Comments: PAT *ʒ́ǝ-śa- (cf. also Bz. á-ʒ́-śa-ra). The morpheme is present only within a compound (see *śV-). It must be kept separate from *p-ʒǝ 'water' q.v.
Comments: PAT *ʒǝ (cf. also Bz. a-ʒǝ́); PAK *psǝ; Ub. def. a-bzǝ́. Shagirov (2, 16) suggests that PAT *ʒǝ should be separated from the AK and Ub. forms and prefers to compare them with PAT *p(ǝ)sǝ '*water,*river' (reconstructed on hydronymic evidence). There exists indeed a PAT root *pǝsa 'to become wet, soak' (e.g. Abaz. psa-ra id., Abkh. (a-ʒǝ́) a-psá-la-ra, Bz. (a-ʒǝ́) a-psǝ́-la-ra 'throw into water') which could be compared with PAK *psǝ ( - but not with voiced bz- in Ub.!). Still it is very hard to separate the AA root *ʒǝ from Ubykh (with which it correlates even in what concerns accent), and the Ubykh one from AK. We think that in PAK there could have occurred a contamination of two original roots: *bzǝ 'water' and *pǝsǝ 'river; liquid' - this would be the best explanation of the whole situation.
Abdokov (1983, 185) compares PAT *ʒǝ with PAK *p:c:a- in *p:c:a-n-tħá "glue" (Ad. pcāntħa, Kab. bʒāntħa). This would be fine in view of the EC data, but Shagirov (1, 81) is probably right in treating *p:c:a-n-tħá as "fish glue" (with *p:c:a "fish").
Comments: PAT *ʒʷa (with emphatic gemination). Phonetically the Ad.-Abkh. comparison is very probable; the semantic change must be postulated as "to pour water (enough) for washing" > "to cut off (water), to dam".