¹ Malay tiga ‘three’, Lg. 22.131–7 (1946). Abbreviations of journals and publications are listed in fn. 1 of that article. Abbreviations of names of languages follow: Bar., Bare'e; Bob., Bobongko; HlBs., Hiligaynon-Bisayan; Jv., Javanese; Ml., Malay; OJv., Old Javanese; PMP, Proto-Malayo-Polynesian; Snd., Sundanese; Tg., Tagalog.

² In a letter kindly written to me by the Dutch scholar P. Voorhoeve.

³ S. H. Ray, The languages of Borneo, Sarawak Museum journal 1:4.1–196 (1913). The word is ascribed to Sibuyau, Lemanak, Balau, Skarang, and Bugau (58); all these are identified as Sea Dayak (7).

⁴ Gajösch-nederlandsch woordenboek met nederlandsch-gajösch register (Batavia, 1907).

⁵ Een maleisch telwoord in het teloegoe?, Tijd. 46.58 (1903).

⁶ De Bare'e-sprekende Toradja's van Midden-Celebes 3.61 fn. (Batavia, 1914).

⁷ Only so in Th. Pigeaud, Javaans-nederlands handwoordenboek (Groningen, Batavia, n.d.); P. Jansz, Practisch javaansch-nederlandsch woordenboek (Semarang, Soerabaia, Bandoeng, 's-Gravenhage, n.d.).

⁸ Old Javanese words are cited from H. H. Juynboll, Oud-javaansch-nederlandsche woordenlijst (Leiden, 1923). The macron over the u of hantělū is taken to have no effect on the following discussion. The word angantělu ‘lay eggs’ is cited without the macron.

⁹ See fn. 11.

¹⁰ De soendasche ö-klank, Bijd. 99.111–8 (1940).

¹¹ In his letter (fn. 2) P. Voorhoeve writes: 'The etymology for tiga generally current among Dutch linguists is that it is an Indonesian word, originally meaning the (three) hearthstones, and used as a tabu-substitute for tělu on account of some unfavorable connotation of this latter word, perhaps because of its likeness to some form of tělur 'egg', as words for egg often also mean testicle. ... I think its auctor intellectualis was Dr. N. Adriani.'

¹² J. Kaufmann, Visayan-English dictionary (Iloilo, P. I., n.d.).

¹³ Adriani, Woordenboek sub tigangi.

¹⁴ Perhaps to be added to these, under an assumption that the difference between o and u will ultimately be explained, is Todjo-Bare'e buga 'alocasia macrorhiza (a plant)', which can be compared with Tg. bigá 'Homalomena' (cited by Dempwolff from Merrill, A dictionary of the plant names of the Philippine Islands 32 [Manila, 1903]), Toba-Batak bira ‘a tuberous plant’, Ml. birah ‘alocasia’. (If the Bare'e word be ignored, the comparison would yield PMP biRaq.) However, Ampana wia ‘alocasia macrorhiza’ can be compared just as well under the easily demonstrable formulas (1) *b became Bar. w (e.g. Bar. wawu ‘pig’ from *babuy, Ml. babi 'id.') and (2) *R disappeared in Bare'e (e.g. Bar. ua 'vein, tendon' from *uRat, Ml. urat ‘id.‘). Adriani considers buga to be a loanword from Parigi, a Toradja language; whether it is or not, it does not clearly support the formulas, because Ampana wia does not explicitly show that Bare'e had u in this word from PMP i or any other PMP vowel: even if it is assumed that the Ampana word went through a stage *wuya or the like, the shape of the word itself does not compel such a theory. That is, the resemblance of buga may be fortuitous.

¹⁵ The assumption could be that one of the two reflexes, probably the second, indicates a borrowing from a dialect or another language. Another example of the l-reflex is Bar. kuli ‘bark’ from *kulit, Ml. kulet ‘skin’. It should be noted that these instances are not specified as belonging to the ‘priestertaal’, the language used in the songs of priests and in poetry, in which l sometimes appears when it is lacking in the corresponding ‘ordinary’ word; e.g. talu ‘three’, whose first vowel then also causes difficulty and suggests a loanword.

