Section A: Understanding the study
What the children did
How to read the numbers
Each number on this page is an average score for a group of children. A child's score for a category is the number of five-second intervals, out of 240 in the 20-minute test, in which the observers saw that kind of behaviour. So a score of 25.8 means that, on average, children in that group did that kind of act in about 26 of the 240 intervals.
The groups are small. Each aggressive and non-aggressive group had six children. The control group had 24 children, which works out at 12 boys and 12 girls. With so few children, one very active child can move a group's average a lot. Keep this in mind whenever two numbers look different.
You will see the word "significant" on this page. A difference is statistically significant when it is large enough, given the size of the groups, that it would rarely happen by chance alone. It does not mean large or important. With groups of six children, only quite big differences reach this standard.
Finding 1: children copied the aggressive model's exact acts
This was the main result, and it was very strong. Children who had watched the aggressive model sat on the Bobo doll and punched its nose, hit it with the mallet, kicked it and threw it in the air. They shouted "Pow!" and "Sock him!" Children who had watched the calm model, and children who had seen no model, almost never did these particular things.
Copied physical aggression
Average number of five-second intervals, out of 240, with an exact copy of the model's physical acts
- Girls
- Boys
Source: Bandura, Ross and Ross (1961), Table 1. Six children per bar, except the control bars (12 each).
Show these numbers as a table
| Condition | Girls | Boys |
|---|---|---|
| Aggressive female model | 5.5 | 12.4 |
| Aggressive male model | 7.2 | 25.8 |
| Calm female model | 2.5 | 0.2 |
| Calm male model | 0.0 | 1.5 |
| No model (control) | 1.2 | 2.0 |
Compare the two "Aggressive" groups with the three other groups. That gap is the study's central finding. The calm-model and no-model children scored close to zero. About 70% of children in those conditions had a score of exactly zero.
The same pattern appeared for the model's words.
Copied verbal aggression
Average number of five-second intervals, out of 240, in which the child repeated the model's aggressive phrases
- Girls
- Boys
Source: Bandura, Ross and Ross (1961), Table 1. Six children per bar, except the control bars (12 each).
Show these numbers as a table
| Condition | Girls | Boys |
|---|---|---|
| Aggressive female model | 13.7 | 4.3 |
| Aggressive male model | 2.0 | 12.7 |
| Calm female model | 0.3 | 1.1 |
| Calm male model | 0.0 | 0.0 |
| No model (control) | 0.7 | 1.7 |
Notice something here that Finding 4 explains: the girls copied the female model's words most, and the boys copied the male model's words most.
The children copied more than aggression. About one third of the children who watched the aggressive model also repeated its non-aggressive comments, "He keeps coming back for more" or "He sure is a tough fella". No child in the other groups said these things. Copying a harmless phrase that only the model had said is strong evidence that the children were reproducing what they had observed, not just becoming more excited or aggressive. And all but two of the 24 children in the aggressive condition performed at least some of the model's unusual aggressive acts.
Statistics note
The researchers could not use ordinary analysis of variance, because so many children scored zero and the scores were very unevenly spread. They used the Friedman test, a test based on ranks that suits matched triplets. The overall effect of condition was significant for copied physical aggression (χ²r = 27.17, p < .001) and copied verbal aggression (χ²r = 9.17, p < .02). Pairwise sign tests showed that the aggressive group differed from both the calm and control groups (physical: both p < .001), while the calm and control groups did not differ significantly from each other. Copying the non-aggressive phrases was tested with Cochran's Q test (Q = 17.50, p < .001).
Finding 2: watching aggression also raised other aggression, but not every kind
Some children copied part of the model's behaviour. They used the mallet to hit things other than the doll, or laid the doll down and sat on it without hitting it. These partial imitation scores were also highest in the aggressive condition.
The model's example also spread to some acts the model never performed. Children in the aggressive condition showed more non-imitative aggression, such as other hitting and hostile remarks like "Shoot the Bobo" or "Cut him", than children who watched the calm model.
But two kinds of aggression did not change with condition: ordinary punching of the doll, and aggressive gun play. Children in every group punched the doll and played at shooting. These are acts young children do without being shown, which is exactly why the researchers had built their main test around unusual acts.
Statistics note
For mallet aggression the overall effect was significant (χ²r = 11.06, p < .01). The aggressive and control groups both did more than the calm-model group, but the aggressive group (mean 20.0) did not differ significantly from the controls (13.3). For non-imitative physical and verbal aggression the overall effect was significant (χ²r = 8.96, p < .02), with the difference coming mainly from aggressive versus calm groups (p = .026). Punching the doll (χ²r = 2.87) and gun play (χ²r = 2.75) showed no significant effect of condition.
