Effects of Non-reinforced Test Trials on Transposition

Ten male albino rats were trained In a brightness discrimination
problem where they were reQuired to choose a sttmulus value of
0.053 ft/cd over one of 0.012 ft/cd. Pairs were matched accordlng to
the number of trials requtred to reach a criterion of 18/20 correct
responses then randomly assigned to one of two grouos for testing In
transposition. Both groups were tested on the orlgtnally positive stimulus and a brighter one: 1.25 ft/cd for Group 8-C and 5.38 ft/cd
for Group 8-D. By testing for transposttlon wtth non-reinforced trials, contrary to the usual method, a tendency toward converging measures
of transposltton was achteved. Transposition for Group 8-D, In the
situation most dissimilar to training, was greater than for 8-C.
These results were discussed from relational or Gestalt, Spence model,
and Adaptation Level positions and It was shown that the results are
contrary to traditional Gestalt predictions. It was oredlcted that, according to underlying assumptions of
the Spence model, with continued non-reinforced trials, per cent of
transposition for both groups would decrease until a chance level of
responding was reached. That this did not occur cannot be explained
by the Spence model. Because the variability was too great with such a small N,
these results did not reach the .05 level of probability.