Transmission of Aggression Through Imitation of Aggressive Models (Bandura, Ross & Ross, 1961)

Background & Theoretical Context

Study: "Transmission of Aggression Through Imitation of Aggressive Models" (Bandura, Ross & Ross, $1961$ ) published in Journal of Abnormal and Social Psychology, Vol. $63(3)$ , pp. $575-582$
Built on prior findings that:
- Children imitate adult behavior in the model’s presence (Bandura & Huston, $1961$ ).
- Observation of a model’s responses facilitates matching reactions in immediate social‐influence settings (Blake $1958$ ; Grosser, Polansky & Lippitt $1951$ ; Rosenblith $1959$ ; Schachter & Hall $1952$ ).
Key research gap: Does imitation generalize to new settings when the model is absent?
Conceptual cornerstones:
- Incidental (observational) learning vs. direct reinforcement.
- Social imitation as a shortcut to acquiring new behaviors without trial-and-error shaping (Skinner $1953$ ).
- Sex-typed behaviors & differential reinforcement histories (Pauls & Smith $1956$ ).

Core Hypotheses

H1 (Facilitation): Children exposed to aggressive models will reproduce similar aggressive acts in a later, separate room.
H2 (Inhibition): Observation of non-aggressive subdued models will generally suppress later aggression relative to a no-model control.
H3 (Sex Matching): Children will imitate same-sex models more than opposite-sex models due to past reinforcement of sex-appropriate behaviors.
H4 (Masculinity Bias): Because aggression is masculine-typed, boys will imitate aggression more than girls; the largest difference expected after exposure to a male aggressive model.

Participants

$72$ preschoolers (Stanford University Nursery School): $36$ boys & $36$ girls.
Age range: $37–69$ months (Mean $=52$ months).
Adult models: $1$ male, $1$ female.
One female experimenter ran all sessions.

Experimental Design & Grouping

$8$ experimental groups ( $n=6$ each) + $1$ control group ( $n=24$ ).
Factors:
- Model Behavior: Aggressive vs Non-aggressive.
- Child Sex: Boy vs Girl.
- Model Sex: Same-sex vs Opposite-sex.
Control children had no exposure to models pre-test.
Matching: Children pre-rated on $4$ aggression scales (physical, verbal, toward objects, aggressive inhibition) $<br>ightarrow$ composite score used to form matched triplets (reliability $r=.89$ ). Each triplet randomly split across aggressive model / non-aggressive model / control.

Materials & Rooms

Exposure Room:
- Child’s corner: potato prints & picture-sticker craft (high interest, diverts attention yet allows unobtrusive observation).
- Model’s corner: tinker toys, mallet, $5$ -ft inflated “Bobo” doll.
Test Room (Generalization):
- Aggressive toys: $3$ -ft Bobo doll, mallet & peg board, $2$ dart guns, tether ball with face.
- Non-aggressive toys: tea set, crayons, dolls, bears, cars, trucks, farm animals, ball.
Arousal Anteroom: premium toys (e.g., fire engine, jet plane); children briefly allowed to play $<br>ightarrow$ withdrawn to induce mild frustration/aggression.

Exposure Procedures

Non-aggressive model: quietly assembled tinker toys; ignored Bobo doll.
Aggressive model: after $ext{≈}1$ minute of tinkering, enacted a standardized 4-part aggressive sequence repeated $ext{≈}3$ times:
1. Lay Bobo on side, sit & punch nose repeatedly.
2. Stand, strike Bobo on head with mallet.
3. Toss Bobo into air.
4. Kick Bobo around room.
Verbal aggression phrases: “Sock him in the nose,” “Hit him down,” “Kick him,” “Throw him in the air,” “Pow!”
Non-aggressive utterances: “He keeps coming back for more,” “He sure is a tough fella.”
Children observed without instructions to imitate; learning, if any, was covert.

Aggression Arousal Phase

All children (experimental + control) experienced toy withdrawal frustration to raise instigation level & equalize baseline tendencies.
Rationale:
- Observation of aggression can paradoxically lower aggression (catharsis); mild provocation ensures capacity to express learned responses.
- Absence of aggression post-provocation in non-aggressive model group would index inhibition.

Test for Delayed Imitation (20-min Session)

Experimenter remained unobtrusively; observers watched via one-way mirror.
Time sampling: $20\text{ min}$ split into $240$ five-second intervals (electric timer).
Primary scorer: male model (blind to most assignments); half sessions double-scored for reliability (product-moment r>.90).

Coded Behaviors (Response Categories)

Imitative Physical Aggression (exact actions vs Bobo).
Imitative Verbal Aggression (quoted phrases).
Imitative Non-aggressive Remarks (quoted comments).
Partial Imitation:
- Mallet aggression vs other objects.
- Sitting on Bobo without hitting.
Non-Imitative Aggression:
- Punch/slap Bobo.
- Other hostile physical/verbal acts.
- Aggressive gun play (dart or pretend shots).
Non-Aggressive Activities & Passive Sitting.

