The topic of this discussion is analogical change of a particular sort, namely that where, pretheoretically, meaning seems to have had a role to play in determining the direction taken by the change. Analogy is an area of linguistic change of the highest theoretical interest. No analogical changes can take place without the involvement of ‘performance’ factors in general, including a perceiving human agent as one of these (cf. Vincent, 1974; Anttila, 1977), for many analogical acts illustrate the human mind in a condition that is at the same time relatively free and relatively fettered. It is free in the sense that the inferential processes used in the cases that interest us are ABDUCTIVE (cf. Andersen, 1973: 775; but contrast Mayerthaler, 1980: 126–7); that is, given some facts, a law may be invoked which allows the reasoner to infer that something MAY, not MUST, be the case, and to act upon that possibility in a creative way. The basic data in this paper include instances of similarity between lexemes, a notion which will receive the closest scrutiny. In the changes we will examine, people appear to invoke a law-like principle to the effect that there is a close relation between similarity of form and similarity of meaning. And they infer, abductively, that cases of similarity in such instances may be failed examples of sameness, and accordingly may replace the state of similarity by one of sameness (cf. Panagl, 1982: 5). The result is therefore the fruit of not always unreasonable guesswork rather than of the logical security of deduction (or induction). The mind is fettered in the sense that it appears condemned to try to make something transparent, or at least formally transparent, out of what is linguistically obscure; its activity is thus constrained by the rest of the linguistic system with which it operates. It is in the condition of an electrician who is summoned to sort out a dangling wire and connects it up to the first other dangling wire that he or she finds.