This research examined in 26 undergraduate students a form of instructional control of certain emergent relations definitional of the stimulus-equivalence relationship. The subjects read 2 pages that explained a paper-and-pencil match-to-sample procedure, and then went on to solve 11 more pages of matching-to-sample problems. Each of the first 10 of these last 11 pages was introduced by an instruction of the form , "Matches means [verb]," for example, "Matches means EATS," followed by the facts that established two related pages, 5 specifying a verb as the meaning of "matches" used equivalence verbs (EQUALS, IS, IS PARALLEL TO, GOES WITH, and MATCHES); another 5 specified nonequivalence verbs (EATS, OWES, PAYS, LIKES, and TEACHES). For 15 of 25 subjects, different verbs differentially controlled the emergence of the untrained relations revealing the symmetry, transitivity, and symmetric-transitivity properties of the original relations. For the remaining 10 subjects, these untrained relations either emerged uniformly despite verb differences (5 subjects), or were absent despite verb differences (5 subjects). The types of verbs or explanations provided by the subjects in response to the 11th problem, which offered no verb but requested an explanation of how the subjects had answered the probes, always reflected each subject's prior equivalence or nonequivalence responding to the earlier probes. Thus, imposing relevant instructions on two directly established interlocking conditional discriminations can account for much of the emergence of the new, untrained relations definitional of equivalence relations.Reprints of this article can be obtained from Jesus Rosales-Ruiz at Department of Behavior Analysis, University of North Texas, Denton, TX 76203. We are grateful to Richard