Key Results (Stats in Text)

Statistical tools: Friedman two-way ANOVA by ranks, sign tests, Cochran $Q$ , Lindquist Type III ANOVA, $t$ tests.

Imitation Measures

Aggressive-model group scored dramatically higher than both comparison groups on:
- Imitative Physical Aggression \chi^2_r = 27.17,\; p<.001.
- Imitative Verbal Aggression \chi^2_r = 9.17,\; p<.02.
- Imitative Non-aggressive Phrases \chi^2_r = 17.50,\; p<.001 (≈ one-third echoed comments; other groups $=0$ ).
Non-aggressive vs Control: no significant difference on pure imitation indices.

Partial Imitation

Mallet Aggression: overall \chi^2_r = 11.06,\; p<.01.
- Aggressive > Non-aggressive (p = .026).
- Control > Non-aggressive (p = .018).
Sitting on Bobo: \chi^2_r = 13.44,\; p<.01; Aggressive group highest.

Non-Imitative Aggression

Physical & verbal (non-Bobo) aggression: \chi^2_r = 8.96,\; p<.02.
- Aggressive > Non-aggressive (p = .005).
No significant condition effect on punching Bobo or gun play.

Sex Effects & Interactions

Boys > Girls on imitative physical aggression ( t=2.50,\; p<.01 ) after aggressive model.
Sex × Model patterns:
- Male model provoked markedly higher aggression in boys than girls (physical $t=2.07$ ; verbal $t=2.51$ ; gun play $t=2.12$ ).
- Female model: girls showed more verbal imitation & non-imitative aggression than boys, but differences ns (small $N$ , high variance).
Non-aggressive male model produced conspicuous suppression vs controls on five aggression indices (sign-test p-values .002–.07). Female non-aggressive model showed no such effect.

Non-Aggressive Behaviors

Sex-typical play:
- Girls > Boys with dolls, tea set, coloring (all p<.05).
- Boys > Girls exploring guns (p<.01).
Condition effects:
- Non-aggressive-model children played more with dolls than Aggressive or Control ( $t≈2.6$ ).
- They also spent >2× as many intervals just sitting quietly ( t=3.07,\; p<.01 ).

Interpretations & Theoretical Implications

Demonstrated observational learning of novel aggressive patterns without direct reinforcement or practice.
Modeling permits rapid acquisition vs laborious shaping by contingencies (supports Bandura’s social learning framework).
Male model held greater influence, perhaps because:
- Aggression is masculine-typed $\rightarrow$ freer imitation.
- Male adult presence is rarer in nursery setting $\rightarrow$ higher reward value (Rosenblith $1959$ alternative).
Findings challenge “catharsis” (aggression drains aggression): exposure fostered, not reduced, later aggression.
Question psychoanalytic “identification with aggressor” notion; aggression learned even with neutral model and without fear/defense dynamics.

Methodological & Ethical Notes

Matching on baseline aggression enhanced internal validity.
One-way mirrors minimized reactivity; inter-rater reliabilities >.90 ensured measurement precision.
Ethical reflection (implicit): induced frustration & exposure to violence in preschoolers raises questions about potential long-term effects—authors note inhibition vs arousal aspects but no follow-up reported.

Practical & Real-World Relevance

Media violence & real-life aggressive role models can socialize children’s aggression by mere observation.
Non-aggressive adult behavior, especially by salient male figures, can proactively inhibit child aggression.
Classroom & parenting interventions can strategically use modeling to teach prosocial or suppress antisocial behaviors.

Key References Cited

Bandura & Huston $1961$ ; Bandura & Walters $1959$ ; Blake $1958$ ; Grosser et al. $1951$ ; Pauls & Smith $1956$ ; Rosenblith $1959$ ; Skinner $1953$ ; Miller & Dollard $1941$ , etc.

Concise Summary (One-paragraph)

Preschool boys and girls observed an adult model behaving either aggressively or non-aggressively toward an inflated “Bobo” doll, then faced a new room filled with aggressive and non-aggressive toys. Children who witnessed aggression later reproduced the model’s novel physical and verbal assaults, partially imitated related acts, and displayed broader non-imitative aggression—far exceeding both children shown a quiet model and those never seeing a model. Exposure to a subdued male model significantly suppressed subsequent aggression. Boys imitated violent male models most, underscoring sex-typed learning. Overall, the study delivers powerful evidence that children learn and generalize aggressive behavior simply by watching others, without any direct reinforcement, challenging catharsis and demonstrating the potency of observational learning in socialization.

