2.3.1 Hostile Attributional Bias and Reactive versus Proactive Aggressive Behavior

The first test was whether hostile attributional bias predicts both reactive aggression and instrumental (proactive) aggression. Reference BerkowitzBerkowitz (1962) and Reference BanduraBandura (1973) argued for 60 years about whether aggressive behavior is an emotional reaction to frustration (Berkowitz) or an instrumental behavior in response to modeling and in pursuit of rewards (Bandura). The debate has been largely resolved through recognition that both scholars are correct but in different contexts: Aggression takes multiple forms, functions, etiologies, and consequences. Berkowitz studied reactive aggression, also known as impulsive aggression, intermittent explosive disorder, and retaliatory aggression. Bandura’s form of aggression is less emotional and more directed at pursuit of rewards, known as instrumental aggression or proactive aggression. Hostile attributional bias should lead to reactive but not instrumental aggression.

Reference Dodge and CoieDodge and Coie (1987) tested this hypothesis by developing a reliable and valid teacher-rating measurement tool to assess these two forms of aggressive behavior. Although the scores for reactive and proactive aggression were positively correlated, they were able to identify five groups of boys: (1) highly reactive and highly proactive; (2) highly reactive but not proactive; (3) highly proactive but not reactive; (4) nonaggressive but still socially rejected; and (5) nonaggressive and well-functioning. They administered to all boys a new test of hostile attributional bias in which they presented videorecorded vignettes depicting provocations by one boy toward another. They asked each boy to imagine being the boy wearing a numbered t-shirt as that boy became the object of an ambiguous provocation by another boy. As depicted in Figure 3, the two groups of reactively aggressive boys were about twice as likely as the other groups to attribute hostile intent to the ambiguous provocation (left side of figure) and to react aggressively to the ambiguous provocation (right side of figure). The proactively aggressive boys did not show these patterns.

Figure 3 Hostile attributional bias and retaliatory aggression among reactive aggressive, proactive aggressive, combined aggressive, and nonaggressive children: Reactive aggressive children display hostile attributional bias.

(adapted from Reference Dodge and CoieDodge & Coie, 1987)

But how did these boys behave in real interactions with other boys? We created play groups in which we asked these boys to play with each other for 45 minutes on each of five consecutive days and videorecorded 551 instances of overreactive aggression and 1,866 instances of proactive aggression. In between sessions, boys were interviewed to assess hostile attributional biases. True to hypotheses, a boy’s hostile attributional bias tendency predicted the rate at which he displayed reactive aggressive behavior (r = .20, p < .05) but not the rate at which he displayed proactive aggression (r = .05, n.s.; Reference Dodge and CoieDodge & Coie, 1987).

By now, dozens of studies have confirmed the significant relation between hostile attributional bias and reactive, but not proactive, aggressive behavior, some by our own group (e.g., Reference Crick and DodgeCrick & Dodge, 1996; Reference Dodge, Lochman, Harnish, Bates and PettitDodge et al., 1997; Reference Schwartz, McFadyen-Ketchum, Dodge, Pettit and BatesSchwartz et al., 1999) and others by independent scholars (e.g., Reference Smithmyer, Hubbard and SimonsSmithmyer et al., 2000; Reference WaltersWalters, 2007). These studies have also shown that other social information processing patterns (e.g., the decision-making judgment that aggression will lead to desired outcomes) predict proactive aggressive behavior.

2.3.2 Hostile Attributional Bias and Depression

Making a hostile attribution does not inevitably lead children to react aggressively. Another mental process must follow; that is, children must also judge that reacting aggressively will lead to a positive outcome such as reducing negative internal emotion or stopping the peer from provoking again. What if children attribute hostile intent to a peer but judge that reacting aggressively would only lead to more problems?

