The existence of a direct value transfer between stimuli in the contex
t of instrumental discrimination learning is demonstrated. In a first
experiment pigeons learned to discriminate instrumentally four success
ively presented target stimuli. Pecks to them were reinforced with gra
ded amounts of reward and penalty (A++, B+, C-, D- -). These stimuli w
ere accompanied by four different neutral stimuli (N-a, N-b, N-c, N-d)
but responses to these had no consequences. After discrimination of t
he target stimuli had been established the neutral stimuli were presen
ted in pairs and in the absence of the target stimuli. These tests rev
ealed a graded preference in accordance with the value transfer hypoth
esis. In a second experiment pigeons were taught to discriminate two t
arget stimuli A+ and B-. Each of these was again accompanied by a neut
ral stimulus (N-a, N-b). Pigeons developed a strong preference for N-a
. A discrimination reversal affecting the target stimuli (B+, A-) led
to a strong reversed preference for N-b. A third experiment employed a
similar design but in it one group of pigeons, unlike the control gro
up, was prevented from pecking the neutral stimuli by transparent barr
iers. As both groups preferred the reward-related neutral stimulus wit
h approximately equal strength it is concluded that classical conditio
ning, rather than pseudo-operant conditioning was responsible for the
value transfer observed. A modification of an algebraic instrumental c
onditioning model incorporating an adventitious classical conditioning
element is shown to emulate the value transfer obtained. The role of
classical conditioning in instrumental discrimination learning is disc
ussed.