Rats were trained to press a bar in a Skinner box and were then given food pellets independently of the bar presses. In one group of rats, a tone followed each bar press, and another occurrence of the same tone preceded the delivery of the food pellet. A second group had the tone following bar press but no tone preceding food pellets. The third group had tone before pellet but no tone following bar press. The fourth group had no tones at all. The bar pressing response was largely extinguished, but the rates of extinction did not differ among the groups. In reconditioning the third group showed more secondary reinforcing effects than groups 1, 2 and 4. The failure to demonstrate secondary reinforcement effects in group 1 is interpreted to mean that prediction of response outcome is a more potent factor than association with reward in a situation where there is a conflict between the two.