REVIEW ARTICLE A GUIDE TO THE FILIPINO LINGUIST TOWARDS READING AND APPRECIATING Blust, Robert A. 1980. Austronesian Etymologies. Oceanic Linguistics, Volume 19, Numbers 1 & 2, Summer-Winter 1980. (189pp) Reviewed by: R. David Zorc, School of Australian Linguistics, Darwin Instiute of Technology PART ONE. BACKGROUND INFORMATION. (l) LANGUAGES USED. A scholar must include a wide range of languages sufficient to give a full enough picture of both the sound system (phonology) of the parent language and the distribution of its daughter languages. In the Philippines, for example, one could theoretically reconstruct Proto Philippine on the basis of a comparison of Ilokano and Tagalog, since they are historically and geographically separated, but this reconstruction would miss numerous important distinctions and give a very limited view of what the parent language had in its lexicon and morphology. In my review of Paz's PPH reconstruction (l981), I commended her treatment of 29 widespread Philippine languages (Zorc 1981). Blust's "present study draws on approximately 200 of the 700 or more AN languages" (9) representing diverse subgroups of the family, and an increase of at least 142 languages since his initial major contribution (l970). Blust email to April | 2018.08.15 | 15 August 2018 The cognation (relationship) of words has to be established by ruling out chance, borrowing and language universals as alternative explanations for similarity: cognates may be phonetically similar, or phonetically very different ---- that is simply irrelevant to determining an etymology. (2) CORRESPONDENCE CLASSES. The process of historical reconstruction involves the search for similarities in four areas: sound, form, function, and meaning, or, in the jargon of the field, the isolation of phonological, morphological, grammatical, and semantic correspondences. These, of necessity, involve a further four degrees of resemblance: (a) identical, (b) regular, (c) irregular, or (d) false correspondences. (2a) Identities yield rather straightforward reconstructions: Tagalog, Ilokano, Malay, and Fijian 'eye' yield a parent form, *mata (the asterisk signifies that the form is hypothetical--we have no surviving proof that it was so). Note that every element is identical: the sounds m-a-t-a, usage as a noun, and semantics. [Only a few of Blust's discoveries are of this type; some examples are his etymologies which are IDENTICAL in form and meaning include: #22 *atag 'cross-beam', #29 *badi 'nervous spell', #33 *balaw 'timber tree', #48 *bantal 'bundle of cloth or clothes', #59 *besiq 'widen an opening' etc.--but the absolute identities are often more limited to form (sounds) rather than to meaning.] (2b) By far the most common type of correspondence set is regular, wherein a sound in one language regularly corresponds to a different sound in another, which probably descends from a distinct phoneme of the parent (proto) language. For example, Tag ka:'in, Ceb ka:'un 'eat' < *ka:'en [note Tag i :: Ceb u < *e] Tag la:yag, Malay layar 'sail' < *la:yaR [Tag g :: Malay r < *R]. (2c) Irregular correspondences often, but by no means always, involve loans. Thus, while *e goes to i in Tagalog (except when a neighboring syllable has *u, e.g., *penuq 'full' > Tag puno'), all instances of *e > a should be viewed with suspicion, e.g., Tag andal 'jostle' < Malay endal 'stuff, push', #120)--numerous loans from Kapampangan and Malay establish this. However, a similar development in Pangasinan is irregular and unexplained, e.g., talo 'three' ( < *telu), pano 'full' ( < *penuq), batik 'run' ( < *betik), sali 'foot'( < *seli), etc. It is difficult to maintain a hypothesis that these are loans (many of the etyma in question are specific to the South Cordilleran subgroup to which Png belongs), since no donor language with the same words and shifts is identifiable. A similar situation exists with Ilokano reflexes of *R--most scholars would agree that [r] is the expected reflex, yet [g] occurs on more numerous forms, some even as doublets Ilokano ba:go, baro 'new' < *baqRuH Ilokano *bibig, bibir 'lip' < *bibiR, so that the treatment of forms with [g] as irregular correspondences is more attractive than a loan hypothesis (viz: a split of *R > r and g might have occured in the history of the language). More recent studies seem to suggest that early Ilokano was heavily influenced by Ibanag, which at that time long ago, was more influential in northern Luzon than was Ilokano. Ibanag regularly has *R > g. This would seem to imply that many Ilokanos at that time were bilingual in Ibanag and Ilokano. (2d) False correspondences involve loans or accidental