Roadblocks in the Grammaticalization Highway: When Phonology Gets in the Way

In lieu of an abstract, here is a brief excerpt:In light of recent proposals that question the unidirectionality of grammaticalization and the necessity of phonetic erosion (or more broadly autonomy), the third person pronominal suffix -nye in Besemah, a little-known Malay language of southwest Sumatra, presents an interesting puzzle. In the majority of constructions -nye attaches directly to the root and has undergone phonetic erosion (i.e., -nye > -e): anak-nye > anak-e ‘their child’. In deverbal noun constructions suffixed with -an, or circumfixed with peN- -an, however, the -nye is separated from the root and surfaces as a separate phonological word (i.e., -nye > anye): langkah-an-nye > langkah-an anye ‘his stride’ (from the root langkah ‘to step’). One might expect that the most phonetically robust and prosodically independent form (anye) is the oldest, but there is strong evidence going back to Proto-Austronesian that it is not. Therefore, what do we make of a case where a suffix by all other accounts is grammaticalizing, but in some constructions is phonetically strengthened and phonologically less bound (i.e., suffix > word)? Arguably, such a case should be added to the growing list of examples of degrammaticalization because -nye is apparently climbing up the cline. However, what is more interesting than this taxonomy is determining the conditions, pressures, and/or motivations by which this change emerged. It turns out that this unexpected alternation between –e, –nye, and anye can be attributed to a morphophonological (or more specifically morpho-prosodic) preference in Besemah for no more than one suffix per word. This preference alongside restrictions on phonological words and routinization has resulted in a number of unusual patterns, most importantly the alternations resulting from the suffix -nye attaching to other suffixes.