Garber and Dodge (Reference Garber, Quiggle, Panak, Dodge, Cicchetti and TothGarber et al., 1991; Reference Quiggle, Panak, Garber and DodgeQuiggle et al., 1992) hypothesized that depressed children make hostile attributions like aggressive children do but also judge that retaliating aggressively would be foolish, and so they respond in an internalizing rather than externalizing way. This team identified four groups of children (aggressive, depressed, comorbid aggressive-and-depressed, and non-disordered) and assessed their hostile attributional biases, emotional reactions, and behavioral judgments in response to hypothetical provocations. Unique profiles of processing patterns emerged for each group. The aggressive group showed a pattern of hostile attributional bias coupled with evaluations that retaliatory aggression would be a “good” way to respond and that they could engage in aggressive retaliation rather easily (these individual social information processing patterns coalesce into an overall defensive mindset). The depressed group also showed a high level of hostile attributional bias, but they evaluated aggressive responses as a “bad” way to respond that they could not implement easily. Furthermore, they reported feeling sad when the provocation occurred and attributed the provocation to global, stable, and internal causes (that is, they infer the peer had hostile intent, but the child must have done something to cause the peer to provoke them), and they evaluated behavioral withdrawal (e.g., hiding, crying) as the best and easiest way to respond. This pattern might be called a depressive mindset. The comorbid group responded with a pattern that combined responses of the aggressive and depressed groups.

3.2.1 Step 1: Attention

As a social stimulus presents itself, the child’s first step of processing is to attend to the social cues. There are literally millions of cues in an environment at any moment, far too many for any person to encode. And cues change with each passing moment. Humans have evolved over millennia so that social encounters take priority and capture particular attention, from the moment of birth when an infant begins to focus on their mother and interpret her smile as comforting. As the initial process through which children encounter a social stimulus, attention influences all downstream processes. It is easy to contemplate the impact that attending to a peer’s face and the peer’s look of surprise (or malice) will have on the way children respond to being bumped by a peer. Both within- and between-child differences predict behavioral responding.

Because there is no such thing as an eidetic (photographic) social memory store, people have evolved to use common heuristics to code the stimulus: “Focus on the other person’s face,” “Attend to tone of voice as much as the words,” “Create a narrative that has meaning and store in memory the narrative rather than the individual parts.” One can easily imagine children’s occasional missteps in which they fail to pay proper attention due to fatigue, preoccupation, or random error; and one can imagine the success that follows when children consciously bring their eyes into “focus” on a particular task.

Between-child differences are also important. People develop heuristics that are unique to their own experience: “Be wary of rapid movements.” “Look for the other’s weak spot.” “Find the good in others.” In rapid time, people encode stimuli into the brain in their unique way. Clear and obvious cues are coded similarly by most people, but, as the cue becomes more ambiguous, individual differences play an increasing role. Some differences are a function of acquired skill, and others are a function of unique experiences that bias attention. Reference SimonSimon (1967) showed that chess masters are expert at noticing an opponent’s early attempts to set up a later attack on the king. One child might have acquired exquisite skill at detecting another’s facial movements (perhaps through numerous past emotional encounters), whereas another child may be oblivious. We know these patterns as “selective attention,” “biased attention,” “hypervigilance,” “attentiveness,” and “attention deficits.”

In the CDP sample, Reference Dodge, Pettit, Bates and ValenteDodge et al. (1995) scored children’s open-ended responses to each of 24 provocation vignettes as indicating attention to relevant social cues (or not). As hypothesized, the sum score for lack of attention to relevant cues predicted higher teacher-rated aggression on the Child Behavior Checklist and more classroom peer nominations for being an aggressive behavior problem (Reference Dodge, Pettit, Bates and ValenteDodge et al., 1995). Reference Dodge and PriceDodge and Price (1994) replicated these findings with the SDP sample. Once again, children’s scores for lack of attention to relevant social cues predicted both higher teacher ratings and more peer nominations. A feature of the SDP sample was the different ages of children (6, 7, or 8), enabling a demonstration that attention to relevant cues improves with each advancing year and the correlation between attention scores and children’s aggressive behavior increases as children get older.

