¹ Unless there is an indication to the contrary, Tagalog is cited from A National Language-English vocabulary³ (published by the Institute of National Language, Republic of the Philippines, Manila, 1947), Malay from R. J. Wilkinson, A Malay-English dictionary (romanised) (Mytilene, 1932), and Javanese from Th. Pigeaud, Javaans-Nederlandsch handwoordenboek (Groningen and Batavia, n.d.). The spelling of these languages, and also of Bisayan, is modified in conformity with the rules given in Isidore Dyen, The Proto-Malayo-Polynesian laryngeals §§6–25, §§30–1 (Baltimore, 1953), except that I write ? and for his q and N (so also in reconstructed forms). Indication of stress in Tagalog is omitted, since if vowel length is regarded as phonemic and therefore given representation in the orthography the position of stress is, within the domain of the word, predictable.

² Additional types of correspondence for which there are only a few examples are noted by Dyen, Laryngeals §§102–11.

³ Otto Dempwolff, Vergleichende Lautlehre des austronesischen Wortschatzes 1.47–8, 50–1, 2.17, 19 (3 vols. = Beiheft zur Zeitschrift für Eingeborenen-Sprachen 15, 17, 19; Berlin, 1934–8).

⁴ Dempwolff 1.76.

⁵ Dyen, Laryngeals §§62–75, §112. I use Dyen's symbols for the phonemes of the protolanguage, and normally prefer his reconstruction to Dempwolff's when the two differ.

⁶ For the details, see Dyen, Laryngeals §§76–107.

⁷ Dempwolff included the Ngaju Dayak and Fijian cases in his comparisons but regarded their finals as ‘unexplained exceptions’ (3.91, 2.134) and failed to notice the coincidence with the ‘irregular’ ? of the Tagalog cognates; he did not cite Old Javanese data. Dyen did not use material from Ngaju Dayak and Fijian in his Laryngeals, and made only limited use of Old Javanese (see his §34).

⁸ Ngaju Dayak is cited from Aug. Hardeland, Dajacksch-deutsches Wörterbuch (Amsterdam, 1859), with -au, dj, and ng replaced by -aw, j, and .

⁹ Dyen, ‘The Ngaju-Dayak “old speech stratum” ‘, Lg. 32.83–7 (1956), has convincingly demonstrated that it is the words showing the correspondences characteristic of Dempwolff's ‘old speech stratum’ that are to be regarded as the inherited segment of the Ngaju Dayak vocabulary.

¹⁰ Fijian is cited from A. Capell, A new Fijian dictionary (Sydney, 1941), with g replaced by .

¹¹ The first gloss is from Pedro Serrano Laktaw, Diccionario tagálog-hispano, segunda parte 538 (Manila, 1914) : ‘venta (ambulante de mercaderías, recorriendo calles y acercándose de casa en casa)‘. The second is from P. Jacobo Enriquez and Maria Odulio Guzman, A Tagalog-English vocabulary 49 (Manila, 1949).

¹² Laktaw 538: ‘vender (mercancías, recorriendo calles y pueblos, y acercándose de casa en casa)‘.

¹³ The source for Old Javanese is H. H. Juynboll, Oudjavaansch-Nederlandsch woordenlijst (Leiden, 1923), with ng replaced by .

¹⁴ ö is the conventional transcription of the Old Javanese vowel sign which represents the long vowel corresponding to the short vowel transcribed ĕ (H. Kern, Verspreide geschriften. 8.185 ['s-Gravenhage, 1918]).

¹⁵ Jv. u-u is the regular correspondent of OJv. ĕ-ö: OJv. h-um-ĕlö, Jv. ulu ‘swallow’; OJv. tĕkö, Jv. tuku ‘buy’; OJv. ma-wĕrö, Jv. wuru ‘drunk’. Further examples are given in R. A. Kern, ‘De Soendasche ö-klank’, Bijdragen tot de taal-, land- en volkenkunde van Nederlandsch-Indië 99.111–8 (1940).

