trial problems to choose between two targets differing in odor, one of them providing 10 µl of
a 50% sucrose solution and the other 10 µl of water. In 9 of the problems, two odors were
used, and the reward ratio was varied systematically over a wide range. In the 10th problem,
three odors were used in an ambiguous-cue (A+/B−, B+/C−) design. The results were
predicted quantitatively, and with substantial accuracy, from a simple theory of learning and …