Finding 3: the calm male model reduced aggression, the calm female model did not
Hypothesis 2 predicted that children who watched a calm model would be less aggressive than children who saw no model. Taken as a whole, the calm and control groups did not differ significantly, apart from the control children using the mallet more.
When the researchers looked at the two calm models separately, a clearer picture appeared. Children who had watched the calm female model behaved like the control children on every measure. Children who had watched the calm male model were less aggressive than the controls on several measures: less copied physical and verbal aggression, less mallet aggression, less other aggression, and less punching of the doll.
The calm models also had a broader effect. Children who had watched a calm model spent more than twice as long as children in the aggressive condition simply sitting quietly, not playing with anything. They also played more with the dolls. The authors concluded that watching a subdued model "not only decreases the probability of occurrence of aggressive behavior but also generally restricts the range of behavior" (p. 581).
How strong is this evidence?
The comparisons between the calm male model group and the controls used sign tests on very small groups, and some p-values are close to the usual cut-off (for example p = .06 for copied physical aggression and p = .07 for punching the doll). The authors admit that comparing subgroups when the overall test is not significant "is likely to capitalize on chance differences", but argue that the consistent direction across measures supports their reading (p. 579). Treat this finding as suggestive, not firm.
Finding 4: the sex of the child and of the model mattered, in an unexpected way
Boys copied more physical aggression than girls, but not more words. Among children who watched an aggressive model, boys copied significantly more of the model's physical aggression than girls did. Boys and girls did not differ in copying the model's aggressive words. So hypothesis 4 was only partly confirmed, as the paper itself says.
The male model had more influence overall. For physical aggression, both boys and girls copied the male model more than the female model. Boys who watched the aggressive man scored highest of all. For verbal aggression, each sex copied the same-sex model more: girls copied the woman's words, and boys copied the man's. So hypothesis 3, same-sex copying, held for words but not for physical aggression.
The authors explained this through sex-typing. In their society, physical aggression was strongly linked with men, so a man's aggression looked "appropriate" and was copied by everyone. Verbal aggression was less tied to one sex, so children copied the adult of their own sex. The children's own comments support this reading.
"Who is that lady. That's not the way for a lady to behave. Ladies are supposed to act like ladies..."
"You should have seen what that girl did in there. She was just acting like a man. I never saw a girl act like that before. She was punching and fighting but no swearing."
Children's comments about the female model, Bandura, Ross and Ross (1961), p. 581
"Al's a good socker, he beat up Bobo. I want to sock like Al."
"That man is a strong fighter, he punched and punched and he could hit Bobo right down to the floor and if Bobo got up he said, 'Punch your nose.' He's a good fighter like Daddy."
Children's comments about the male model, Bandura, Ross and Ross (1961), p. 581
The authors noted their own limit here: there was only one man and one woman acting as models. Any difference between "the male model" and "the female model" might come from these two individuals' personalities rather than their sex (p. 581).
Statistics note
Sex differences were tested with t tests within the aggressive condition only. Boys copied more physical aggression than girls (t = 2.50, p < .01). With the male model, boys showed more copied physical aggression (t = 2.07, p < .05), copied verbal aggression (t = 2.51, p < .05), non-imitative aggression (t = 3.15, p < .025) and gun play (t = 2.12, p < .05) than girls. With the female model, girls showed more copied verbal and non-imitative aggression than boys, but these differences were not significant. Because the main tests were non-parametric, no overall test of the sex of the model, or of its interaction with the sex of the child, was possible.
The four hypotheses: what happened
| Hypothesis | Result | What it means |
|---|---|---|
| 1. Watching aggression produces copied aggression | Strongly supported | Children reproduced the model's unusual acts and words; the other groups almost never did |
| 2. Watching a calm model reduces aggression | Partly supported | Only the calm male model reduced aggression compared with the control group |
| 3. Children copy the same-sex model more | Partly supported | True for words; for physical aggression both sexes copied the man more |
| 4. Boys copy aggression more than girls | Partly supported | True for physical aggression, not for words |
Try this
A classmate looks at the first chart and says: 'Girls who watched the aggressive woman scored 5.5 and girls who watched the aggressive man scored 7.2. So girls copy men more. The difference is 1.7, which proves it.'
What is wrong with this reasoning?
Show the reasoning
Three things. First, each of those bars is the average of only six girls, so one or two girls behaving differently could easily produce a gap of 1.7. Second, the authors did not test this particular difference, and they explain that the kind of statistical test they had to use could not test the effect of the model's sex overall. Third, there was only one female and one male model, so even a real difference could reflect those two people rather than men and women in general.
What you can say is that the pattern fits the authors' interpretation: for physical aggression, the male model seemed to be copied more by both sexes. The word "proves" is not justified.