Date of Award:


Document Type:


Degree Name:

Doctor of Philosophy (PhD)




J. Grayson Osborne


Characteristic distributions of errors across fixed ratio schedules of reinforcement were studied for two types of discrimination paradigms. Two experiments studied error patterns as a function of hypothesis behavior in two species of animals, children and pigeons.

Three key zero-delay matching-to-sample and two key simultaneous discrimination were reinforced for both species of animals on fixed ratio schedules of reinforcement. Experiment 1 involved children on matching-to-sample and simultaneous discrimination, and Experiment 2 involved pigeons on matching-to-sample and simultaneous discrimination. Both species of subjects experienced experimental conditions in which shift or stay response hypotheses were selectively reinforced using a high speed digital computer.

Data protocols were scored into four exhaustive error classes; winstay, lose - shift; win-shift, lose-stay; win - stay, lose-stay; and win-shift, lose-shift errors. These four error types were scored by frequency of occurrence and response latency for the ordinal positions of the fixed ratio. Two types of error patterns were defined for individual subjects. A standard error pattern was defined as having 15% more first half ratio errors than last half ratio errors. A reversed error pattern was defined as having 15% more last half ratio errors than first half ratio errors.

Experimental results indicated that selective reinforcement of particular response hypotheses produced only small effects for either species of animal on matching-to-sample or simultaneous discrimination. Response latencies for matching-to-sample and simultaneous discrimination were divided into two classes. The first class included long latency responses occurring immediately after reinforcement for children and pigeons. The second class included shorter latencies in the succeeding ordinal positions of the ratio for children and pigeons.

A majority of standard error patterns were produced when the total errors were separated into specific error types for low accuracy subjects of both species. The standard error pattern was lost for total errors due to a very high frequency of win-shift, lose-shift errors which were not distributed in any characteristic pattern. Higher accuracy subjects of both species tended to show a majority of reversed error patterns or no patterning when total errors were separated into error types. These subjects had very low frequencies of win-shift, lose-shift errors. The high frequency of win-shift, lose-shift errors in both species of animals across discrimination paradigms could be due to the complexity of the discrimination, a developmental age base for human subjects, or a 0-second intertrial interval.



Included in

Psychology Commons