¹ The term ‘linguistic meaning-form’ is defined as a linguistic form together with its linguistic meaning. (The terms ‘linguistic form’ and ‘linguistic meaning’ are employed in this definition as defined by L. Bloomfield, Language 264.) The comparative method operates upon units consisting of forms with their meanings. It is impossible to consider the comparability of linguistic forms without simultaneously considering whether their meanings are similar. The relationship of a language known only through entirely undeciphered inscriptions (even if these are written in an already familiar script) cannot be demonstrated by the comparative method. The unit of comparison is taken to be the linguistic meaning-form rather than the morpheme with its meaning, because, under favorable conditions, elements larger than morphemes can be compared and reconstructed; it would be difficult to avoid concluding that the sequence *es-ti ‘is’ existed in Proto-Indo-European. The number of occurrences of such favorable conditions probably varies inversely with the time interval between the proto-language and the daughter languages. For this reason, perhaps, the reconstruction of morphemes with their meanings is more often achievable than that of larger elements when the interval between the reconstructed language and the daughter languages is as great as it is in the case of Proto-Indo-European and Proto-Malayo-Polynesian.

² Cf. Bloomfield, Language 314 f.

³ Works are referred to in this article according to the list given in Dyen, The Malayo-Polynesian word for ‘two’, Lg. 23.50–5 fn. 1 (1947). The same system of indicating the languages and the reconstructed phonemes is employed here as there, with the following additions and changes: (1) AMBt. = Angkola-Mandailing-Batak (words quoted from H. J. Eggink, Angkola- en Mandailing-Bataksch-Nederlandsch Woordenboek, Verh. v. h. Kon. Bat. Gen. v. Kunst. en Wet., Vol. 72, Pt. 5 [1936]), Bk. = Bikol, Bt. = Batak, KBt. = Karo-Batak (words quoted from M. Joustra, Karo-Bataksch Woordenboek [Leiden, 1907]), Fi. = Fiji, To. = Tonga, Fu. = Futuna, Sm. = Samoa; (2) in a reconstruction quoted from Dempwolff, h represents his ' and q his h in any position. A reconstruction not specifically attributed to Dempwolff and involving *q or *h rests on theories stated in a forthcoming treatment of the Proto-Malayo-Polynesian laryngeals. Parentheses in a reconstruction are equivalent to Dempwolff's brackets; they include two or more phonemes between which no choice can be made.

⁴ Since Jv. r is taken to indicate a borrowing from a Javanese dialect when it reflects PMP d (cf. Dyen, The Tagalog reflexes of Malayo-Polynesian D, Lg. 23.227–38, §2.2), it can be taken similarly when it reflects PMP Z.

⁵ Cf, Dyen, Tagalog reflexes (see fn. 4).

⁶ The appearance of Fi., To., Fu., Sm. e (instead of anticipated o) from PMP. e cannot yet be explained; cf. Dempvv. 2.134, 178.

⁷ The ? of this word is not expected in view of the statement of the Futuna reflex of *Z or its cluster with a preceding nasal in §4. Since the unexpected correspondent occurs in a single language and in only a single instance, the instance can be directly cognate only (1) if the word has undergone some secondary change not now statable or (2) if the formula is in error. Under the first condition *qunZan may have become *u?a which suffered metathesis to u?a. If ^? were taken to be a regular Futuna reflex, the instances of X (which are more numerous) would presumably have to be assigned to unstatable secondary changes. It is also possible that Fu. u?a was borrowed from a dialect or language in which *-nZ-became ?; the word would then be only secondarily cognate.

⁸ With assimilation of the expected initial r (cf. §4 and Dempw. 2.84) to the following l.

⁹ The Fiji and Samoan cognates reflect a cluster formed with a preceding nasal, cf. §4; for an explanation, cf. Dyen, On the history of the Trukese vowels, Lg. 25.420–36, §26.3 (1949). The theory is there inaccurately attributed to Dempwolff.

¹⁰ The symbol 0 means ‘nothing’; the parenthesis (h0) indicates that the meaning-forms compared do not allow of a distinction between the original presence of *h and the absence of any consonant (see fn. 3 end).

¹¹ For the appearance of -u-, see §13.

¹² Bk. rogo? ‘blood’ exhibits initial r by analogy. The explanation is essentially like that given for Tg. initial l instead of d from *D in Dyen, Tagalog reflexes 233.

^12a PMP R usually becomes Hv. z (Dempw. 2.81) but unaccountably disappears in some words (Dempw. 2.89), as here.

¹³ Dempwolff also compares Tg. ya?ó ‘journey’.

¹⁴ Cf. fn. 8.

¹⁵ Dempwolff's explanation (2.55) is that the appearance of j in both Ngaju-Dayak words is due to palatalization.

¹⁶ Cf. R. J. Wilkinson, A Malay-English dictionary (romanised) sub tujoh (Mytilene, Greece, 1936).

¹⁷ HlBs. turo? ‘point out with the finger’ (Phil. lang. 11.18) must be a loanword under this theory, perhaps from Tagalog. Bk. turo? ‘show, point, teach’ contains the regular Bikol reflex of *Z (without preceding *l).

¹⁸ It is probable that many of the words ascribed to the ‘younger speech-level’ of Ngaju-Dayak are actually loanwords from Malay. It is assumed that in such loanwords NgD. final k is the equivalent of Ml. final q.

¹⁹ Another instance would be colloquial Ml. taroq ‘put’ beside literary taroh ‘put, lay a bet’ (cf. colloquial Ml. taroh 'bet') if assigned to *taRuq, Tg. tá:go? 'conceal', TBt. taru, Jv. tòh ‘lay a bet’, NgD. taroh ‘bet’, Fi. tau ‘lay down’, Saa elu ‘lay, put’, To. tau?-aki ‘lay in the sun’.

²⁰ For evidence that a final h may sometimes have been lost in Malay by analogical wrong division, see my forthcoming treatment of the Proto-Malayo-Polynesian laryngeals.