My objective is to examine some aspects of patterns in normal word substitution errors for their bearing on claims for psychologically effective semantic fields. By “psychologically effective,” I mean to assert an active role in the normal function of a significant part of the communicative system. In so doing, I associate some empirical observations about a particular data base of speech errors with claims for specific processing mechanisms in language production. By way of preliminaries, the nature of the data base is first discussed, and second, the manner of my interpretation of it. Following those remarks, I report results of analyses of several classes of error and discuss their relation to some proposals that have been made regarding semantic fields.