similarities. In a family, such as Austronesian, where so many influences have been felt Spanish - throughout the Philippines, more so in the lowland Christianized languages Portuguese - mainly in Indonesia Indic/Sanskrit via Malay Arabic - directly in Islamicized groups, indirectly via Malay Persian - via Malay Chinese - Southern Min and other varieties in commerce and cuisine Dutch - mainly in Indonesia English - starting with the Thomasites in 2007, and especially after World War II etc.), a scholar cannot be expected to be a master of all potential sources. Hence, a beginner might be excused for making spurious reconstructions, e.g., [x]gadapun 'water-jar' ( < Sp garrafo'n 'large decanter') or [x]mani' 'peanuts' ( < Mex-Sp mani' 'peanut'). [The raised x indicates a false correspondence.] But even a veteran scholar might include foreign material in what may be considered a legitimate etymon (see 2.8 below). Spanish anca(s) 'rump' > Tagalog angkas 'ride on s.t. with s.o.' (precisely where a second person would ride on a carabao or horse). Blust in his earlier work, as did Dempwolff before him, included appropriately coded loan material in order to exemplify regular correspondences in the languages treated. However, some readers mis-interpreted this as "a literal claim that Sanskrit or even Arabic loanwords were found in Proto-Austronesian," (29) which clearly was not the case. Section 8 acknowledges the problems involved in identifying intimate borrowings and the fact that Malay has had a strong influence in insular Southeast Asia. While it is most difficult when reflexes are regular (especially in the case of identity correspondences, see l.2a), I would suggest a rule of thumb that excludes new etyma where Malay and only one Philippine language are witnesses. This would then affect reconstructions #70, 120, 128, 131, 135, 155, 177, 195, 275, 302, 382, 417, 440, 44l. The problem of Hispanic influence has affected two reconstructions. The inclusion of Chamorro man~a 'custom, habit, tradition' [ < Sp man~a 'skill, knack; habit' (in the phrase malas man~as 'bad habits')] at #297, leads to a reconstruction fraught with difficulties [*mana, *manaq, or *man~a 'inherit(ance)'] which could be the result of widespread Hispanic/Portuguese influence. A similar problem is encountered at #l33 (*gana, *ganas 'raw energy, animal appetite' < Sp gana 'desire, appetite', ?s 'plural') for at least the Philippine evidence??it is hard to see how Old Javanese could contain anything but inherited material, but difficult to determine which forms are legitimately cognate. (3) RECONSTRUCTION OF PARENT LANGUAGE PHONOLOGY. Blust, as well as most researchers nowadays, uses the PAN phonology reconstructed by Dempwolff as revised by Dyen (1947, 1951, 1953a, 1965). The following table indicates the status of the PAN phonemic inventory to date [excluding the subnumerals as discussed in Dyen 1953b (*R1?R5) or Tsuchida 1976 (*S1?S6), see also 2.5 and 3.1 below for further details]: PAN AUTHOR NOTES ACCEPTABILITY *a Demp yes *b Demp yes *B Prentice & Nothofer no *c Dyen Demp & Dahl *k' yes?Blust; Wolff rejects *C Dyen Dahl *t2 yes, on Formosan evidence *d Demp Dahl d1 yes, but Wolff rejects *D Dyen Demp *d., Dahl *d2 problematic (Tagalic, Javanese, and Paiwan/Formosan evidence need comprehensive re?investigation) *d3 Dahl maybe; Blust rejects *e Demp Demp * (schwa) yes *e' Dyen (Mid front vowel) no *g Demp yes, but Wolff rejects *h Demp Revised by Dyen yes, but = *S in Blust *H Dyen Extended by Zorc yes, on Formosan evidence *i Demp yes *j Dyen Demp & Dahl *g' yes *k Demp yes *l Demp yes *L Dyen&Ts Formosan evidence maybe; Dahl rejects; recent paper by Dyen & Tsuchida may re?instate *m Demp yes *n Demp Dahl *n1 yes *n~ Demp Possibly only PMP problematic next to *i *N Dyen&Ts Dahl *l? yes, on Formosan evidence *nj Demp (Orthographic "ng") yes *o Dyen (Mid back vowel) no *p Demp yes *q Dyen Demp *h yes *r Dyen Demp *l. yes, but Wolff rejects *R Dyen Demp & Dahl * yes *s Dyen Demp & Dahl *t' yes *S Dyen&Ts yes, on Formosan evidence *O? Tsuchida problematic; Formosan & Malagasy evidence need re?investigation *t Demp Dahl *t1 yes *T Dyen Demp & Dahl *t. yes?Blust; Wolff & Dahl reject based on Indic influence in Javanese *u Demp yes *w Dyen Demp *v yes, but Dahl rejects *W Dyen Probably *u marker no *X Dyen Probably = *S no *y Dyen Demp *j yes, but Dahl rejects *z Dyen Demp & Dahl *d' yes?Blust; Wolff rejects *Z Dyen Demp *d' yes, some problems *? Dyen & Zorc yes, some problems *: Zorc Vowel length Blust rejects * Zorc Vowel shortness Blust rejects (4) SUBGROUPING. Unlike determining correspondence classes, the establishment of subgroups