Blust "Resurrection of Proto-Philippiines"

  As seen already, in arguing for the separation of ‘Outer Philippines’ from other ‘Extra-Formosan’ languages Reid has fallen into a common logical trap, namely to attempt to draw positive conclusions from negative evidence, when no definite conclusion is possible without further information (Blust and Chen 2017).  Moreover, as noted in Blust (1996:141), Dempwolff’s ‘facultative nasal’ is far less likely to appear in Tagalog than in Toba Batak, Javanese, Malay, Ngaju Dayak or Malagasy.  As concluded there, the most straightforward interpretation of this pattern is that medial prenasalization is undergoing a process of loss in Tagalog.  While Tagalog belongs to the group of languages in which Reid (1982) believes medial prenasalization was innovated, comparative evidence shows that it (and presumably other Central Philippine languages) is reducing the frequency of such structures in the lexicon, and simply has not proceeded as far in this process as languages in more northerly parts of the Philippines.

  On closer inspection the argument that the languages of northern Luzon belong to a primary branch of Reid’s ‘Extra-Formosan’ based on the asserted absense of medial prenasalization in native forms is thus not well-supported.  What, then, is the position of the languages of northern Luzon?

  As can be seen from Appendix 1, the evidence for linking the Northern Luzon languages with other Philippine languages is far more extensive than the evidence for inclusion of Bilic within a larger Philippine group.  Some of the relevant comparisons have already been shown in Table 4, where 21 of the 30 comparisons link Northern Luzon with Bilic, 16 link Northern Luzon with GCP, and smaller numbers link Northern Luzon with both Kalamianic and Batanic languages.  In Appendix 1 the number of comparisons that include both Northern Luzon and GCP languages far exceeds any other combination of microgroups, appearing in over 74% of the cognate sets cited there (905 of 1,222).  One reason for this is the rich body of available evidence for each of these groups, as seen in Table 5, which lists all languages with at least 1,000 citations in the ACD on June 21, 2019 (Northern Luzon languages above the dotted line, GCP languages below it):


Blust "Doubletting"

2.4.  The ‘facultative nasal’.
   There is one other restriction on the definition of doublets that must be mentioned before proceeding.  Many entries in Dempwolff (1938) contain either simple or prenasalized medial obstruents, as *mata‘ ‘eye’, or *punti‘ ‘banana’, but   in others the evidence of medial prenasalization is contradictory.  Thus, based on reflexes of *tubuh in Tagalog, Toba Batak and Javanese, of *tumbuh in Ngaju Dayak, Fijian, Sa’a and the Polynesian languages, and of both in Malay (tubuh ‘body’, tumbuh ‘grow’), and Malagasy (za)tuvu ‘youth, lad’, tumbu ‘excess’, mi-tumbu ‘grow, increase’) Dempwolff posited 
*tu(m)buh ‘prosper, grow’.  He called the parenthetic segment a ‘facultative nasal’, as opposed to the nasal in e.g. *punti‘ ‘banana’, which appears in all diagnostic witnesses.  It is important to understand that Dempwolff used this notation to mark contradictory evidence (ambiguities are marked with square brackets).  In principle, then, all proto-forms with a facultative medial nasal could be written as doublets (in modern orthography *tubuq/tumbuq, etc.).

  As noted in Blust (1996) the facultative nasal in AN languages poses a serious challenge to the Neogrammarian hypothesis, since it cannot readily be accounted for as a product of morphology, conditioned sound change, borrowing or a change in progress.  With respect to the present argument this variable feature creates many allofams that differ only in the presence or absence of this element.  Since what I call ‘true doublets’ vary in many ways and show far less obvious patterning, I adopt a conservative position for cases like *tubuq/tumbuq, and treat the longer form as a product of convergent nasal insertion.


Second, although variation is found both in attested languages and in proto-languages the record to date suggests that it is more common in the latter, a pattern which implies that there is a flaw in the methodology used to reconstruct proto-languages.  But if there is a flaw, what is it?  I have taken what some might consider extreme precautions to avoid inflating the evidence for doubletting by 1. requiring that doublets be phonologically similar but non-identical forms that are found in the same language, 2. by separating disjuncts from doublets, 3. by distinguishing true doublets from co-morphemes (bases that contain a common submorphemic root), and 4. by disallowing doublets that are distinguished only by a ‘facultative nasal’.  Despite this tightening of conditions on the definition of ‘doublet’ I have found evidence for nearly 1,000 allofams in a comparative dictionary that to date probably is no more than 25% complete.  A few proto-doublets may reflect convergent sporadic change in widely separated languages that produces the false appearance of a cognate set having a shape different from its actual prototype, but this could hardly explain all, or even a significant part of the data.