Background & Theoretical Context

Study: "Transmission of Aggression Through Imitation of Aggressive Models" (Bandura, Ross & Ross, $1961$ ) published in Journal of Abnormal and Social Psychology, Vol. $63(3)$ , pp. $575-582$ . This groundbreaking study aimed to empirically investigate the phenomenon of observational learning, specifically how aggressive behaviors can be acquired through mere observation.
Built on prior findings that:
- Children imitate adult behavior in the model’s presence (Bandura & Huston, $1961$ ). Previous research had shown children's tendency to mimic observed actions when the model was still actively present.
- Observation of a model’s responses facilitates matching reactions in immediate social-influence settings (Blake $1958$ ; Grosser, Polansky & Lippitt $1951$ ; Rosenblith $1959$ ; Schachter & Hall $1952$ ). This indicated that a model could influence immediate behavioral responses.
Key research gap: Does imitation generalize to new settings when the model is absent? The primary objective was to determine if learned behaviors would transfer to a different environment and persist even after the model was no longer present, differentiating it from mere mimicry.
Conceptual cornerstones:
- Incidental (observational) learning vs. direct reinforcement. The study explored whether learning could occur without explicit rewards or punishments given directly to the child for their imitation.
- Social imitation as a shortcut to acquiring new behaviors without trial-and-error shaping (Skinner $1953$ ). This challenged the dominant behaviorist view that all learning required direct reinforcement and gradual shaping of behavior.
- Sex-typed behaviors & differential reinforcement histories (Pauls & Smith $1956$ ). The researchers considered how societal norms around gender might influence which aggressive behaviors are more readily imitated by boys versus girls.

Core Hypotheses

H1 (Facilitation): Children exposed to aggressive models will reproduce similar aggressive acts in a later, separate room. This predicted that observing aggressive behavior would directly facilitate its reproduction.
H2 (Inhibition): Observation of non-aggressive subdued models will generally suppress later aggression relative to a no-model control. It was hypothesized that quiet, non-aggressive models would lead to less aggressive behavior from the children.
H3 (Sex Matching): Children will imitate same-sex models more than opposite-sex models due to past reinforcement of sex-appropriate behaviors. This hypothesis stemmed from the idea that children are more likely to emulate behaviors deemed appropriate for their gender, as learned through prior social experiences.
H4 (Masculinity Bias): Because aggression is masculine-typed, boys will imitate aggression more than girls; the largest difference expected after exposure to a male aggressive model. This specifically predicted that boys would exhibit more aggression, especially when observing a male aggressive role model, given societal perceptions of aggression as a male trait.

Participants

$72$ preschoolers (Stanford University Nursery School): $36$ boys & $36$ girls. The sample comprised an equal number of male and female children from a university nursery school.
Age range: $37–69$ months (Mean $=52$ months). The children were all relatively young, averaging just over four years old.
Adult models: $1$ male, $1$ female. Two adult confederates acted as the aggressive or non-aggressive models.
One female experimenter ran all sessions. A consistent experimenter ensured standardized interactions with the children across all conditions.

Experimental Design & Grouping

$8$ experimental groups ( $n=6$ each) + $1$ control group ( $n=24$ ). The total sample of $72$ children was divided into distinct groups to test the hypotheses systematically.
Factors:
- Model Behavior: Aggressive vs Non-aggressive. Children either observed an adult behaving aggressively or non-aggressively.
- Child Sex: Boy vs Girl. The gender of the child was a variable of interest in relation to imitation.
- Model Sex: Same-sex vs Opposite-sex. The gender of the adult model was varied to examine its influence on imitation based on the child's sex.
Control children had no exposure to models pre-test. They served as a baseline to compare the effects of model exposure.
Matching: Children pre-rated on $4$ aggression scales (physical, verbal, toward objects, aggressive inhibition)
ightarrow composite score used to form matched triplets (reliability $r=.89$ ). Each triplet randomly split across aggressive model / non-aggressive model / control. This rigorous matching procedure was implemented to ensure that the groups were comparable in baseline aggression levels, thereby increasing the internal validity of the study and ruling out pre-existing differences as an explanation for the results.

Materials & Rooms

Exposure Room: This room was set up to allow the child to observe the model's behavior.
- Child’s corner: potato prints & picture-sticker craft (high interest, diverts attention yet allows unobtrusive observation). These appealing, quiet activities kept the child engaged but able to observe the model without direct interaction.
- Model’s corner: tinker toys, mallet, $5$ -ft inflated “Bobo” doll. The Bobo doll was the primary object of interaction for the aggressive model, and the tinker toys were used by the non-aggressive model.
Test Room (Generalization): This room was designed to measure generalized imitation of aggression and other behaviors.
- Aggressive toys: $3$ -ft Bobo doll, mallet & peg board, $2$ dart guns, tether ball with face. These toys were specifically chosen to elicit aggressive responses or allow for the imitation of the model's aggressive actions.
- Non-aggressive toys: tea set, crayons, dolls, bears, cars, trucks, farm animals, ball. These items provided alternatives for non-aggressive play.
Arousal Anteroom: premium toys (e.g., fire engine, jet plane); children briefly allowed to play
ightarrow withdrawn to induce mild frustration/aggression. This phase was crucial for standardizing the children's emotional state before the test, aiming to increase the likelihood of aggressive expression if it had been learned.

