This experiment examined the retention of generalized equivalence classes.
Seventeen subjects completed matching-to-sample training, in which two-elem
ent stimulus compounds were presented as sample stimuli, and an equivalence
test, in which those compounds were separated and each element was tested
separately for its entry into equivalence relations with the other stimuli.
A generalization test was then conducted, in which dimensional variants (a
long the dimension of hue) of one member of each class were presented as sa
mple stimuli on test trials for two generalized symmetry relations and one
generalized equivalence relation. Equivalence and generalization retention
tests were conducted for 8 subjects 2-3 months following their initial labo
ratory session. This experiment also employed a verbal protocol analysis, i
n which subjects' overt verbal behavior was recorded during both experiment
al sessions and examined in light of subjects' performances. More subjects
demonstrated the retention of symmetry relations than within-compound or eq
uivalence relations, and all subjects who showed the maintenance of equival
ence and within-compound relations also showed the maintenance of symmetry
relations. Generalized classes were shown to have remained stable over time
for some subjects, while other subjects demonstrated the emergence of subj
ect-defined generalized classes during the generalization retention test. V
erbal protocol analyses showed that all subjects emitted a relatively great
er frequency of verbal descriptions of stimulus relations under extinction
testing situations than during training, and class-consistent names for gen
eralized class-members appeared to be likely, but not necessary for the gen
eralization of class membership.