depends on the isolation of shared differences (rather than similarities) ?? these constitute innovations which form the backbone of a subgrouping argument. The subgrouping hypothesis of each researcher must be made explicit. Since l977, Blust has evolved a sub grouping that recognises the indigenous languages of Formosa as falling into at least one first?order subgroup of Proto Austronesian, and the numerous languages spoken outside of Formosa taken together as the Malayo?Polynesian family. I am not in a position to take issue with the finer details of this hypothesis (but see 2.3). However, Blust's statement is a marked improvement over his earlier study, where (akin to a procedure used by Dempwolff) certa in reconstructions were labelled PAN (Dempwolff's UAN) despite a limited distribution in Indonesian and Philippine languages [which would clearly indicate a label of "Proto Western?Malayo?Polynesian" (in Blust's termino logy) or "Proto Hesperonesian" (in Dyen's)]. Hence, Blust's indication by a code [1 = Proto Austronesian (spoken around 5000 B.C.), 2 = Proto Malayo?Polynesian (spoken around 4000 B.C.), and 3 = Proto Western Malayo?Polynesian (spoken around 2000 B.C.)] is an honest appraisal of the status and also the approximate time?depth of each etymology. Any scholar who may disagree with these assignments can easily re?compute the status based on the scope of the languages represented and his own subgrouping hypothesis. (5) DOUBLET vs DISJUNCT. Blust has made a very useful distinction between reconstructions that are formally and semantically similar (doublets = Dempwolff's 'Nebenformen', e.g., *adaduq/*anaduq 'long', *kambing/*kanding 'goat') and reconstructions which have an overlap of cognate sets (disjuncts, e.g., Fijian kumi which could come from either *kumis or *gumi 'beard', Tagalog gata'? from *Rataq or (irregularly from) *getaq 'coconut milk'). However, Blust recognises that the term 'doublet' is still used to describe several quite distinct phenomena (27): phonologically similar reconstructions, e.g., *bingaq and *bingaR [volute shell], *baNaw and *baNaR Smilax ( = true doublets) and etyma containing a monosyllabic root [or "phonestheme", which he has since treated in more detail in a recent article (Blust, in press b)], e.g., *ket 'sticky, adhesive' or *ngaC, *ngeC, *ngiC, *nguC 'gnash the teeth as in anger'. To this list can be added suspect synonyms, e.g., *beRngi and *Rabi:?iH 'night'. I propose that the terms and abbreviations used can be adapted to include: doublets (Dbl), disjuncts (Dsj), monosyllabic roots (Mon), and synonyms (Syn). Where there is still some potential ambiguity as to the mixture of types, a convention can be adopted to mark suspect morpheme boundaries, e.g., *ti+ku?, *te+ku? 'bend, curve' (Dbl+Mon). At least some of the remaining difficulties would appear to have to do with the quality of specific language evidence rather than problems in labelling reconstructions. (6) TEST, CRITERION, AND WITNESS LANGUAGES. In an earlier paper, I indicated that a reconstruction is founded upon three different kinds of language evidence (Zorc 1982:ll4). (l) A test language is one wherein a phoneme directly reflects one??and only one??proto phoneme, e.g., Paiwan ts < *C, Ilokano e < *e. (2) A criterion language is one where a phoneme can relate to two or more proto phonemes, but with the evidence of additional languages, the comparativist can "triangulate" on the most probable corres pondence set, e.g., Akl ? < *q or *? and Iban ? < *? or *h so Akl ? + Ib ? < *?. (3) A witness language is useful primarily in determining the antiquity of an etymon--not its phonemic shape. Thus, with the loss of final consonants and/or numerous mergers in the Oceanic languages, a word might be descended from any of several etyma within a range of formal and semantic similarities. Amongst the languages of the North Cordilleran subgroup (Ibanag, Atta, Agta, Yogad, etc.), only Malaweg differentiates *s from *t, and is thus a test language for both phonemes; whereas Ibanag, which merges the two, as well as all final voiceless stops into glottal stop, is a witness language. (l) The "Brief History of Research" is an excellent survey of the field since Reland proposed 23 Malagasy-Malay comparisons in 1708 (representing the earliest Austronesian etymologies known), through Humboldt, Klaproth, van der Tuuk, Brandstetter, to Dempwolff (l938). Although there seems to have been a suspension of innovative etymological work in the two or more decades after Dempwolff, this need not reflect a tacit assumption that "the work had been done". (3) I suspect that this slow-down was in part due to the