Comment: 3d class in Kryz. and Arch., 4th class in Rut. The word is actively contaminating with PL *ma[q:ʷ] 'song, tale' (q.v.), which explains some irregularities (e.g., loss of pharyngealization in Lezg. and Rut.). External evidence suggests that the PL form must have had a medial *-r-, but all the languages which have preserved the root frequently drop *-r- before uvulars; unfortunately no direct Tabasaran or Agul reflexes are present.