34 SISSANO. 
Galavi keina, and Boniki kaena. It is possible that this will be held 
sufficient warrant for the inclusion of these forms in conjunction with 
this ai stem. 
When modulation is applied to the ai stem we find a prefixed palatal 
mute, surd in kai, sonant in gai, the difference not diagnostic but 
resting upon the habit of the particular speech. ‘The variety of hai 
(8) is controlled by local conditions. Codrington establishes the rule 
that the g of Nggela speech becomes h in the passage to Vaturanga, 
close neighbors in the Solomons. We lack this statement for his un- 
particularized language of New Georgia, but as it is within the same 
circle of speech the same rule probably holds. Accordingly we regard 
hai as a variant of the gai series. As to 12 we make this note: Cod- 
rington records gair as from Murray Island, but in Ray’s more recént 
and far more complete study of the speech of the languages of the 
islands of Torres Straits nothing in the least resembling this is recorded 
either in Miriam or Mabuiag, the two languages which cover what- 
ever region might appear as Murray Island in any source from which 
Codrington might draw. ‘The Arag gae (11) is scarcely distinguish- 
able from the gai type and occurs in the close vicinity thereof. 
From 13 to 23 we encounter a perplexing variant; yet because it is 
included within regions of the predominance of the kai type we must 
regard it as associated with that stem. It appears to be a kai mutant 
with one or more prefixes of a nature which we can not yet identify 
in sense value. In examining the mutants of the stem we need pay 
little attention to the passage g-ng and k-nk, for these are common 
in the Pacific. As variants set upon the gai stem we note regai (13) 
and tangae (17) immediately associable with gae (11). As variant 
set upon the kai stem, with the same mutation to kei as appears in 
ei from ai (3, 5), we note tankei (18). From 13 and 17 depend in 
succession 14-16; from 18 depend 19-22. In 13-15 we find the prefix 
ra (re), in 17-19 the prefix ta (te); in 20-21 we have both prefixes, for 
the mutation te—ke (21) is normal in these languages. In 22 in a group 
of closely allied village dialects te has become e through a habit of 
dropping t which characterizes these languages. In 16 we seem to 
have a duplication of the gai stem gei, after the manner of kei (18) 
and a degradation form ga as in 14-15. In 23 we leave this partic- 
ular area and are at some distance in the Solomon Islands, yet the 
relation with the type 17-22 is unmistakable; we regard da as the 
variant of ta (17-18), and go as a further deterioration of gai through 
ga (14-16). 
In 24, 26-28 we find demolition forms of kai-gai with which we have 
familiarized ourselves in the next preceding paragraph. ‘The go of 
25 may be, as we have found needful in 23, a residual of gai; but by 
reason of its occurence in a region of the predominance of the kau 
stem (35-43) it will be preferable to assign this to that source. 