reorganisation of academic priorities during and after World War II (when teaching and learning "foreign" languages was foremost in the visions and finances of politicians and educationists). Also, scholars needed to understand more clearly the correspondences of reflexes in the numerous languages for which data was finally being published and the subgrouping relationships within the entire family. That might best be characterised as a period of consolidation, resulting in: the refinement of several correspondence sets (*q, *h, *D, *Z), including the introduction of new phonemes based on Formosan evidence (*S, *C, *N/L, *?, *H), and a productive debate over higher and lower order Austronesian subgroups. These endeavors consumed (certainly not wasted) a good deal of time, and were the necessary prerequisites for further careful etymological work. Blust's statistical analysis of the contributions to Austronesian lexical reconstruction are fair and accurate, and not at all self-serving. He has, after all, almost single-handedly doubled the 2,215 etymologies made available by Dempwolff (1938), and improved well over 600 reconstructions with the presentation of criterial evidence, thereby correcting either the phonological or semantic assignments. (2) In section 2, Blust outlines the scope of his study and other details of a comparativist's work, e.g., cognate decisions. Most of these have been discussed in l.l?6 above. Blust has taken great pains to establish the regular correspondences for each of the many languages he is dealing with, but admits that errors could crop up (10). Nevertheless, he has applied stricter criteria than in his previous studies and limits irregularities of most kinds to notes rather than the main entries for each new etymology. (3) Section 3 details Blust's view of the Austronesian family tree (see l.4 above). My only reservation is the large corpus of etyma that appear to be limited to Formosan, Philippine and/or Indonesian languages, which has led me to classify reconstructions of this type as PHF (Proto Hesperonesian-Formosan). Many of these are probably not innovations, such as terms relating to rice agriculture (*beRas, *pa:jay, *benSiq, *Semay, *Ri?ek, etc.), which were presumably lost in the Oceanic move. Some may be the result of borrowing or inter-influence going back quite far in time. Many too probably have Oceanic cognates as yet undiscovered--Blust has certainly presented many Eastern Austronesian cognates in his publications. But the residue is still significantly large enough to caution me to conservatism--if even a handful turn out to be innovations, they could substantiate a subgrouping hypothesis different from that proposed by Blust and Dahl. Reconstructions that might be thus affected are: #17, 23, 39, 41, 42, 55, 56, 58, 105, 158, 165, 212, 231, 270, 274, 353, 404, 409, 410b, 413, 431. (4) Section 4 deals with orthographic conventions for both data and etymologies. It is essential to have a uniform orthography, which sometimes differs from the original sources, so that data can be compared without requiring the reader to be a master of numerous and varying conventions. I applaud this procedure and commend it to other scholars. Note that only the following have been retained from the original sources: Paiwan tj vs ts and Palauan ch; and that Casiguran Dumagat e = schwa, while e' = mid front vowel (which is an inverse of the convention in the Headland dictionary). The Philippine convention of not writing intervocalic glottal stop is followed, and this alone I find unfortunate--it would have been clearer to indicate ??? in forms and languages where it does occur since the reconstruction of *??? vs *?O/? can be at issue (see #4, 5, 7l, 197, 228, 263, 275, 287). Blust's suppression of the complex convention of indicating ambiguities in reconstructions (such as *[qSO/]a(nj)bek 'mat', in favor of a more straightforward PHN *a(m)bek) is welcome, given the evidence of the data and the level of a reconstruction. The indication of homorganic nasals (rather than *njC) is also appreciated. Granting that it can be justified in a more abstract phonological interpretation, the evidence of some languages would appear to indicate non-homorganic clusters (e.g., PPH *hamtik 'wasp', PHN *halimtang 'cross-piece', possibly also PHN *kamding 'goat', and PHN *limtaq < PHF *qalimeCaq 'leech') which would be lost by the previous convention. The list of test and criterion languages for various PAN phonemes is useful, but does not indicate the degree of reliability of some languages, e.g., Malay, Samal, and Gorontalo are not particularly good witnesses for initial *S. Paiwan may turn