¹⁶ There is a single case in which Tagalog has a doublet. Beside the forms la:ko? and mag-la:ko? cited in §6 from Laktaw and Enriquez and Guzman (forms in -? are also recorded in Leonard Bloomfield, Tagalog texts with grammatical analysis 362 [University of Illinois studies in language and literature 3:2–4; Urbana, 1917] and Manuel Cruz and S. P. Ignašev, Tagal'sko-russkij slovar' 177 [Moscow, 1959]), there exist lako ‘sell, peddle around’ (A National Language-English vocabulary 94) and la:ko ‘merchandise, sell’ (Preliminary studies on the lexicography of the Philippine languages 7.9 [Publications of the Institute of National Language 1:1–11; Manila, 1937–40]). Dempwolff 1.76 states that the form of the Tagalog cognate of *palu (§1) given by Laktaw lacks ? and thus constitutes a doublet with the forms with ? given by other sources; but the only form I find in Laktaw 990 is pa:lo?.

¹⁷ Cited from J. Kaufmann, Visayan-English dictionary (Iloilo, P.I., n.d.).

¹⁸ Dempwolff 3.76 regards Tg. -al-, Toba Batak -r-, and Ml. -r- as infixes, but this treatment is unconvincing.

¹⁹ Dempwolff 1.90 (with reference to 1.87–8) considers both vowels of Jv. k-ruyu to be unexplained exceptions to the regular development of *e in Javanese. However, *-eR regularly becomes -ö in Old Javanese—e.g. ‘hear’ < ‘take care of, do one's utmost’ < *ibeR, linawĕ-lawö ‘it lasted a long time’ < *laweR (reconstructible from the OJv. base lawö in comparison with Tg. la:wig ‘length or duration of time’)—and to ĕ-ö in Old Javanese there regularly corresponds u-u in Javanese (fn. 15). Hence the vowel correspondences of both OJv. and Jv. ‘hear’ < are regular.

²⁰ For the u of the first syllable, see fn. 15. The only other reconstruction with *e-aw of those given in §9 (3) for which I have found a Javanese cognate is *delaw: Jv. dula ‘look at’.

²¹ The reconstruction of *-ey was proposed by Dyen, ‘On the history of the Trukese vowels’, Lg. 25.420–36 (1949); see 421 fn. 5.

²² Laryngeals §53.

²³ Dempwolff 3.25 cites the Ngaju Dayak word as (converted to my spelling) e-wau, thus indicating that he regarded the e- as a prefix not derivable from the etymon and that he interpreted Hardeland's ewau as a trisyllable with syllabic u. Hardeland's dictionary does not make a distinction between disyllabic au and diphthongal aw, and I do not know that such a distinction is phonemic in Ngaju Dayak. However, the Ngaju Dayak translation of the New Testament (Soerat brasih Djandji Taheta ain Toehan toentang Djoeroesalamat itah Jesoes Kristoes³ [Leiden, 1926]), which follows Dutch orthographic conventions, does distinguish between aoe = my au and au = my aw. The word in question occurs in John 11.39 and 19.40, where it is spelled ewau; I therefore write ewaw in place of Dempwolff's ewau.

²⁴ Dempwolff 2.52 (*e), 48 (*b). Loss of *h is illustrated e.g. by NgD. ihop ‘drink’ < *hi Rup (Dyen, Laryngeals §119).

²⁵ Tg. a can come from *e as well as from *a (Dempwolff 1.74–5). In Malay, *e normally becomes ĕ, but there are no cases to contradict an assumption that when *e is followed by a laryngeal it is reflected as a. The Javanese and Old Javanese forms exhibit regular loss of *h and contraction of vowels (Dyen, Laryngeals §53); that *behew would contract to OJv. bö (regularly corresponding to Jv. bu) seems reasonable, since the ö developed from the final *-ew is the long counterpart of the short vowel ĕ which is the normal reflex of the penultimate *e. The development of Fi. bo from *behew would be a case of the contraction of like vowels after the regular loss of *h, since both *e and *ew yield Fi. o.

²⁶ Tongan po-a ‘have a fishy smell’ (C. Maxwell Churchward, Tongan dictionary 412 [London, 1959]) can be derived from the same combination of elements as NgD. bewan.

