¹ The works referred to are:

Dempw. = O. Dempwolff, Vergleichende Lautlehre des austronesischen Wortschatzes; Zeitschr. f. Eing.-Spr., Berlin:

1. 1.

   1. Induktiver Aufbau einer Indonesischen Ursprache; 15. Beiheft, 1934.

2. 2.

   2. Deduktive Anwendung des Urindonesischen auf Austronesische Einzelsprachen; 17. Beiheft, 1937.

3. 3.

   3. Austronesisches Wörterverzeichnis; 19. Beiheft, 1938.

Bloomfield = L. Bloomfield, Tagalog Texts with Grammatical Analysis; University of Illinois Studies in Language and Literature, Vol. 3, Nos. 2–4, 1917:

Part II: Grammatical Analysis, No. 3.

Part III: List of Formations and Glossary, No. 4.

Laktaw = P. S. Laktaw, Diccionario Tagálog-Hispano, Segunda Parte; Manila, 1914.

Encarnación = J. F. Félix de la Encarnación, Diccionario Bisaya–Español, Tercera Edición, Aumentada ... por J. Sanchez; Manila, 1885.

Except where specially noted, Bisayan words are taken from the following and quoted after the abbreviation Bs. without a note if listed under all of the three dialects Sebu, Hiligaynon, and Samar-Leyte:

Prelim. Stud. = Preliminary Studies on the Lexicography of the Philippine Languages; Publications of the Institute of National Language 1.1–11, Manila, 1937–40.

Words and meanings other than those of Tagalog, Bisayan, and Malay are taken from Dempw. 3.

Reconstructions stated without comment are likewise taken from Dempw. 3.

Abbreviations of names of languages: Bs. = Bisayan; HlBs. = Hiligaynon Bisayan; Hv. = Hova (Malagasy dialect); Jv. = Javanese; Ml. = Malay; NgD. = Ngaju-Dayak; MP = Malayo-Polynesian; PMP = Proto-Malayo-Polynesian; SeBs. = Sebu Bisayan; SLBs. = Samar-Leyte Bisayan; TBt. = Toba-Batak; Tg. = Tagalog.

² See Dempw. 3 sub ḍuva̔. In writing reconstructions 1 find it convenient to use a set of symbols which is in part different from Dempwolff's. Hereafter I quote reconstructions in my own symbols. In the following list of equivalent symbols, mine precede the colon and Dempwolff's follow it: c : k', D : ḍ, e : ә, intervocalic h : ‘, ˀ : intervocalic and final h, ñ : ń, r : ḷ, R : γ, s : t’, T : ṭ, w : v, y : j, z : d'. I write nothing for Dempwolff's initial and final ʽ. No difference in phonological interpretation is implied except with regard to the substitutions for Dempwolff's h and ʽ.

³ This bound form is abstracted from Tg. dalawá ‘two’, ikalawá ‘second’. The form dalawá' appears in Dempw. 3 sub Duwa by a typographical error, but is quoted correctly elsewhere.

⁴ From *n-Duwa, Dempw. 2.183 f.

⁵ The languages before Fiji are classed as Indonesian, Fiji and Saa as Melanesian, and the last three as Polynesian.

⁶ Dempwolff (1.78) treats this correspondence as one of the ‘Tg. vokalische LU. [= Lautunstimmigkeiten], die nicht durch Assimilation oder auf andere Art zu erklären sind’ (1.74 f.) Although apparently a change of *u to Tg. a must be assumed in tracing Tg. lansáŋan ‘street’ from *lancuŋ (TBt. latsuŋ, Ml., Jv. lancoŋ 'go for a walk'), the conditions of the change in this isolated example are not understood as against the regular Tagalog retention of *u. Likewise difficult to understand is the change of *i to Tg. a—as against the regular retention of *i—in the single instances haygàn 'border, until' if it is traced to *hiŋga ‘border’ (Jv. iŋga ‘until’, Ml. hiŋga ‘border, until’, NgD. hiŋga ‘border’, iŋga ‘until‘). Other than mere phonetic change may have been involved in history of these two words.

⁷ See Dempw. 1.73 f. for further examples of the changes indicated. An example of *e becoming u(o) is *kebut, Tg. kubòt ‘wrinkle’. It is clear that a phonemic difference u : o must be set up in describing Tagalog because of the single minimally contrasting pair pùˀ ‘ten’ and pòˀ ‘sir’. Tg. u and o otherwise do not contrast and the rules for their occurrence can be stated (see Bloomfield, Part II 134 f.). It is therefore a fair assumption that Tg. u and o arose from the same sound.

⁸ Dempwolff's *w and *y may turn out to be positional variants of *u and *i respectively. Since this point would require a separate discussion, I treat *w and *y as phonemes in this article.

⁹ The complete list is given in §1 in the comparison of the words for ‘two’.

¹⁰ I treat only of non-final *ey and *ew. Whenever *ey and *ew are mentioned in the following discussion, reference is made to them only in non-final positions. If final *ey and *ew (thus far not reconstructed) did occur, it may be that they had a different history.

¹¹ S. van Ronkel (Maleisch Woordenboek, 3d ed.; den Haag, 1930), sub jakin and warta respectively, quotes mĕjakinkan (= mĕyakinkan) ‘deal earnestly with’ and mĕwartakan ‘inform’, which seem to indicate the occurrence of ĕy and ĕw across morpheme boundaries. The first element of these two words is the common Malay prefix mĕ(ŋ))-.

