Electric-shock punishment produces stimulus-generalization gradients w
ith animals. The purpose of the present experiments was to determine w
hether response-cost punishment produces similar gradients with humans
. In Experiment 1, psychophysical procedures were used to generate a s
et of 10 horizontal lines that differ in length such that adjacent lin
es were indiscriminable, then reinforcement gradients were obtained wi
th those lines to show that gradients can be produced with those stimu
li. In Experiment 2, subjects pressed a lever for points exchangeable
for money on a VI schedule in the presence of each stimulus, then poin
ts were lost immediately after each response in the presence of Stimul
us 6 (the middle value). Response rate decreased in the presence of St
imulus 6, but it had a similar increase in the presence of all other s
timuli. Those results suggest that point loss and not line length was
the discriminative stimulus for further point loss. To minimize stimul
us control exerted by immediate point loss, and thus enhance control b
y line length, in Experiment 3 points were lost only at the end of the
component correlated with Stimulus 6, and in Experiment 4 point loss
was arranged on VR and VI schedules. Neither delayed nor intermittent
point loss produced gradients: Unpunished response rate was similar in
the presence of all line lengths. The present results suggest that th
e technique of reinforcing responding in the presence of several stimu
li during baseline then punishing responding in the presence of only o
ne of those stimuli does not produce with humans and response-cost pun
ishment the gradients obtained with pigeons and electric-shock punishm
ent.