Introduction (1) The problem: explaining analogical changes within paradigms a. Paradigm leveling---e.g., loss of /iu/#/ie/ alternations in early NewHigh German (Paul, Wiehl, and Grosse 1989, `to fly' MHG Early NHG NHG 1sg vliuge fleuge fliege 2sg vliugest fleugst fliegst 3sg vliuget fleugt fliegt 2pl vlieget fliegt fliegt b. Analogical extension---e.g., loss of /i/#/ e/ alternations in early NewHigh German (ibid., `to give' MHG Early NHG NHG Following pattern of `to dig' 1sg gibe gibe gebe grabe 2sg gibest gibst gibst gr abst 3sg gibet gibt gibt gr abt 2pl g ebet gebt gebt grabt . Some forms in the paradigm are rebuilt to match/differ systematically from another form . That form is the pivot, or base, of the analogical change (2) Which form in the paradigm acts as the base/pivot? Numerous typological tendencies (Kuryl/owicz 1947; Ma nczak 1958; Bybee 1985; Hock 1991) . Bases tend to be the most frequent members of the paradigm (token frequency)