out to be a far more critical test language than many of the others listed in evidence for *d, *D, *j distinctions, yet it is omitted (perhaps rejected?). (5) In section 5, Blust rejects various proposals for the revision of PAN phonology: Prentice (1974, for *B), Wolff (1974, against *r and *d), Dahl (1976, against *w and *y, for *d3), Dyen (1978, for *e' and *o, although the latter is possible at #28, 235, 354), and Zorc (1978, for contrastive accent). He tentatively accepts my proposals for the laryngeals *H and *? (Zorc 1982), noting the counter-evidence of some Iban and Sasak forms. He outlines two problematic correspondence sets: Iban ?? :: Malay ?h (where I suspect the Iban form is secondary and therefore would reconstruct *?q, giving priority to the Malay evidence), and Iban ?h :: Malay ?O/ (where I cannot explain the Iban form, but would reconstruct *?h or *?S giving priority to the current reflexes in Philippine or Formosan languages). He reiterates his position (Blust 1974) that certain languages of north Sarawak require the retention of *S as a sibilant in their immediate proto language, thereby necessitating *S for PMP (in all but final position) rather than just for PAN. Dahl (l976) and I (Zorc 1982) have maintained that only Formosan evidence can warrant the reconstruction of *S, and I have since argued (Zorc 1983:13?20) that (a) the north Sarawak reflexes were the result of accent phenomena yielding strengthened reflexes of *b, *d, *D, *j and (b) those languages appear to have lost *S and *H > *h in all positions. Since Blust and I are apparently at an impasse over this issue, scholars will have to determine if certain *S reconstructions are warranted -- my hypothesis suggests that #66, 83, 408, 409, 410 are, but #47, 194, 405, 406, 407 are not (due to lack of substantiating evidence from Formosa). However, this difference is, after all, minor because, as Blust points out (15), "information...is extractible from the cognate sets themselves once these witnesses are known" and the indication of *S can be interpreted as a shorthand for ambiguities, e.g., *kaSir = *ka[hHS]iR. (6) Section 6 briefly summarises his criteria for the reconstruction of morpheme boundaries. Blust's identification of rather standard affixes, e.g., *ma? [adjective], *?en [stative], *[S]a? [attributive] (with my suggestion for the addition of *S) is straightforward. I question the morpheme division at #23 (*qati?mela 'flea')??whereas a series of frozen affixes, including PAN *qaLi?, can be identified, neither *qati? nor a base *mela is warranted, as Paiwan qatjim+tjim appears to testify. If anything, a *qa? prefix and *?a suffix may be involved as indicated by Ilokano ti:mel, Ifugao ti:mol 'flea'. It is also possible that an *aR infix is present in #95 (i.e., *d&?&aya?), both ?g? and ?ag? have a pluralising function in Bisayan (see Zorc l977:115f). (7) The useful distinction between doublets and disjuncts is covered in section 7. I have commented on this above (l.5), but would note again that some problems involve not only the evaluation of reconstructions, but also of the languages used (see l.6). (8) Section 8 acknowledges the problems involved in identifying intimate borrowings and the fact that Malay has had a strong influence in insular Southeast Asia. While it is most difficult when reflexes are regular (especially in the case of identity correspondences, see l.2a), I would suggest a rule of thumb that excludes new etyma where Malay and only one Philippine language are witnesses. This would then affect reconstruc tions #70, 120, 128, 131, 135, 155, 177, 195, 275, 302, 382, 417, 440, 44l. The problem of Hispanic influence has affected two reconstructions. The inclusion of Chamorro man~a 'custom, habit, tradition' [ < Sp man~a 'skill, knack; habit' (in the phrase malas man~as 'bad habits')] at #297, leads to a reconstruction fraught with difficulties [*mana, *manaq, or *man~a 'inherit(ance)'] which could be the result of widespread Hispanic/Portuguese influence. A similar problem is encountered at #l33 (*gana, *ganas 'raw energy, animal appetite' < Sp gana 'desire, appetite', ?s 'plural') for at least the Philippine evidence??it is hard to see how Old Javanese could contain anything but inherited material, but difficult to determine which forms are legitimately cognate. (9) The last section of the introduction contains a brief forward look into Austronesian etymological studies. There is great promise in the forthcoming work of Blust and Anceaux, and in that of other scholars' endeavors to reconstruct lower?order proto languages. As more and more descriptions and dictionaries of AN languages become available, so