An Analysis of the Impact of Reinforcement on Response Variability on Multiple Behavioural Dimensions
Kong, X. (2016). An Analysis of the Impact of Reinforcement on Response Variability on Multiple Behavioural Dimensions (Thesis, Doctor of Philosophy (PhD)). University of Waikato, Hamilton, New Zealand. Retrieved from https://hdl.handle.net/10289/10252
Permanent Research Commons link: https://hdl.handle.net/10289/10252
A series of four experiments were carried out to examine whether increases in variability in responding on some dimensions of a behaviour as a result of reinforcement generalised to other dimensions of the same behaviour. In Experiment 1, adult participants drew 300 rectangles on a computer screens; variability in the Area, Shape and Location dimensions of the rectangles drawn was measured. One group of participants received reinforcement when they drew rectangles varying on all three dimensions (VAR) using a threshold contingency to determine reinforcement while another group received reinforcement regardless whether they varied on any of the dimensions of the rectangles drawn (YOKE). Results showed that the variability of each dimension, as measured by the U-value, was higher for the VAR group than for the YOKE group. In Experiment 2, three groups of adult participants received reinforcement when they varied on two of the three dimensions of the rectangles drawn. Results showed that reinforcing variability in the Shape and Location dimensions and in the Area and Location dimensions resulted in higher variability in these dimensions compared to the dimension where variability was not directly reinforced, as measured by U-values. However, for the group who received reinforcement when varying in the Area and Shape dimensions, the variability across the three dimensions did not differ. Generalisation of reinforced variability was examined by comparing the variability in dimensions that were not subject to reinforcement for varying from Experiment 1 and Experiment 2. For the Shape and Location dimensions, results showed that the variability was higher when the two dimensions occurred with other dimensions that were subject for reinforcement for varying compared to when they occurred with other dimensions that were not required to vary. However, no significant difference was found for the Area dimension. Therefore, there appeared to be generalisation of reinforced variability for only two of these three groups of participants, suggesting the dimensions were not orthogonal. The non-orthogonality of the three dimensions of the rectangles was suggested as a possible confounding variable so that it not be concluded that generalisation of reinforced variability across dimensions had been obtained. Experiment 3 used a task with more independent dimensions (Shape, Colour and Pattern) to control for the non-orthogonality issue found in Experiment 2. Adult participants created objects by selecting elements from three dimensions on the computer screen. One group of participants received reinforcement when they varied on the use of elements on all three dimensions using a threshold contingency while the other group received reinforcement independent of the elements chosen. Results showed that the variability in using the elements for all three dimensions was similar between the two groups; whether the group received reinforcement when the dimensions varied did not make a difference. Inspection of the elements selected on each trial revealed that participants from both groups, in general, responded in stereotypic patterns; however, the stereotypy in the responses was not picked up by the measure of U-value. In Experiment 4, a new task asking participants to colour t-shirts on the computer screen was created to examine the effect of reinforcement on the variability in colours used. Adult and adolescent participants were instructed to colour t-shirts on the computer screen in three phases. In Phase 1, no feedback was provided; in Phase 2, positive feedback was provided when participants used a colour they had not used before (including those used in Phase 1); and in Phase 3, no feedback was provided. Results showed that, after reinforcement for using novel colours, participants who used only a small number of colours initially used more colours; however, the increase was not found for participants who initially used a moderate or high number of colours. U-value as the measure of variability was examined by using two sets of simulated data. It was found that highly stereotypic response patterns can result in extremely high U-values. Also, the number of options/categories used affects the values of U, which makes comparisons between groups difficult. Overall, results from these experiments support the notion that the level of variability in responses can be controlled by reinforcement. It is possible that learned variability can generalise over unreinforced dimensions. However, the examination of generalisation of reinforced variability across dimensions is not conclusive when the dimensions are interdependent. The results and simulation showed that U-value as a measure of variability needs to be used with caution because it does not capture potential stereotypy in responses and the values of U can be ambiguous.
University of Waikato
All items in Research Commons are provided for private study and research purposes and are protected by copyright with all rights reserved unless otherwise indicated.
- Higher Degree Theses