Transmission of Aggression through imitation of aggressive models

Albert Bandura, Dorothy Ross and Sheila Ross (1961)

 

Theory

consider the limitations of behaviourism and how watching somebody else can be a good way of learning. This is known as "modelling", and would be a good way of describing how, for example, language is learnt.

Very brief summary.  Full version here

Both male and female children were used. They witnessed an adult, who was either male or female, behave in either an aggressive or passive way. Other children did not witness either (the control group) The experimenters made sure that children who were more aggressive before the study started, were evenly spread amongst the groups. Having witnessed the behaviour of the adult the child was then frustrated by being denied access to interesting toys. The child’s level of aggression was then recorded. This aggression could be an exact copy of what the adult had done or the child could make up interesting and original ways to be aggressive.

 

Hypotheses

  1. Participants exposed to aggressive models will reproduce aggressive acts resembling those of the models to a significantly greater extent than both participants exposed to non-aggressive models and those not exposed to any models at all (control group); (This is the major, overall, hypothesis.)
  2. Participants exposed to subdued, non-aggressive models will display significantly less imitative aggression than control group participants will.
  3. Participants will imitate the behavior of same-sex models to a greater degree than opposite-sex models.
  4. Boys will be more likely to imitate aggression than girls, especially when they are exposed to aggressive male models.

Method

Subjects

36 boys and 36 girls. Stanford University Nursery School. 3 - 5 years old.

Experimental Design

Independent groups matched on aggression (51 potential experimental subjects observed independently; Observer's ratings 0.89 using Pearson product-moment, p is less than 0.0005). 8 experimental groups of 6 subjects, also a control group of 24. Aggressive / Non-Aggressive model x sex of subject x sex of role model (2 x 2 x 2 = 8). Why match on aggression?


Experimental Conditions

Children exposed to either aggressive or non-aggressive condition individually. Child occupied with potato printing. Experimenter led model to opposite corner of room. Non-aggressive condition, model plays quietly with tinker-toy set, ignoring Bobo doll. Aggressive condition, model plays quietly for first minute, then hits Bobo repeatedly, in a prescribed fashion. The model also uses verbal (scripted) aggression. Both conditions last 10 minutes. at the end of the experimental condition the experimenter bids the goodbye to the model; this implies that the experimenter is supporting the Models behaviour. This is important because later versions of this experiment demonstrated that vicarious modelling can effectively facilitate imitation. Vicarious modelling means observing the model being either rewarded or punished for their behaviour.

Aggression arousal.

After being taken to another room, child is aggressively aroused by the experimenter. This was achieved by allowing the child to become interested in toys, and then, after two minutes, stopping them. It was explained that they are not for that child to play with, but are to be kept for the use of others.

There are two reasons why 'aggression arousal' was used:

  1. Rosenbaum & deCharms (1960) found that immediately after observing aggressive behaviour, subjects were less likely to be aggressive, which, in this experiment, means that children in the 'aggression' condition are less likely to be aggressive than the children in either of the other two conditions.
  2. The arousal gives an opportunity for the 'non-aggressive' children to demonstrate the anticipated inhibiting effect.

Test for delayed imitation

Tested in yet another room. An adult remains in the room with the child ostensibly quietly dealing with some paperwork. This is to prevent the child from leaving the room prematurely. Toys, half aggressive (e.g. dart gun) and half non-aggressive are always placed in a specific order. Observed by two independent observers who were supposed to have no knowledge of group assignment, for half of the trials, and by one observer for the rest. One of the observers was the male model. Even though he did not observe any child he had modelled for, he knew that the child had been exposed to a female model, or had been a control. The observers are behind a one-way mirror.

Observations carried out by counting the number of time units (5 seconds) in which aggressive or non-aggressive behaviour had occurred. Children observed for 20 minutes. High inter-scorer reliability.

Response Measures

  1. Imitation of Physical Aggression
  2. Imitative Verbal aggression
  3. Imitative non-aggressive verbal responses
  4. Mallet aggression
  5. Sits on Bobo doll
  6. Punches Bobo doll
  7. Non-imitative physical and verbal aggression
  8. Aggressive Gun Play

Results

Effect of same sex model: The male model influences the boys more than the girls and vice versa. Same sex imitation has been explained as the result of reinforcement in the home for sex appropriate behaviour by parents. So children find it more rewarding to copy same sex parents (Fauls and Smith, 1956).Aggressive models cause the children to be more aggressive, and this extends to behaviour that was not specifically modelled (non-imitative aggression). Non-aggressive models cause the children to exhibit less mallet aggression and sitting on Bobo, compared with the controls. 'Aggressive gun play' and 'punching Bobo' is not affected by modelling. Boys were found to be more aggressive than girls. Aggression is expected more from boys, so this would explain why boys are more aggressive; simply because they are reinforced in the home and elsewhere for aggressive behaviour.

Not only were specific violent acts imitated but also the children generalised their violence, for example, playing aggressively with plastic farm animals.

