Schulze (email 8/1/02) 'Re: -VtsV and VstV-'. According to his data (lexical words 8000, 120,000 in texts), the sequence -VtsV- never occurs. The same holds for the glottalic variant Vt's and voiced variants -Vds- and VdzV-.

In general, Udi dislikes (homorganic) sequences that start with a stop and end with a fricative. Such sequences are usually metathesized (or avoided completely). One exception is the plural morpheme (oblique cases) -g^- and the Nizh absolutive/dative plural morpheme -xo that can be added to stems ending in a homorganic stop without further effects.

But note that non-homorganic clusters stop + fricative can occur...., see chapter 2.6.

Metathesis is some kind of automatic 'must'. Except for the forms mentioned above, there are no competing forms. Udi likes metathesis. This holds not only for the clusters mentioned above, but also for lexemes. A nice example is Nizh andar vs. Vartashen adamar 'man/person' (again stop goes into the second position!). My favorite is Udi e%l%m 'donkey' < h.elem < *h.emel < *h.imal < Arabic h.ima:r 'donkey'. But there is more the like that, however, calls for diachronic treatment.

Perceptual enhancement

Shifting the stop or affricate in metathesis from pre-consonantal position to postvocalic position enhances the perceptibility of the consonant given the presence of the vowel transition (Hume 1997, 1998).

The case involving the affricate and fricative is especially interesting. /c^s/ -> [sc^]. So what's bad about the input? You have an affricate with a fricative release precedinga fricative. This is like the cases where you get an /h/ following a fricative. It is also reminiscient of the constraint against sequences of two sibilants in English, leading to epenthesis, e.g. church + s -> churches. The release of the affricate wouldn't be very perceptible before a fricative. However, if you had a closure between the frications, then you'd get more of a distinction. Modulation yields more syntagmatic distinctiveness.