¹⁶ Ml. sělah, with about the same meaning, perhaps resulted from contamination with Ml. cělah ‘cleft, rift, cavity between two connected objects’.

¹⁷ Otherwise the evidence of tógisi cannot be applied to the history of togo.

¹⁸ An example of what may turn up in the case of minimally supported formulas is the following. It is possible to construct the formula: *R becomes Bar. r (perhaps dividing *R into *R ₁, which disappears in Bare'e, and *R ₂, which becomes Bar. r), on the basis of the assignment of Bar. siru ‘sip’ to Dempwolff's reconstruction *siRup (TBt. sirup, Ngaju-Dayak surup 'id.', with assimilation), and of Bar. kora 'stiff, dry, fast' to *keRaŋ (Toba-Batak horaŋ, Ngaju-Dayak kehaŋ 'dry'). This formula would make it possible to assign Coast-Bare'e man-tirís-ani 'let trickle through a cloth', Ampana and Lalaeo (another Bar. dialect) mon-tirís-i ‘strain, let trickle through a funnel-shaped basket (pontiris-i) to purify’ to *tiRis and ignore Adriani's suggestion that these are ‘from Malay’. Such an assignment might now be enough to lead us to regard the resemblance of Bar. tógisi to *tiRis as coincidence. This example illustrates the point that minimally supported formulas are different from well supported formulas (even though a boundary between them would be difficult to draw) in that the chance factor has not been completely eliminated in the former type. This chance factor may appear in a coincidence of correspondences. Although it cannot be denied that the formula attempted above has some probability, this must be considered low in view of the following. Dempwolff reconstructed *siRup apparently because he was persuaded that it was a by-form, perhaps with a prefix, of *hiRup (Tagalog hí:gop, Ml. (h)irop ‘sip’, Ngaju-Dayak ihop ‘drink’; Ngaju-Dayak hirup ‘sip’ could be a Malay loanword). Adriani also quotes the latter words in connection with Bar. siru. The *R of *hiRup is certain, but it does not follow that *R must be reconstructed in the other case. Instead, *r could be satisfactorily reconstructed, with a resultant disconnection of the two reconstructions. Furthermore, Bar. kora could be assigned to *keras ‘hard’ (Toba-Batak horas, Ml. and Jv. kěras ‘id.‘); in this way the formula attempted in this footnote is eliminated, unless indeed Bare'e tiris should otherwise prove assignable to a dialect in which *R became r.

¹⁹ If tógisi is a borrowing from another language (not another dialect), its application to our problem would be too unclear to allow us to consider it. This remains true even if togo is also a borrowing, but from a different language. If togo is a borrowing from the same language, its history could be attacked only after identifying the source language. However, the assumption that another dialect is involved is not perfectly clear in the case of Bar. g from *R. Apparently other dialects than Ampana have pujiti ‘polyp’ (Adriani, Woordenboek sub puji), with assimilation of the final vowel to the preceding one if the word is from *-uRita. Since phonemic principles have probably not been adhered to (witness to date of Adriani's publications, and the probable complementation of Bar. c and s) the sequences -ugi-, -ui-, and -uji- may conceivably be phonemically alike. If then Bar. graphic g occurs here, let us say, for a palatal voiced stop (recorded g or j) varying with an alternant which is not always recorded before vowels (here before i), the problem of the history of togo disappears: we have /toyo/ from *telu(h). But questions concerning the value of the written symbols can only be answered by a descriptive study of the language; they do not concern the comparatist qua comparatist, even though he cannot escape the errors which derive from a faulty description. A systematically accurate description is one of the comparatist's assumptions. Here then the describer's statements are taken at face value.

²⁰ See Adriani, Woordenboek sub soga.

²¹ If a typographical error is not involved. There is no indication that this is the case in an appended ‘Aanvullingen en verbeteringen’ to Adriani's Woordenboek.