4.1.9, 4.1.12, 4.1.13 - Vicarious reinforcement and development

Acquisition is not performance

Acquiring information about a behaviour and choosing to perform it are distinct outcomes.

Acquisition means learning the action or its representation. Performance means displaying it in a particular situation. A child can remember how an adult acted while deciding that copying would bring disapproval. Another child can describe a movement but lack the physical skill to reproduce it. Observed behaviour is evidence from which learning is inferred, not a perfect direct measure of everything stored in memory.

The observer need not receive the model’s reward or punishment. Watching a peer’s careful contribution praised can change expectations about contributing; watching disruption sanctioned can make an observer reluctant to disrupt. A direct consequence instead follows the observer’s own action.

This distinction challenges a simplistic interpretation of reinforcement: “If nobody rewarded the observer, nothing could have been learned.” Social learning permits acquisition through attention and retention without an immediate reward for copying. Consequences may strongly affect whether the behaviour is expressed. They can also influence attention and future opportunities to learn, so it would be equally simplistic to claim that motivation never affects acquisition in any situation.

For example, a fictional learner watches a complicated trick that earns a model a reprimand. The learner can retain the steps but refrain from using them. A later safe demonstration showing accurate knowledge would reveal something that spontaneous behaviour alone concealed. The research question is therefore whether differing observed consequences affect knowledge of the action, willingness to display it, or both.

Bandura 1965: design

Bandura manipulated the consequences to an aggressive film model and then changed the incentives offered directly to the observing children.

The aim was to test whether consequences administered to a model affected the acquisition of matching responses or primarily their performance. The sample was 66 children from Stanford University Nursery School, 33 boys and 33 girls, aged 42–71 months. Children were randomly assigned to three conditions, each with 11 boys and 11 girls. The available nursery population limits representativeness; random assignment to conditions is not random sampling from all children.

Each child watched an approximately five-minute film individually. A male model performed distinctive physical and verbal aggression towards a Bobo doll. The central behaviour was the same across conditions; the ending varied. In the model-rewarded condition, another adult praised the model and supplied treats. In the model-punished condition, the adult reprimanded and physically punished the model. In the no-consequences condition, neither ending was shown. No-consequences did not mean that children saw no model or no aggression.

After exposure, children entered a playroom with a Bobo doll and other toys. During a ten-minute test, observers behind a one-way mirror recorded matching responses in specified categories. This first phase measured spontaneous performance without an offered reward for copying. The observers did not know the assigned film condition, reducing expectation-based scoring bias.

Then all children were offered the same direct positive incentives: sticker pictures and juice for reproducing the model’s physical and verbal actions. The experimenter asked them to show what the model did and report what he said. This second phase sought to reveal acquired responses that the first phase might not elicit. The study therefore combined an independent-groups comparison of model consequences with repeated observation of children before and during direct incentives.

This is the 1965 experiment, not the 1961 aggressive/nonaggressive/no-model comparison or the 1963 live/human-film/cartoon comparison. Keeping the manipulated variable clear is necessary to understand what each study can conclude.

Bandura 1965: findings and inference

The effect of observed punishment depended on whether children were being tested for spontaneous performance or offered incentives to reproduce what they had seen.

In the first phase, children in the model-punished condition produced fewer different matching responses than children in the rewarded and no-consequences conditions. The rewarded and no-consequences groups did not differ significantly. This does not support the claim that a model must be rewarded for observers to imitate it. It does support an inhibitory effect of observed punishment in this setting.

When direct positive incentives were introduced, the earlier differences between the three model-consequence conditions disappeared statistically. Children who had copied relatively little could now reproduce modelled acts. Bandura interpreted this as evidence that observed consequences had affected performance more than acquisition: withholding an action was not equivalent to failing to learn it. Children had acquired more than their initial spontaneous behaviour revealed.