of the family. Thus in Karo in Sumatra, in Balinese next to Java, in Tontemboan in Celebes, etc., the word for “three” is tĕlú.
III. Old Javanese likewise has tĕlú; and how important Old Javanese is will be shown in § 6.
IV. Nias has no pĕpĕt; where other tongues have ĕ, Nias has an o. But this o has a peculiar pronunciation, it is articulated further back in the mouth than the o of a different origin. If I represent the front o by o1 and the back one by o2, I get (for example) the equations: Nias bo2li, “price” = Original IN, and likewise Gayo, Malay, etc., bĕli, but Nias o1no1, “child” = Original IN, and likewise Old Javanese, Tagalog, etc., anak. The peculiarity in the articulation of the o2 accordingly points to an originally peculiar sound, in fact to the pĕpĕt.
V. Iloko knows no pĕpĕt; where other languages have a pĕpĕt, Iloko puts an e. But the consonant, which immediately follows this e, is doubled ; thus the equivalent of Original IN, and likewise Old Javanese, Malay, etc., lĕpas, “free”, is Iloko leppás. This doubling of the consonant does not occur after an e of any other origin. Now Madurese says lĕppas; it also doubles the consonant, but leaves the ĕ unchanged. If we compare the Mad. procedure with the Iloko, it follows that the Iloko e, after which the consonant is doubled, points back to an original pĕpĕt.
VI. Talautese lacks the pĕpĕt; an a occurs where other languages have ĕ. But after this a the liquid l is articulated differently than it is after an a which descends from an Original IN a. So in Tal., too, we have an indication of the existence of the pĕpĕt in Original IN.
VII. Hova possesses no pĕpĕt; for an ĕ of other languages it puts in an accentuated syllable an e, in an unaccentuated one an i. Original IN, and likewise Karo, etc., tĕlĕn, “to swallow”, has therefore in Hova the equivalent télina. Now before this i < ĕ Hova preserves Original IN l unchanged, whereas before a Hova i < Original IN i it becomes d; thus tĕlina < Original IN tĕlĕn, but dimi, “five” < Original IN
