A form of learning (1) in which an organism choosing repeatedly between two alternatives A and B is initially rewarded for choosing alternative A rather than B until a behavioural preference for A is established, and is then rewarded for choosing B rather than A until the preference is reversed, and so on. Birds and mammals show progressive improvement in switching from A to B and back again, whereas fish show little or no improvement. Also called reversal learning. See also overtraining reversal effect, serial learning (2). Compare non-reversal shift.