Abstract
Undergraduates' button presses occasionally made available points that were exchangeable for money. Lights over left and right buttons were respectively correlated with multiple random-ratio random-interval components. During interruptions of the multiple schedule, students filled out sentence-completion guess sheets. When shaping of these guesses produced performance descriptions (e.g., "press slowly" for the left button and "press fast" for the right), button-pressing rates typically were consistent with the verbal behavior even when rates were opposite to those ordinarily maintained by the respective schedules. When shaping instead produced contingency descriptions (e.g., the button works "after a random number of presses" or "a random time since it worked before"), pressing rates were inconsistently related to the descriptions; for some students descriptions of ratio contingencies generated higher corresponding pressing rates than were produced by descriptions of interval contingencies, but for others contingency descriptions and pressing rates were unrelated.