Other investigators have created different ways of measuring children’s selective or biased attention (such as distraction and total attentive capacity) to social cues. Reference GouzeGouze (1987) asked 4- and 5-year-old boys to pay attention to a video screen that alternately displayed 12 short vignettes in which two puppets were either fighting with each other or sharing with each other, and to press a button as quickly as possible when a red light turned on in the corner of the screen. The average latency of pressing the button while the puppets were fighting, minus the average latency of responding while the puppets were sharing, was an indication of hyper-attention to aggressive stimuli. These children were then observed on the playground for at least 50 minutes each. A boy’s hyper-attention-to-aggression score predicted a higher rate of physically aggressing against peers. Heightened focus on aggression and threat by others is the first processing step in the pattern of defensive mindedness.

3.2.2 Step 2: Interpretation

As cues are encoded in the brain, they get interpreted. People have evolved to interpret social cues along specific dimensions that foster survival (is the cue a threat?), security (is the cue calming?), and growth (does the cue indicate an opportunity?).

The most basic dimension of interpretation is threat versus no threat. Infants initially interpret all disruptive cues (e.g., a loud sound, sudden movement) as threat (not consciously, but instinctively) and respond with crying, a facial expression of fear, and a psychophysiological stress response. As infants grow and gain experience, they learn to distinguish actual threats from “accidents” or non-threats, typically by age three or four. A unique characteristic of humans is the ability to imagine, guess, or interpret the intentions of another person, commonly called “theory of mind.” Many animals have information processing systems that include interpretation of another animal’s cues as threatening or not; only humans can contemplate that the other can have intentions. We learn that not all disruptive cues are intentional threats; sometimes, the other person causes a bad outcome for us inadvertently through an accident or even for a good intention (e.g., a physician gives a painful shot to a child to provide immunization). The distinction between another’s intentions and outcomes is hard for young children to realize but an important lesson to learn.

Both skills and biases differentiate across children. Not all children learn the skill of accurate interpretation at the same pace. Particularly important are skills of recognizing different emotions, both in other people and in the self. Five-year-old children readily recognize when another person is displaying a positive emotion (“good”) versus a negative emotion (“bad”), but they have only a rudimentary ability to recognize the differences among negative emotions (e.g., “sad” versus “mad”). Their interpretation of their own internal cues (such as the meaning of an accelerating heart rate) is also developing during early childhood.

Innate tendencies and specific socializing experiences lead children to acquire unique heuristic biases in interpretation. Poor Charlie Brown always interpreted Lucy’s words as indicating she would benignly hold the football in place while he tried to kick it; he was always surprised when she picked up the ball as his toe came close to touching it. At the other extreme, early adverse experiences lead some children to become biased toward interpreting others’ intentions as hostile.

As with attention, numerous constructs are captured by this processing step, depending on the stimulus, including attribution of another’s intention under conditions of ambiguity, skill of interpreting another’s intention accurately under conditions of clarity, and skill at recognizing another’s emotional state. All three constructs have been associated with children’s aggressive behavioral problems.

The most replicated finding is the association between a bias to make an attribution of hostile intent and aggressive behavior problems. As I described in my dissertation study (Reference DodgeDodge, 1980), I presented hypothetical cartoon stories to aggressive and nonaggressive boys in which they imagined being the object of an ambiguous provocation by a named classroom peer, who was known to be aggressive or nonaggressive. As Figure 5 depicts, the aggressive group had a .25 probability of attributing hostile intent to the peer, in contrast with a significantly lower probability of .17 for the nonaggressive group. The bias held both when the peer was a known aggressive boy and when the peer was a known nonaggressive boy, it held across all ages studied (7, 9, and 11), and the magnitude of effect increased as boys got older.

Figure 5 Hostile attributional biases by and toward aggressive children.