Exposure Procedures

Non-aggressive model: quietly assembled tinker toys; ignored Bobo doll. The non-aggressive model demonstrated subdued, task-oriented play, providing a contrast to the aggressive condition.
Aggressive model: after $ext{≈}1$ minute of tinkering, enacted a standardized 4-part aggressive sequence repeated $ext{≈}3$ times. This sequence was meticulously pre-defined to ensure consistency across trials and models:
1. Lay Bobo on side, sit & punch nose repeatedly.
2. Stand, strike Bobo on head with mallet.
3. Toss Bobo into air.
4. Kick Bobo around room.
Verbal aggression phrases: “Sock him in the nose,” “Hit him down,” “Kick him,” “Throw him in the air,” “Pow!” These specific phrases were uttered by the aggressive model to allow for later measurement of imitative verbal aggression.
Non-aggressive utterances: “He keeps coming back for more,” “He sure is a tough fella.” These neutral comments were also made by the aggressive model, providing a contrast to aggressive verbalizations.
Children observed without instructions to imitate; learning, if any, was covert. This emphasized that learning was incidental and observational, rather than being explicitly requested or trained.

Aggression Arousal Phase

All children (experimental + control) experienced toy withdrawal frustration to raise instigation level & equalize baseline tendencies. This manipulation was implemented after the observation phase but before the test phase.
Rationale:
- Observation of aggression can paradoxically lower aggression (catharsis); mild provocation ensures capacity to express learned responses. The researchers wanted to avoid the possibility that merely observing aggression could reduce a child's own aggressive drive, ensuring any learned aggression would have an outlet.
- Absence of aggression post-provocation in non-aggressive model group would index inhibition. If children from the non-aggressive model group showed less aggression even after provocation, it would indicate that observing a subdued model inhibits aggression.

Test for Delayed Imitation (20-min Session)

Experimenter remained unobtrusively; observers watched via one-way mirror. The experimenter stayed in the room but did not interact with the child, and trained observers recorded behavior from behind a one-way mirror, minimizing demand characteristics.
Time sampling: $20 ext{ min}$ split into $240$ five-second intervals (electric timer). This systematic observation method allowed for quantitative data collection on the frequency and duration of various behaviors.
Primary scorer: male model (blind to most assignments); half sessions double-scored for reliability (product-moment r>.90). Using the male model as the primary scorer (who was unaware of the specific experimental condition for each child) helped reduce observer bias. High inter-rater reliability confirmed the consistency of the behavioral coding.

Coded Behaviors (Response Categories)

Imitative Physical Aggression (exact actions vs Bobo): This included direct replication of the model's specific physical attacks on the Bobo doll (e.g., punching its nose, hitting with mallet on head, kicking).
Imitative Verbal Aggression (quoted phrases): Exact reproduction of the model's aggressive verbalizations (e.g., "Sock him!", "Pow!").
Imitative Non-aggressive Remarks (quoted comments): Reproduction of the model's neutral or non-aggressive verbalizations (e.g., "He keeps coming back for more").
Partial Imitation:
- Mallet aggression vs other objects: Using the mallet aggressively on toys other than the Bobo doll.
- Sitting on Bobo without hitting: Mimicking the model's posture without necessarily continuing the aggressive action.
Non-Imitative Aggression:
- Punch/slap Bobo: Aggressive acts towards the Bobo doll that were not exact replicas of the model's actions.
- Other hostile physical/verbal acts: Aggressive behaviors directed at other toys or expressed verbally, not seen in the model's specific sequence.
- Aggressive gun play (dart or pretend shots): Use of dart guns or pretend shooting with other toys, indicative of general aggressive play.
Non-Aggressive Activities & Passive Sitting: Time spent playing with quiet toys or simply sitting without engaging in any activity.

Key Results (Stats in Text)

Statistical tools: Friedman two-way ANOVA by ranks, sign tests, Cochran $Q$ , Lindquist Type III ANOVA, $t$ tests. A variety of statistical methods appropriate for the experimental design and type of data were employed.

