Arnold Chikobava suggests that the reason for stem reduction must have been a non-fixed intensive stress in the Old Georgian language. In comparing the Kartvelian languages he presumed that non-reduced stems attested in Megrelian and Svan must be primary relative to the Georgian reduced stems; compare Georgian ghrdzil-i with Megrelian dzhirgil-i and Tchan (Laz) dzhirdzhil-i (gum /in mouth/); and Georgian vashli with Megrelian ushkuri (apple).
Chikobava suggested that the week stress in Megrelian and Svan made it possible to preserve the original non-reduced stems in these languages, while non-fixed intensive stress in Georgian caused the reduction of these stems. Stress must have come at the end of a word in one group of stems (with reduction at the beginning of a word, as in *ghirdzil-i >  ghrdzil-i  (gum /in mouth/); and at the beginning of a word in another group of stems (with reduction at the end of a word: *vashali  > vashli /apple/).

An analogous process of reduction is evidenced in the modern Tchan (Laz) language where nonfixed intensive stress is not attested. The possibility that the decessive clusters are the result of such a reduction  (e. g. k'ibir-i > k'ibr-i /tooth/) is thought to explain the reduction (Togo  Gudava,  Zanuri (Megrul-Tchanuri) srulxmovnobis axsnis cda [An Attempt at Explaining Full Vocalisation in Zan (Megrelian-Tchan)], in: Sakartvelos mecnierebata ak'ademiis moambe, Tbilisi, 1950, vol.11, N7, p. 466);
Other assumptions: Vowels found along sonants