(adapted from Reference DodgeDodge, 1980)

One hypothesis was not supported: I speculated that if aggressive boys are simply disorganized, the magnitude of the association between their attribution of intent and their behavioral response would be smaller than that for nonaggressive boys. No evidence emerged to support this hypothesis, indicating that the processing pattern displayed by aggressive boys is rule-governed and biased rather than arbitrary and disorganized.

To enhance the realism of the stimuli, Reference Dodge, Pettit, McClaskey and BrownDodge et al. (1986) videorecorded six vignettes with child actors in which one child actor ambiguously provoked another child (e.g., knocked his puzzle over). Child research participants watched the vignettes and made interpretations about the provocateur’s intentions. Later, the same children were asked to play with unacquainted peers (who were in fact child actors), in which one peer ambiguously knocked over the child’s puzzle. The rate at which a child attributed hostile intent to the video stimuli predicted the likelihood that the child would react to an actual provocation with retaliatory aggression. Since those first studies, over 400 published reports have replicated the basic finding. At least three meta-analyses (Reference Orbio de Castro, Veerman, Koops, Bosch and MonsshouwerOrbio de Castro et al., 2002; Reference Tuente, Bogaerts and VelingTuente et al., 2019; Reference Verhoef, Alsem, Verhulp and CastroVerhoef et al., 2019) concluded the finding is robust, not accountable by any third variable, and shows a pattern of hostile attributions driving reactive aggressive behavior.

Are aggressive children less accurate than others at identifying others intentions? The second interpretation construct is accuracy at detecting others’ intentions. Reference Dodge, Murphy and BuchsbaumDodge et al. (1984) created 300 videorecorded vignettes, each depicting a provocation in which the peer actor’s intent varied as: (1) hostile; (2) prosocial; (3) accidental or benign; (4) ambiguous; or (5) merely present, that is not involved at all in the provocation. Socially rejected, aggressive behavior-problem children were less accurate than average and popular nonaggressive children overall. To replicate the findings, Reference Dodge, Pettit, McClaskey and BrownDodge et al. (1986) presented six of the vignettes (two prosocial intent, two hostile intent, and two accidental) to aggressive children and matched nonaggressive children and found the nonaggressive children were more accurate than aggressive children at identifying prosocial intent, whereas the aggressive children were more accurate than the nonaggressive children at detecting truly hostile intent.

The third construct in this processing step is emotion recognition. It is so important to learn to recognize emotions in others to get along with them and in recognizing one’s own emotions to regulate them. Many young children can tell whether they are feeling “good” versus “bad” but are not able to tell whether their bad feeling is “sad” or “mad.” Numerous investigators have created instruments to assess emotion recognition. One of the most-used instruments is a set of photographs developed by Tottenham depicting facial expressions of different emotions (Reference Tottenham, Tanaka and LeonTottenham et al., 2009). Reference Acland, Peplak, Suri and MaltiAcland et al. (2023) presented photographs depicting happiness, sadness, anger, and fear to assess children’s skills at accurately recognizing emotions. The findings are clear: aggressive children are less accurate than other children at recognizing fear and sadness in others’ faces. Reference Schwenck, Gensthaler and RomanosSchwenck and colleagues (2014) created a morphing task in which clinically diagnosed conduct-disordered girls (without callous-unemotional traits, meaning they were likely reactively aggressive rather than cold bullies) and matched controls were presented 60 film clips, each 9 seconds long, in which a neutral face gradually morphed into a face expressing a particular emotion. Girls were asked to press a key as soon as they recognized the emotion and to label it. The conduct-disordered girls were slower than the controls to recognize happy, sad, and fearful faces, but not slower to recognize anger. They were also less accurate than the controls at recognizing sadness and fear but not anger.