²⁷ For the loss in Ngaju Dayak of an antepenultimate syllable with initial laryngeal or vowel, cp. baha ‘shoulder’ < *(?0)abaRa(h) (Bisayan ?abága [Rodolfo Cabonce, Visayan-English dictionary 18 (n.p., n.d.)], Toba Batak abara, Sa'a ahala, apala ‘shoulder’; in the absence of Malay and Tongan cognates, the initial is reconstructed ambiguously as *(?O)-in accordance with Dyen, Laryngeals §135) and pero ‘bile’ < *?apeju(h). Another possible example is hej an ‘staircase, ladder’, the etymon of which is either *harezan or *Rezan (for details, see my article ‘The Proto-Malayopolynesian word for “ladder, staircase” ‘, forthcoming in JAOS).

²⁸ Dyen, ‘Dempwolff's *r‘, Lg. 29.363–4 fn. 18 (1953). In this discussion I shall not maintain Dyen's distinction of several varieties of *r, since it does not affect the question of the reconstruction of *-ew.

²⁹ Dyen's gloss is ‘pursue’; I follow J. Warneck, Tobabataksch-deutsches Wörterbuch 41 (Batavia, 1906).

³⁰ Dyen discussed the possibility of regarding Jv. buru as a Malay loanword and pointed out that the inclusion or exclusion of the Javanese word does not affect the reconstruction of the final, but only that of the medial consonant.

³¹ The only other possible example of Tg. -aw for anticipated -o? < *-ew is in the case of *lakew, where, beside the la:ko? cited in §6 and the lako and la:ko noted in fn. 16, there is a form la:kaw ‘roam’ given in Preliminary studies 7.56. The authenticity of this form, however, is questionable, since it is not recorded in the other Tagalog sources which I have consulted. Bisayan, which in the case of bahó? (Sebu, Samar-Leyte dialects) ‘bad odor’, báho? (Hiligaynon dialect) ‘odor’ (Preliminary studies 2.36) < *behew and tábo? (Samar-Leyte dialect) ‘dipper’ (but Hiligaynon tabó, Dyen, Laryngeals §97) < *tabew agrees with Tagalog in having -o? for *-ew_y also has lakáw (Sebu dialect) ‘go on foot, go away’, (Samar-Leyte dialect) ‘walk, roam’ (Preliminary studies 7.56; the word is also recorded, without specification of dialect, in Juan Félix de la Encarnación, Diccionario bisaya-español³ 181 [Manila, 1885] and Cabonce 133), but appears to lack forms in -o? comparable to Tg. la:ko?.

³² For convenience I retain Dyen's reconstruction of final vowel as the source of those instances of *-u which are not derivable from *-ew, since it is not the purpose of this paper to subject that reconstruction to critical review. His attempt to demonstrate the survival of distinctive morphophonemic treatment of words with original *-V in several major branches of Malayopolynesian must be adjudged unsuccessful : the ? in the Tongan word for ‘father’ as given by Dempwolff, upon whom Dyen relied for his Tongan material (Laryngeals §35), is spurious, going back to what is apparently a misprint in one of the entries in Dempwolff's source (A. Colomb, Dictionnaire toga-français et français-toga-anglais [Paris, 1890]), and most of the Javanese morphophonemic irregularities which Dyen cited can be explained in other ways. Nevertheless, his reconstruction may still be the most likely way to account for the cases of unanticipated final ? in Philippine cognates of words with final vowel in other languages (other than those cases which now can be traced back to *-ew) and for the morphophonemic irregularities in Tagalog, Bisayan, and Bikol which constitute the bulk of his evidence.

³³ For examples of Old Javanese doublets with -ö and -u, see Kern, ‘De Soendasche ö-klank’ 114–6.

³⁴ In fairness it must be admitted that it is also possible that archaizing scribes may have occasionally substituted unhistorical -ö for an etymologically correct -u.

³⁵ See Dyen, The Ngaju-Dayak “old speech stratum” '.

³⁶ Wilkinson gives only ‘sorrowful’; the second gloss is from H. C. Klinkert, Nieuw Maleisch-Nederlandsch woordenboek⁴ 996 (Leiden, 1930).

³⁷ P. Jansz, Practisch Javaansch-Nederlandsch woordenboek³ 776 (Semarang, n.d.). The word is not in Pigeaud.

³⁸ Laktaw 1222.