¹ In his Vergleichende Lautlehre des austronesischen Wortschatzes, Zeitschr. f. Eing.- Spr., Beihefte 15 (1934), 17 (1937), 19 (1938). (These will be referred to as Dempwolff 1, 2, and 3 respectively.) For a fuller reference and an account of the equivalences between our transcriptions, see Dyen, The Malayo-Polynesian word for ‘two’, Lg. 23.50 fn. 1 (1947). For other differences, see Dyen, Proto-Malayo-Polynesian *Z, Lg. 27.534-40, esp. fn. 3 (1951).

² Tg. bigqás ‘polished rice’ beside bigás is probably a product of contamination with digqás 'third polishing (of rice)'; note Hiligaynon Bisayan bugás 'hulled rice', digqás 'pound, grind well's. The contamination may have been favored first by the near homonymy of bigás and digqás resulting from the change of *e to Tg. i (but to HIBs. u), and second by the loss of the postconsonantal q in some Tagalog dialects, which resulted in digás; thus bigqás could have arisen at the digqás-digás isogloss.

³ I now use Merina instead of, though equivalent to, Dempwolff's Hova, thus following the practice of French scholars and of Otto Chr. Dahl, Malgache et maanjan (Oslo, 1951). For a few words on this point, see his p. 8.

⁴ Otherwise the distinction between borrowed words and inherited words may not be discoverable without resort to earlier records, since morph-alternation in most of the MP languages so far systematically compared works in such a way that loanwords can easily acquire the alternations of inherited words.

⁵ For Malagasy. From Dahl's statement (Malgache et maanjan 6) that the individuals speaking different dialects of Malagasy always succeed in understanding one another, we can conclude that Malagasy is a single language. The term Malagasy can therefore be used in general statements, though in specific citations the name of the dialect will be given (Merina, Sakalava).

⁶ See Dempwolff 2.89, where this reflex is treated as one of the ‘unerklärte Ausnahmen’.

⁷ There is no satisfactory explanation of the final e.

⁸ We expect a final e from *e.

⁹ Mer. viha with the same meaning has an inexplicable h.

¹⁰ Colloquial Malay has taroq ‘put’ (with inexplicable q) and taroh ‘wager’.

¹¹ NgD taroh ‘stake (in betting)‘ is taken to be a loanword, perhaps from Malay.

¹² NgD mă-rabut ‘tear off’ is taken to be a loanword, probably from Malay.

¹³ When the Tagalic languages (Tagalog, Bisayan, Bikol) suggest the presence of *q next to a consonant, I follow the practice of inserting it in the position indicated by Tagalog. No doubts are raised here by Bs., Bk. bágqaŋ ‘molar’. The comparison of Bs. búgqat ‘heavy’ with *beR₂qat causes no difficulty, but Bk. gobát ‘heavy’ does—because (besides the metathesis) it has no q. It may be that the Bikol word shows the effects of dialect mixture.

¹⁴ See fn. 5.

¹⁵ The explanation of the o is not clear.

¹⁶ The appearance of dz is the result of analogic change, Dempwolff 2.94.

¹⁷ Cf. Dempwolff 2.45 ff., particularly 51 ff. To apply the theory expressed in this paragraph one must divide the NgD words cited by Dempwolff into three groups: (1) those that contain a feature of the ‘older stratum’, (2) those that contain a feature which by Dempwolff's formulas contradicts a feature of the 'older stratum'—hence a feature of the 'younger stratum', and (3) those that contain no distinguishing feature. The last group is assigned by Dempwolff to the ‘younger stratum’, but its ambiguous position should be recognized. It then turns out that the vast majority of the words in (2) have Malay equivalents which they closely resemble in shape and meaning. There is a residue whose explanation is not obvious; but in the present state of our knowledge of Ngaju-Dayak and of its contacts with other languages, this is not surprising. It follows that Aichele's theory, as quoted by Dempwolff—'dass dort [in Borneo] eine Literatur-Sprache existiert haben muss, ähnlich wie das Kawi auf Java'—is otiose as a way of explaining the 'older stratum'.

¹⁸ If this reconstruction is successful, it is our only example of *-ew. The reconstruction is an attempt to combine Dempwolff's *buRaw (based on TBt. buro ‘pursue’, taken with the Tg. and NgD words) and *buru (based on TBt. buru ‘chase’, NgD kă-buro ‘chased away’, Saa huru ‘run’, taken with the Ml. and Jv. words). In reaching my reconstruction I take TBt. buru as probably a loanword and NgD -buro as a Malay loanword (leaving the final o unexplained) which has acquired a NgD prefix. The reconstruction *-ew is not interdependent with that of *R ₃; the latter depends on the assumption that Jv. buru is inherited and not a loan from Malay. In either case, however, we can reach the reconstruction of *-ew. The reflexes assigned to *-ew (Tg. -aw, TBt. -o (?), Jv. -u, Ml. -u, NgD -aw) are parallel to those assigned to *-ey (Tg. -ay, TBt. -e, Jv. i, Ml. i, NgD -ey, Mer. -i); e.g. *qatey, Tg. qatáy, TBt. ate-ate, Jv. ati. Ml. (h)ati ‘liver’, NgD atey ‘feelings’, Mer. ati ‘liver’. Now Saa has -e from *-ey, as in s-ae ‘liver, feelings’, and we should perhaps expect Saa -o from *-ew rather than the -u of huru; but the discrepancy is explained as assimilation. The interesting part of this Saa word is its r; for Saa l reflects *R ₁ (*paR ₁i, Saa heli 'ray') and *R ₂ (*uR ₂at, Saa ule-ule 'tendon, vein'). If it is true that Dempwolff's two reconstructions above can be replaced by one, and if the association of Saa huru is correct, we have further evidence of a distinction other than *R₁ and *R₂.

¹⁹ This is a combination of Dempwolff's *kaRus and *qaRus; the initial reflexes are dealt with in my forthcoming monograph on the PMP laryngeals. Jv. òs-òs ‘hiss’, cited by Dempwolff under his *haRus, is discarded because of its deviant meaning. NgD hărus-an (as well as its by-form with metathesis, rahus-an) ‘river-bed’ is taken to be or contain a loanword, perhaps from Malay.

²⁰ There is no satisfactory explanation of Ml. e from *i.

²¹ Cf. Mer. tavéz-ina ‘to be made savory’ beside tavi ‘fat, savory’ (from *(tT)abe(q0), Tg. tabáq, TBt. tabo ‘fat, savory‘), in which the z must be analogical in origin.

²² Malgache et maanjan 59.

²³ Cited directly from S. Elbert, Trukese-English and English-Trukese dictionary (1947). Otherwise Trukese words are cited in the same way as in Dyen, On the history of the Trukese vowels, Lg. 25.420-36 (1949).

²⁴ If Saa r reflects *R ₃ (see fn. 18 end), then Saa ahala- ‘shoulder’ indicates *R ₂ in the PMP reconstruction.