Imitation Measures

Aggressive-model group scored dramatically higher than both comparison groups on:
- Imitative Physical Aggression ext{( ext{r}=27.17, ext{ }p<.001)}. This was a highly significant finding, indicating that children exposed to the aggressive model reproduced physical aggression far more frequently.
- Imitative Verbal Aggression ext{( ext{r}=9.17, ext{ }p<.02)}. Similarly, specific aggressive phrases from the model were mimicked significantly more by the aggressive-model group.
- Imitative Non-aggressive Phrases ext{( ext{r}=17.50, ext{ }p<.001)} (≈ one-third echoed comments; other groups $=0$ ). Even the non-aggressive utterances made by the aggressive model were imitated by a substantial portion of the children in that group, while almost none in other groups did.
Non-aggressive vs Control: no significant difference on pure imitation indices. Observing a non-aggressive model did not lead to increased imitation of neutral behaviors compared to not seeing a model at all.

Partial Imitation

Mallet Aggression: overall ext{( ext{r}=11.06, ext{ }p<.01)}.
- Aggressive > Non-aggressive (p = .026). Children in the aggressive model condition were more likely to use the mallet aggressively on other objects besides the Bobo doll.
- Control > Non-aggressive (p = .018). Interestingly, the control group also showed more mallet aggression than the non-aggressive model group, suggesting an inhibitory effect of the non-aggressive model.
Sitting on Bobo: ext{( ext{r}=13.44, ext{ }p<.01)}; Aggressive group highest. Children who saw the aggressive model were significantly more likely to sit on the Bobo doll, even if they didn't hit it.

Non-Imitative Aggression

Physical & verbal (non-Bobo) aggression: ext{( ext{r}=8.96, ext{ }p<.02)}.
- Aggressive > Non-aggressive (p = .005). Children who observed the aggressive model displayed more generalized aggressive behaviors not directly linked to the Bobo doll.
No significant condition effect on punching Bobo or gun play. While there was generalized aggression, specific non-imitative punching of Bobo and gun play did not show statistically significant differences across conditions.

Sex Effects & Interactions

Boys > Girls on imitative physical aggression ( t=2.50, ext{ }p<.01 ) after aggressive model. This confirmed H4, indicating boys were more prone to directly copy physical aggression.
Sex $imes$ Model patterns:
- Male model provoked markedly higher aggression in boys than girls (physical $t=2.07$ ; verbal $t=2.51$ ; gun play $t=2.12$ ). This strongly supported the masculinity bias hypothesis, showing the significant impact of a male aggressive model on boys' aggressive behaviors.
- Female model: girls showed more verbal imitation & non-imitative aggression than boys, but differences ns (small $N$ , high variance). While trends suggested female models had a greater influence on girls for some types of aggression, these effects were not statistically significant, possibly due to smaller sample sizes within specific sub-groups and greater variability in responses.
Non-aggressive male model produced conspicuous suppression vs controls on five aggression indices (sign-test p-values .002–.07). This was a crucial finding for H2, demonstrating that the presence of a calm, non-aggressive male model significantly reduced children's aggressive tendencies compared to the control group.
Female non-aggressive model showed no such effect. The female non-aggressive model did not produce the same inhibitory effect on aggression as the male non-aggressive model.

Non-Aggressive Behaviors

Sex-typical play:
- Girls > Boys with dolls, tea set, coloring (all p<.05). Girls demonstrated significantly more engagement with toys traditionally associated with female play.
- Boys > Girls exploring guns (p<.01). Boys were significantly more likely to engage in play with

Background & Theoretical Context

Study: "Transmission of Aggression Through Imitation of Aggressive Models" (Bandura, Ross & Ross, $1961$ ) published in Journal of Abnormal and Social Psychology, Vol. $63(3)$ , pp. $575-582$ . This groundbreaking study aimed to empirically investigate the phenomenon of observational learning, specifically how aggressive behaviors can be acquired through mere observation.
Built on prior findings that:
- Children imitate adult behavior in the model’s presence (Bandura & Huston, $1961$ ). Previous research had shown children's tendency to mimic observed actions when the model was still actively present.
- Observation of a model’s responses facilitates matching reactions in immediate social-influence settings (Blake $1958$ ; Grosser, Polansky & Lippitt $1951$ ; Rosenblith $1959$ ; Schachter & Hall $1952$ ). This indicated that a model could influence immediate behavioral responses.
Key research gap: Does imitation generalize to new settings when the model is absent? The primary objective was to determine if learned behaviors would transfer to a different environment and persist even after the model was no longer present, differentiating it from mere mimicry.
Conceptual cornerstones:
- Incidental (observational) learning vs. direct reinforcement. The study explored whether learning could occur without explicit rewards or punishments given directly to the child for their imitation.
- Social imitation as a shortcut to acquiring new behaviors without trial-and-error shaping (Skinner $1953$ ). This challenged the dominant behaviorist view that all learning required direct reinforcement and gradual shaping of behavior.
- Sex-typed behaviors & differential reinforcement histories (Pauls & Smith $1956$ ). The researchers considered how societal norms around gender might influence which aggressive behaviors are more readily imitated by boys versus girls.