shall "opportunities for uncovering more of the early Austronesian lexicon" (30), and "the culture?historical inferences that sometimes follow from them" (31). I would only add to this by expressing the hope that the Philippines, in the fullness of time, will develop an economic and academic climate that would support students and scholars in researching the immensely intriguing linguistic "Garden of Eden" available in the archipelago. (10) The list of language abbreviations and sources of material is excellent. Blust uses upper-case (capital) letters for languages, which facilitates the ready recognition of each new entry. Blust's abbreviations make a great deal of sense, but not all languages have been assigned a code, which prompts me to encourage a "meeting of Austronesian minds" in standardising these, see Zorc (1984) for suggested Philippine language abbreviations. Various considerations involve clarity, impact, economy (one, two, or three letters), and format (small or upper case). We may never achieve total conformity, but attempts could be made to find some common core of agreement, if possible. PART THREE. THE RECONSTRUCTIONS ?? DETAILS AND SUGGESTIONS. (1) DUBIOUS PHONEMES? Blust has waved the wand of the "independent evidence requirement" at many scholars, yet accepts *T on the basis of Javanese evidence alone. He even assumes its validity in cognate sets for which no Javanese forms are available (on the assumption that monosyllabic roots form part of the etymon, e.g., #154, 234, 337, 352, 434)??the assumption is legitimate on logical grounds, but the distinctness of *T must be held highly suspect. I have been sceptical of Central Philippine evidence supporting a *d/*D distinction (see Zorc l977:2ll?l6), because *d/*D/*j have fallen together, yet complex morphophonemics, loans, and re?analysis have resulted in irregular reflexes that have led different scholars (Dempwolff, Dyen, Dahl, Wolff, etc.) to widely?varying solutions. As Blust indicates, reflexes of *r are difficult to determine, but given Wolff's objections (1974), every possibility of identifying false cognates or an alternate etymon with *D or *R should be exhausted, as with #101 (revised as *deRung by Blust (in press a)), l69 (revised to *qingaR), 194 (Dsj *kahir could be dropped), 232 (*kuDis, with r forms as secondary or irregular developments), 302 (*naRa, with *r evidence interpreted as loans), etc. I find several of the reconstructions with *r quite solid (#43, 144, 145, 335, 365, 366, 368, 390), but some are questionable on semantic or formal grounds (70, 147, 158, 210, 223, 363, 373, 416). Following on from Wolff (1982) and my additional comments on the role that accent may have played in the development of irregular reflexes (Zorc l983), reservations might now be expressed over at least some of the following reconstructions with *c: #86, 87, 88, 89, 90, 91, 265, 347, 350, 354, 367. Similarly, several reconstructions with *g will need further evaluation: #21, 71, 115, 116, 128 ? 162, 250, 251, 356, 369, 388, 417. Some of these (e.g., 7, 132, 138, 139, 144, 146, 153, 355, 418, 427) may well serve to substantiate the validity of *g, despite Wolff's objections (1982). (4) NASAL CLUSTERS IN THE NORTHERN PHILIPPINE LANGUAGES? Reid (1982) demonstrated that genuine Ilokano and Bontok cognates of etyma reconstructed with *NC clusters have a simple consonant (viz: without a nasal). I have taken issue with the ensuing subgrouping hypothesis that NPh languages are closer to Formosan and CPh languages to PMP (Zorc 1984), but suggested that the development of nasal clusters in NPh languages (viz, the reconstructional implications of the hypothesis) needed further investi gation. Many of Blust's new etyma have NPh witnesses without a nasal reflex (#l, 93, 118, 172, 209, 242, 246, 258, 277, 337, 352, 436), which I interpret as supporting Reid's observations. It is therefore noteworthy that the following NPh etyma do have *NC clusters: (166) Ilk indang 'overseer', if there are no other Ph cognates, perhaps < Javanese? (173) Png kanding, but note Bon gelding 'goat' (311) Knk nangket 'glutinous', but Cas niket 'honey' (334) Ilk pantar 'treeless plain; shore' (?) (355) Note: Knk pongdaw, pongdol 'to pollard, cut the top off', Bon pongdol 'prune a tree' (388) Ilk sangga 'collar/flange at base of blade' (?) (411) Ilk tamban [sardine] (441) Cas dangkal, pos < Tag via Malay jengkal 'handspan' Few of these are of unimpeachable quality, such that they clearly support the continuation of *NC clusters into NPh languages. It might therefore be necessary to accept their inclusion within the respective cognate sets with some reservation.