This experiment examined whether association with a common probability of reinforcement (either .1 or .9) would cause pigeon subjects to treat dissimilar stimuli the same in later test sessions. The stimuli were from four basic-level conceptual categories: people, flowers, cars, and chairs.
Twenty-four photographs from each category made up the training set. The four categories will be referred to as C1, C2, C3, and C4. In original training, C1 and C2 were followed by reinforcement with a probability of .1 and C3 and C4 were followed by reinforcement with a probability of .9.
In reassignment training, choice trials were given only with C1 and C3. After the pigeons had pecked the stimuli, two colored “buttons” appeared on the computer screen below the pictures. Choice of the left was reinforced for C1 and choice of the right was reinforced for C3.
In testing sessions, twenty-four novel stimuli per category were introduced. Choice trials were given with familiar stimuli from C1 and C3, familiar stimuli from C2 and C4, novel stimuli from C1, C2, C3, C4.
The subjects responded correctly to the familiar C1 and C3 stimuli, scoring 90.28%, and to the novels, scoring 74.24%, even though they had never been trained to do the latter. Likewise they responded correctly to the familiar C2 and C4 stimuli at 87.50% and to the novels at 77.08%, even though these stimuli had never before appeared on choice trials. These results show that association with a differential probability of reward can induce pigeons to treat categories of stimuli the same on later test sessions.
Tracy Garrett, ’98
Sponsor: Suzette Astley