Core Hypotheses

H1 (Facilitation): Children exposed to aggressive models will reproduce similar aggressive acts in a later, separate room. This predicted that observing aggressive behavior would directly facilitate its reproduction.
H2 (Inhibition): Observation of non-aggressive subdued models will generally suppress later aggression relative to a no-model control. It was hypothesized that quiet, non-aggressive models would lead to less aggressive behavior from the children.
H3 (Sex Matching): Children will imitate same-sex models more than opposite-sex models due to past reinforcement of sex-appropriate behaviors. This hypothesis stemmed from the idea that children are more likely to emulate behaviors deemed appropriate for their gender, as learned through prior social experiences.
H4 (Masculinity Bias): Because aggression is masculine-typed, boys will imitate aggression more than girls; the largest difference expected after exposure to a male aggressive model. This specifically predicted that boys would exhibit more aggression, especially when observing a male aggressive role model, given societal perceptions of aggression as a male trait.

Participants

$72$ preschoolers (Stanford University Nursery School): $36$ boys & $36$ girls. The sample comprised an equal number of male and female children from a university nursery school.
Age range: $37–69$ months (Mean $=52$ months). The children were all relatively young, averaging just over four years old.
Adult models: $1$ male, $1$ female. Two adult confederates acted as the aggressive or non-aggressive models.
One female experimenter ran all sessions. A consistent experimenter ensured standardized interactions with the children across all conditions.

Experimental Design & Grouping

$8$ experimental groups ( $n=6$ each) + $1$ control group ( $n=24$ ). The total sample of $72$ children was divided into distinct groups to test the hypotheses systematically.
Factors:
- Model Behavior: Aggressive vs Non-aggressive. Children either observed an adult behaving aggressively or non-aggressively.
- Child Sex: Boy vs Girl. The gender of the child was a variable of interest in relation to imitation.
- Model Sex: Same-sex vs Opposite-sex. The gender of the adult model was varied to examine its influence on imitation based on the child's sex.
Control children had no exposure to models pre-test. They served as a baseline to compare the effects of model exposure.
Matching: Children pre-rated on $4$ aggression scales (physical, verbal, toward objects, aggressive inhibition) $\rightarrow$ composite score used to form matched triplets (reliability $r=.89$ ). Each triplet randomly split across aggressive model / non-aggressive model / control. This rigorous matching procedure was implemented to ensure that the groups were comparable in baseline aggression levels, thereby increasing the internal validity of the study and ruling out pre-existing differences as an explanation for the results.

Materials & Rooms

Exposure Room: This room was set up to allow the child to observe the model's behavior.
- Child’s corner: potato prints & picture-sticker craft (high interest, diverts attention yet allows unobtrusive observation). These appealing, quiet activities kept the child engaged but able to observe the model without direct interaction.
- Model’s corner: tinker toys, mallet, $5$ -ft inflated “Bobo” doll. The Bobo doll was the primary object of interaction for the aggressive model, and the tinker toys were used by the non-aggressive model.
Test Room (Generalization): This room was designed to measure generalized imitation of aggression and other behaviors.
- Aggressive toys: $3$ -ft Bobo doll, mallet & peg board, $2$ dart guns, tether ball with face. These toys were specifically chosen to elicit aggressive responses or allow for the imitation of the model's aggressive actions.
- Non-aggressive toys: tea set, crayons, dolls, bears, cars, trucks, farm animals, ball. These items provided alternatives for non-aggressive play.
Arousal Anteroom: premium toys (e.g., fire engine, jet plane); children briefly allowed to play $\rightarrow$ withdrawn to induce mild frustration/aggression. This phase was crucial for standardizing the children's emotional state before the test, aiming to increase the likelihood of aggressive expression if it had been learned.

Exposure Procedures

Non-aggressive model: quietly assembled tinker toys; ignored Bobo doll. The non-aggressive model demonstrated subdued, task-oriented play, providing a contrast to the aggressive condition.
Aggressive model: after $\approx1$ minute of tinkering, enacted a standardized 4-part aggressive sequence repeated $\approx3$ times:
1. Lay Bobo on side, sit & punch nose repeatedly.
2. Stand, strike Bobo on head with mallet.
3. Toss Bobo into air.
4. Kick Bobo around room.
Verbal aggression phrases: “Sock him in the nose,” “Hit him down,” “Kick him,” “Throw him in the air,” “Pow!” These specific phrases were uttered by the aggressive model to allow for later measurement of imitative verbal aggression.
Non-aggressive utterances: “He keeps coming back for more,” “He sure is a tough fella.” These neutral comments were also made by the aggressive model, providing a contrast to aggressive verbalizations.
Children observed without instructions to imitate; learning, if any, was covert. This emphasized that learning was incidental and observational, rather than being explicitly requested or trained.

