This leaves us in methodological limbo. It would be helpful if evidence of thesedistinctions could be found in more distant languages. Javanese does not appear to be closely related to Malay, Sundanese, or Madurese, so the distribution of this comparative feature is geographically compact but genetically disparate. To raise further suspicions, the distinction is not found in the closest relatives of Malay that lie outside the ‘diffusion zone’ of Malay influence in western Indonesia, namely the Chamic languages (Thurgood 1999:124ff). It might appear that the reconstruction of *-ey and *-ew presents problems similar to those for e.g. *c, but there are important differences. Whereas *c fills a structural gap in a system that must be reconstructed with a corresponding voiced palatal stop and nasal, *-ey and *-ew fill no gaps, but simply elaborate a system of diphthongs that is symmetrical and intrinsically complete without them. Similarly, although *c is consistently distinguished from *s even in submorphemic roots, *-ey and *-ew do not Figure in any known root (Blust 1988a). Until a better basis can be discovered for assigning *-ey or *-ew to PAN or any other early proto language, then, it seems best to treat the correspondences in question as simply unexplained.