288 THE POLYNESIAN WANDERINGS. 



We have already seen that the Tongafiti folk have found it necessary 

 to particularize smoke as fire-vapor. The same usage is found, though far 

 less commonly, in Melanesia. We have in Moanus kasumoan, in which 

 moan is fire although by no means of our ahi-fire stem ; Lakon ahau av and 

 Lo hiev are clearly homogenetic with auahi; but these are all. 



The stem asu we find intact in a considerable group of languages in the 

 New Hebrides and two in the southern Solomons. This stem is altered by 

 frontal accretion of three palatals in Moanus, Nifilole, and Nggao, severally 

 too remote to be regarded as local interinfluence ; of two linguals in Ulawa- 

 Bululaha and Alite, which only once more in all our material, sa (337) bad 

 Tanna ra, are found interchanging; of one labial. I note Codrington's 

 opinion that "Nggao ngganggahu is the Mota gagavu thick, clouded." But 

 is it ? Moanus kasu is nearer phonologically and far nearer in the migration 

 track, for a derivation from Mota to the Solomons is an upstream movement 

 and against the current. I do not incline to regard this frontal accretion 

 as a Melanesian device; it seems more reasonable to look upon it as the 

 remnant of an earlier stem form, but what that earlier and now vanished 

 radical initial may have been we may not seek to know. 



Savo shows asu with the slight change from surd to sonant. Two lan- 

 guages in the Solomons and one in the New Hebrides have asu with equally 

 slight change from sibilant to aspirate. Motlav, Volow and Arag, all in 

 the New Hebrides, carry the change a small step farther by the alteration 

 of the final vowel. A step beyond in the change of the final vowel brings 

 us to Lakon. Efate and Sesake have the asu stem unchanged save for the 

 terminal accretion of the vowel a. 



Thus far we have kept close to the primal form. Now, while holding 

 fast to the s, we are to find a greater vowel change in Malekula Pangkumu 

 ese and one which sacrifices the differentiation of the two vowels which has 

 hitherto been found to persist. To this branch of the stem belongs the 

 Vuras es. With it I have included Ambrym walehi provisionally ; the ehi 

 can not be a mutant of afi fire, for that is av in Ambrym ; I have thought 

 it possible that ehi is a further mutant upon ese, and wal is as yet unac- 

 countable. Santo osun with a terminal accretion introduces the o-branch 

 which will be noticed in Rotuma as well and in three of the Vanua Lava 

 languages with terminal abrasion. Norbarbar shows in one of its forms a 

 terminal abrasion deriving from either ahu or alio, but probably the former. 

 Lo suggests in hi a frontal abrasion, and that principle may also account 

 for a part of the second Norbarbar form suio. One Buka form, uruhu, quite 

 plainly suggests in uhu another vowel change of asu, leaving ur unaccount- 

 able yet somewhat smacking of wal in Ambrym walehi, or perhaps in these 

 words ruhu and lehi are akin to Alite rasu ; the second Buka form is mani- 

 festly a devolution form from the former. 



In the Malay asap we recur to the definitive compound of the Tongafiti, 

 as-ap, in which the latter component is again afi fire. The Malagasy seems 

 like an asu derivative. 



In the Semitic there is seen the initial palatal which appears here in 

 Melanesian, the former vowel is the same, the s is present, and the final n 

 is found once at least in Melanesia. So far as this one stem is concerned 

 the resemblance is striking. 



