 1 may be found in 25-26, 
there assumed by a variant of stem nu. 
We next consider a group of assumed final consonants with the u 
stem. In 13 ni—jo—g we find a palatal mute, in 21 ngup we are dealing 
not with u modifier but with nu stem and therefore the labial mute 
p is to be omitted from this series; in 27-30 we find a group of inter- 
related forms included within a very frequent mutation type, the 
liquids r in 27-28 and 1 in 29, and the proximate nasal n in 30. In 
31-34 and 36 we encounter a most interesting group and probably 
not so easy of resolution as appears on superficial examination. I had 
at first regarded these forms as the compacted ni-u plus an assumed 
ra. It is quite possible that in this set of forms we are dealing with a 
vocalic variant of stem nu with the addition of a modifier ra. ‘This 
is clear in 34 ru-ra, where we encounter the frequent n-r mutation. 
In 36 nuia, which is distinctly in series with 31-34, if we regard a as 
devolution from ra we shall be left with a compaction nui instead of 
niu. This can scarcely be admitted as correct metathesis. It seems 
more satisfactory to regard the compaction as nu-ia, in which is 
readily recognizable ia as mutant upon ra, as to which we note that 
the interchanges of vowel and liquid have been set forth on pages 
35 and 50. ‘This assumed ra becomes by abrasion of the final vowel 
the parent of the r-I-n forms in 27-30. In 24 ne-k we find a palatal 
mute assumed by a variant of ni stem. Vella Lavella 35 néri is not 
to be associated with the wide variation in this vocable; the intrusion 
of r is prohibitory of ascription to niu; ne might seem a ni derivative 
as in several instances in this series; ru might be derived from the nu 
stem; but to interpret the vocable as conjoint of these two elements 
would be to regard a primitive stem as modifying itself, a process 
wholly foreign to the genius of these languages. Accordingly we must 
rule neru outof the identifications. “The stem nis remarkably constant; 
we find its mutation to liquids in 14 liu and 34 rura, and to labial 
mute d in 32 diura. In the Polynesian system we note in Mangaia 
the preservation of nu which in our Melanesian material we find of 
considerable frequency. It is found in 19-21, 27-34, 36. A group 