Aggression Arousal Phase

All children (experimental + control) experienced toy withdrawal frustration to raise instigation level & equalize baseline tendencies. This manipulation was implemented after the observation phase but before the test phase.
Rationale:
- Observation of aggression can paradoxically lower aggression (catharsis); mild provocation ensures capacity to express learned responses. The researchers wanted to avoid the possibility that merely observing aggression could reduce a child's own aggressive drive, ensuring any learned aggression would have an outlet.
- Absence of aggression post-provocation in non-aggressive model group would index inhibition. If children from the non-aggressive model group showed less aggression even after provocation, it would indicate that observing a subdued model inhibits aggression.

Test for Delayed Imitation (20-min Session)

Experimenter remained unobtrusively; observers watched via one-way mirror. The experimenter stayed in the room but did not interact with the child, and trained observers recorded behavior from behind a one-way mirror, minimizing demand characteristics.
Time sampling: $20\text{ min}$ split into $240$ five-second intervals (electric timer). This systematic observation method allowed for quantitative data collection on the frequency and duration of various behaviors.
Primary scorer: male model (blind to most assignments); half sessions double-scored for reliability (product-moment r>.90). Using the male model as the primary scorer (who was unaware of the specific experimental condition for each child) helped reduce observer bias. High inter-rater reliability confirmed the consistency of the behavioral coding.

Coded Behaviors (Response Categories)

Imitative Physical Aggression (exact actions vs Bobo): This included direct replication of the model's specific physical attacks on the Bobo doll (e.g., punching its nose, hitting with mallet on head, kicking).
Imitative Verbal Aggression (quoted phrases): Exact reproduction of the model's aggressive verbalizations (e.g., "Sock him!", "Pow!").
Imitative Non-aggressive Remarks (quoted comments): Reproduction of the model's neutral or non-aggressive verbalizations (e.g., "He keeps coming back for more").
Partial Imitation:
- Mallet aggression vs other objects: Using the mallet aggressively on toys other than the Bobo doll.
- Sitting on Bobo without hitting: Mimicking the model's posture without necessarily continuing the aggressive action.
Non-Imitative Aggression:
- Punch/slap Bobo: Aggressive acts towards the Bobo doll that were not exact replicas of the model's actions.
- Other hostile physical/verbal acts: Aggressive behaviors directed at other toys or expressed verbally, not seen in the model's specific sequence.
- Aggressive gun play (dart or pretend shots): Use of dart guns or pretend shooting with other toys, indicative of general aggressive play.
Non-Aggressive Activities & Passive Sitting: Time spent playing with quiet toys or simply sitting without engaging in any activity.

Key Results (Stats in Text)

Statistical tools: Friedman two-way ANOVA by ranks, sign tests, Cochran $Q$ , Lindquist Type III ANOVA, $t$ tests. A variety of statistical methods appropriate for the experimental design and type of data were employed.

Imitation Measures

Aggressive-model group scored dramatically higher than both comparison groups on:
- Imitative Physical Aggression \chi^2_r = 27.17,\; p<.001. This was a highly significant finding, indicating that children exposed to the aggressive model reproduced physical aggression far more frequently.
- Imitative Verbal Aggression \chi^2_r = 9.17,\; p<.02. Similarly, specific aggressive phrases from the model were mimicked significantly more by the aggressive-model group.
- Imitative Non-aggressive Phrases \chi^2_r = 17.50,\; p<.001 ($\approx$ one-third echoed comments; other groups $=0$ ). Even the non-aggressive utterances made by the aggressive model were imitated by a substantial portion of the children in that group, while almost none in other groups did.
Non-aggressive vs Control: no significant difference on pure imitation indices. Observing a non-aggressive model did not lead to increased imitation of neutral behaviors compared to not seeing a model at all.

Partial Imitation

Mallet Aggression: overall \chi^2_r = 11.06,\; p<.01.
- Aggressive > Non-aggressive (p = .026). Children in the aggressive model condition were more likely to use the mallet aggressively on other objects besides the Bobo doll.
- Control > Non-aggressive (p = .018). Interestingly, the control group also showed more mallet aggression than the non-aggressive model group, suggesting an inhibitory effect of the non-aggressive model.
Sitting on Bobo: \chi^2_r = 13.44,\; p<.01; Aggressive group highest. Children who saw the aggressive model were significantly more likely to sit on the Bobo doll, even if they didn't hit it.

Non-Imitative Aggression

Physical & verbal (non-Bobo) aggression: \chi^2_r = 8.96,\; p<.02.
- Aggressive > Non-aggressive (p = .005). Children who observed the aggressive model displayed more generalized aggressive behaviors not directly linked to the Bobo doll.
No significant condition effect on punching Bobo or gun play. While there was generalized aggression, specific non-imitative punching of Bobo and gun play did not show statistically significant differences across conditions.