^11a Dempwolff (1.74 f.) includes this change among the Tagalog ‘Lautunstimmigkeiten’ referred to in note 6. But Tg. a from *e before a syllable containing a reflex of *u and *i is exemplified a number of times—at least 12 times before *u and 3 times before *i. The frequency suggests that it is due to a regular change, but one that is not characteristic of the standard Tagalog dialect. It is interesting to note that in neighboring Bikol penultimate *e regularly became a if the ultima did not contain *e; see C. E. Conant, The Pepet Law in Philippine Languages, Anthropos 7.942 (1912).

¹² I have adopted Dempwolff's use of brackets to enclose an incompletely determined part of a reconstruction. The choice between two phonemes so enclosed cannot be made from the forms compared.

¹³ Bisayan is a Philippine language which will undoubtedly prove to be closely related to Tagalog. Although the two languages are geographically separated by a body of water, a complete dialect analysis may demonstrate that there is no language boundary between them. Thus far the chief differences between Bisayan and Tagalog are their respective reflexes of *e, *d, *D, *j, *z. Although at first glance the difference may seem to be considerable, I believe that a complete comparison will show that the differences are hardly greater than those to be expected in somewhat divergent dialects. Consequently, except where one of the above reconstructed sounds is involved, I treat the Bisayan evidence as the equivalent of Tagalog evidence.

¹⁴ Whether Bs. u and o are different phonemes, as they are in Tagalog, is indeterminate at present. Even if they should be distinct phonemes, it is quite clear from many comparisons that both reflect *u and *e indistinguishably.

¹⁵ For analogical changes of originally final stops in Hova, see Dempw. 2.84.

¹⁶ Quotable from Sebu and Hiligaynon.

¹⁷ Other examples of a reconstruction beginning with vowel or h and having about the same meaning as a reconstruction with initial t are *ha(m)buR beside *tabuR ‘strew’ and *i(m)pun beside *[t]i(m)pun ‘gather’. It may be that a prefix *t- is involved which replaced an initial *h.

¹⁸ Dempwolff (1.50) specifically abjures the reconstruction of ˀ. I believe that the reconstruction of ˀ where Dempwolff writes intervocalic and final h will permit a simpler treatment of phenomena found in the Philippine languages. For example, the ˀ in Bs. luyˀa—which in my opinion cannot be secondary—within Dempwolff's system would have to be traced to the same ‘weicher zwischensatz’ (written ʽ) as the intervocalic h of Bs. buhat ‘work, raise’. I consider that Dempwolff's interpretation results in needless complexities at this point. The PMP laryngeals will be treated fully in a later article.

¹⁹ Cf. Dempw. 2.152. For exceptional Saa ˀ for *ˀ, see 2.154.

²⁰ Dempwolff (2.153) classes it under ‘occasional assimilations’.

²¹ Dempw. 3, sub Duwa.

²² I assume with Dempwolff (1.91) that Jv. r as well as ḍ can reflect *D.

²³ Dempw. 1.96.

²⁴ For this reason Dempwolff was forced to assume the irregular loss of intervocalic *w in explaining Jv. -ro (1.93) from *Duwa.

²⁵ So cited by Dempwolff. Despite the fact that beside lowal Laktaw cites lual, I take it that Dempwolff's citation is correct. According to the general practice of not writing ˀ between vowels in transcribing Philippine languages, lual would seem to imply /luˀàl/. Laktaw rather consistently writes uä for /uˀa/, e.g. lutuän ‘be cooked with’. Furthermore Bloomfield in Part III cites no Tagalog bases containing /uˀa/, this sequence occurring only at morpheme boundaries (cf. lutúan ‘be cooked with’ sub lútoˀ ‘a, cooking’, for /lutúˀan/ by Part II, Par. 14). It appears then that Laktaw's uwa and unmarked ua can be considered equivalent writings.

²⁶ Dempw. 1.86 f.

²⁷ Cf. op.cit. 50, 53. PMP *h and *R are reflected in exactly the same way in Javanese and can be assumed to have fallen together. Both disappear in all positions and the crasis that accompanies their disappearance in intervocalic position is identical.

²⁸ Jv. w as well as b reflects *b, op.cit. 35, 41.

²⁹ The suggestion in §7 that Jv. tiyup (if it is not a Malay loan-word) comes from *teyup (= *t- + heyup) is in complete accord with these indications that *ewa probably did not become Jv. o. Javanese treats *y and *w similarly; if *ew + vowel suffered crasis, *ey + vowel should likewise be expected to undergo crasis. Yet Jv. tiyup would indicate that *eyu became Jv. iyu without crasis. Examples of the crasis of *e-u are not available, but since the crasis of *e-a, *e-e (*teReb, Tg. tigíb ‘overloaded’, Jv. tub ‘full’; only example, with surprising Javanese vowel) and *a-u (*sahut, Tg. sahót ‘vow’, Jv. sot 'curse') took place, there is reason to believe that further research will provide an example of the crasis of *e-u. The possibility that the prior change of *ey to Jv. i may have hindered crasis seems to be eliminated by the example of *diRus, Jv. dus 'bathe', TBt. duris (with metathesis) ‘sprinkle’, Ml. diros ‘wet’, HlBs. digos ‘bathe’.

³⁰ Cf. Dempw. 2.129, 149, 171.

³¹ Cf. Dempw. 1.50 f., 2.79.

³² Dempw. 2.17, 48.

³³ Dempwolff (1.78) believes that dalawà ‘two’ contains a prefix. It is perhaps more likely that the additional element is an old reduplication.

³⁴ If this explanation is adopted, it would seem to follow that Bs. dohaon (ˀ/dohaˀon/, Encarnacion sub doha) ‘be made two of, be divided’ resulted from a dissimilation affecting the second h of an earlier *duhahon. (The suffix -on is the Bisayan equivalent of Tg. -in [from *-en], and the normal sandhi consonant in Bisayan as in Tagalog is h.)