Because of the small sample size (6 in each condition) it is not possible to tell whether the results are significant between specific conditions. By lumping together boys and girls, this gives 4 conditions with 12 subjects in each; This enables significance to be detected more easily.

Qualitative data is given in the form of quotes from the children who comment on the misappropriateness of the aggressive female's actions.

Discussion

Skinner (1953) behaviour has to be reinforced to be learnt; this is not the case here. Bandura developed his Social Learning Theory, which emphasises the importance of the imitation of suitable role models. You may wish to consider which models would be most influential for particular children. Also consider how the child would be rewarded or reinforced for copying a model's behaviour. Bandura explains that the children's behaviour could be explained by using Freud's idea of identifying with the aggressor.

Evaluation

Theoretical Issues

The social learning theory explanation of aggression contrasts with:

  1. Instinct theories, for example Freud's Catharsis(meaning that aggression is necessary and has to be released from time to time perhaps in a sublimated form such as sport) and Lorenz's evolutionary explanation.
  2. Drive Theories, for example Dollard et al's frustration-aggression hypothesis.

Modelling can be either direct or vicarious. There are six regulatory mechanisms that are applicable to vicarious reinforcement (Bandura 1977):

  1. informative
  2. motivational
  3. emotional learning
  4. influenceability, which is affected by the Models response to reinforcement; whether the observer likes the reward or not.
  5. Modification of Models status, high if rewarded, low if not.
  6. Evaluation, unjust punishment will lessen the effect.

Subsequent Research

Bandura Ross and Ross (1963) had four-year-olds watch a film of Johnny and Rocky (adults). Johnny had attractive toys. In one condition Rocky aggresses and wins the toys and in the other condition Rocky aggresses and is beaten up by Johnny. Not surprisingly the children in the first condition, where Rocky is rewarded, imitate the aggression.

Bandura (1965) had three conditions:

  1. model rewarded
  2. model punished
  3. model neither rewarded nor punished

Children in condition two did not copy, but did copy when later on they were offered rewards. Therefore behaviour can be produced sometime after modelling and behaviour is learnt even when the model is seen to be punished. This experiment demonstrated vicarious reinforcement.

Our behaviour according to Bandura (1986) is itself regulating, that is self-governing the by self-evaluation. There are three stages of self regulation:

  1. self observation
  2. Judgemental process or personal standard, based on reactions of others and observation of respected Models
  3. self response processes, satisfaction, pride upon reaching or exceeding one's personal standard or self dissatisfaction, self criticism, for not achieving one's personal standard.

Bandura and Kupers (1964) had children observe a model rewarding himself with sweets for either a high or low standard of playing bowling. Children adopted similar standards to the level of reward that they had observed.

Methodology

  1. The films in later experiments were only three and half minutes long and lacked plots or justification for the violence portrayed. The experiment thus lacks ecological validity.
  2. Children who had not seen a Bobo doll before were five times as likely to aggress towards it (Cumberbatch, 1995). Further, children said that they felt they were expected to perform aggressively towards the doll. The experiment suffered from demand characteristics, but having said that the specific nature of the imitative aggression and the generalised aggression to other toys would suggest that the modelling was effective.
  3. Were there any long-term effects? Hicks (1965) found that 40% of the Model's acts could be reproduced six to 8 months after one showing of a film lasting under 10 minutes.
  4. Aggression? Or just play? Would the child aggress towards another human? Support for this view comes from Johnson et al. (1977) who found that play aggression correlates with ratings of aggression by peers (0.76) and teachers (0.57). Note that this shows why it was important for Bandura to control for rated aggression.
  5. The experiment was conducted in a permissive setting which was not ecologically valid.

Applications and implications of social learning theory

copy cat hijackings. Air hijackings were unknown in the US prior to 1961. Then some Cuban airlines planes were hijacked which sparked off a wave of hijackings culminating in a peak of 87 hijackings in 1969 (Mischel, 1986).

Phobias. Observing somebody else being scared of something is enough to start a phobia. Vicarious modelling can be used to remove phobias.

Read the extract from Banyard and answer the questions.

  1. How many subjects were there and what was their mean age?
  2. What were the 3 IVs?
  3. Why was it necessary to distribute the aggressive children evenly throughout the groups?
  4. To find out which children were the most aggressive they pre-tested them. How did they do this?
  5. What is the problem of rating a child’s behaviour in this way and how did they attempt to overcome this?
  6. They achieved a high positive correlation. What would this look like on a scattergraph? What does the high positive correlation mean in this case?
  7. Summarise stage 1 and 2 of the procedure.
  8. During stage 3 children were taken to another room where their aggressive behaviour was observed for 20 minutes through a one way mirror. The experimenter recorded what the child was doing every 5 seconds. Why was a one-way mirror used? Why did a researcher remain in the room? What problem might this have caused for the results?
  9. Behaviour was recorded under a number of headings. Give 3 examples of imitative behaviour, 1 example of incomplete imitation and 1 example of non-imitations.

Return to Gary Sturt's Homepage