Sex Effects & Interactions

Boys > Girls on imitative physical aggression ( t=2.50,\; p<.01 ) after aggressive model. This confirmed H4, indicating boys were more prone to directly copy physical aggression.
Sex $\times$ Model patterns:
- Male model provoked markedly higher aggression in boys than girls (physical $t=2.07$ ; verbal $t=2.51$ ; gun play $t=2.12$ ). This strongly supported the masculinity bias hypothesis, showing the significant impact of a male aggressive model on boys' aggressive behaviors.
- Female model: girls showed more verbal imitation & non-imitative aggression than boys, but differences ns (small $N$ , high variance). While trends suggested female models had a greater influence on girls for some types of aggression, these effects were not statistically significant, possibly due to smaller sample sizes within specific sub-groups and greater variability in responses.
Non-aggressive male model produced conspicuous suppression vs controls on five aggression indices (sign-test p-values .002–.07). This was a crucial finding for H2, demonstrating that the presence of a calm, non-aggressive male model significantly reduced children's aggressive tendencies compared to the control group.
Female non-aggressive model showed no such effect. The female non-aggressive model did not produce the same inhibitory effect on aggression as the male non-aggressive model.

Non-Aggressive Behaviors

Sex-typical play:
- Girls > Boys with dolls, tea set, coloring (all p<.05). Girls demonstrated significantly more engagement with toys traditionally associated with female play.
- Boys > Girls exploring guns (p<.01). Boys were significantly more likely to engage in play with aggressive toys like guns than girls.
Condition effects:
- Non-aggressive-model children played more with dolls than Aggressive or Control ( $t\approx2.6$ ). This suggests that exposure to a non-aggressive model elicited more quiet, constructive play.
- They also spent >2x as many intervals just sitting quietly ( t=3.07,\; p<.01 ). This reinforces the inhibitory effect of the non-aggressive model, leading to calmer, less active behavior.

Interpretations & Theoretical Implications

Demonstrated observational learning of novel aggressive patterns without direct reinforcement or practice. This was a primary conclusion, showing that children can learn complex behaviors simply by watching, without needing to perform them or be rewarded.
Modeling permits rapid acquisition vs laborious shaping by contingencies (supports Bandura’s social learning framework). The study provided strong evidence against purely Skinnerian views of learning, suggesting modeling is a much more efficient mechanism.
Male model held greater influence, perhaps because:
- Aggression is masculine-typed $\rightarrow$ freer imitation. Children internalize gender norms and are more likely to imitate behaviors that align with their perceived gender roles.
- Male adult presence is rarer in nursery setting $\rightarrow$ higher reward value (Rosenblith $1959$ alternative). An alternative explanation is that male figures are less common in a nursery environment, making them more novel and potentially increasing their influence.
Findings challenge “catharsis” (aggression drains aggression): exposure fostered, not reduced, later aggression. The results directly contradicted the psychoanalytic concept of catharsis, showing that observing aggression did not reduce aggressive impulses but rather increased them.
Question psychoanalytic “identification with aggressor” notion; aggression learned even with neutral model and without fear/defense dynamics. The study suggested that aggressive learning is a broader phenomenon than just identifying with a feared or powerful aggressor.

Methodological & Ethical Notes

Matching on baseline aggression enhanced internal validity. This control measure ensured that initial differences in aggression levels among children did not confound the results.
One-way mirrors minimized reactivity; inter-rater reliabilities >.90 ensured measurement precision. These methodological details underscore the rigor and objectivity of the data collection.
Ethical reflection (implicit): induced frustration & exposure to violence in preschoolers raises questions about potential long-term effects—authors note inhibition vs arousal aspects but no follow-up reported. Modern ethical guidelines would likely require more extensive debriefing and long-term follow-up for studies involving such manipulations with young children.

Practical & Real-World Relevance

Media violence & real-life aggressive role models can socialize children’s aggression by mere observation. This study has profound implications for understanding the effects of violent media and the importance of positive role models.
Non-aggressive adult behavior, especially by salient male figures, can proactively inhibit child aggression. The inhibitory effect of the non-aggressive male model suggests a powerful tool for promoting prosocial behavior.
Classroom & parenting interventions can strategically use modeling to teach prosocial or suppress antisocial behaviors. The findings provide a foundation for educational and parenting strategies emphasizing positive role modeling.

Key References Cited

Bandura & Huston $1961$ ; Bandura & Walters $1959$ ; Blake $1958$ ; Grosser et al. $1951$ ; Pauls & Smith $1956$ ; Rosenblith $1959$ ; Skinner $1953$ ; Miller & Dollard $1941$ , etc.