Reference Dodge, Laird, Lochman and ZelliDodge et al. (2002) administered Reference Ribordy S., Camras L., Stefani and SpaccarelliRibordy et al.’s (1988) Emotion Recognition Questionnaire instrument, in which the interviewer presented each of 16 situational vignettes about an episode that happened to another child (four each depicting each of fear, sadness, anger, and happiness) and asked children to identify the emotion the peer was experiencing. Poorer accuracy in identifying fear and sadness predicted children’s aggressive behavior problems as rated by both parents and classroom teachers.

It is important not only to recognize emotions in others but also to recognize emotions in oneself. Reference Greenberg and KuscheGreenberg and Kusche (1993) developed the Interview of Emotional Experience to ascertain the emotions children report they would experience in response to situations that normally elicit different emotions of fear, sadness, anger, and happiness. With the Fast Track sample, Reference Dodge, Laird, Lochman and ZelliDodge et al. (2002) found that problems in recognizing sadness and fear in oneself predicted children’s aggressive behavior problems.

3.2.3 Step 3: Energized Emotional Experience

Through neural associations potentiated across generations and laid down through acquired experiences, interpretation of social cues ignites emotional reactions. Phenomenologically, when we interpret a social cue, we construct an emotional response, perhaps fear, joy, anxiety, or anger. There is considerable debate in the science of emotion about whether emotions are prewired responses to specific cues or are dynamically constructed during social interaction (Reference Rosen and LevensonRosen & Levenson, 2009). Either way, an emotion is not merely a passive and epiphenomenal event: It is accompanied by whole bodily reactions. Emotions are neural calls to action. Upon interpreting a social cue, the prefrontal cortex signals other regions of the brain (through neural synapses) and the peripheral nervous system to secrete testosterone, release cortisol into the bloodstream, change the heart’s rate of beating, and alter breathing, depending on the interpretation.

Those neural pathways follow general human maps but also form large individual differences. The same interpretation of a cue (perhaps as a threat) could trigger anger in one child and paralyzing hopeless fear in another child. Emotions motivate behavior; without emotion, we wither. The function of an emotion is to set an energized direction for a behavioral response, which we call a goal. Goals and goal-setting (even unconsciously) are thus essential parts of this step of processing.

Reference Rose and AsherRose and Asher (1999) presented hypothetical social conflict situations to children and asked them what their goal would be if the conflict happened to them. Children who consistently generated a goal of “revenge” toward others, even toward their own friends, had the most problems in relating to peers, including aggression.

In our lab, Reference Crozier, Dodge and FontaineCrozier and colleagues (2008) led a study of the effect of experienced provocation on the sympathetic nervous system. We brought CDP participants into the laboratory and tracked children’s heart rate while the children watched videorecorded vignettes and answered questions. The average profile of heart rate changed while children watched the vignettes, as shown in Figure 6. As the interviewer told the children to anticipate the beginning of a video, their average heart rate increased, but when the video began, their heart rate declined as they attended to the story. This decline in heart rate as children focus attention is a well-documented response pattern. At the moment the video depicted a provocation (such as the focal child being bumped in the back), children’s heart rates soared. Over time, heart rates returned to baseline. This pattern of cardiac response to provocation (sharp increase followed by gradual return to baseline) held for both aggressive and nonaggressive children, but two differences were apparent, particularly among boys. First, aggressive children, relative to nonaggressive children, responded with greater psychophysiological reactivity (higher spikes in heart rate) immediately following the presentation of the provocation. Second, their return to heart rate baseline took longer. Further analyses indicated that children who responded with these two patterns of heart rate reactivity (high spike following provocation and long lag until return to baseline) also displayed biased social information processing responses. The combined psychophysiological and mental responses indicate a defensive mindset.

Figure 6 Mean second-by-second heart rate reactivity in response to provocation for aggressive and nonaggressive boys.

(adapted from Reference Crozier, Dodge and FontaineCrozier et al., 2008)

Also in our lab, Reference Carré, Iselin, Welker, Hariri and DodgeCarré et al. (2014) studied the effect of experienced provocation on the endocrine system. The hypothalamic–pituitary–gonadal (HPG) axis, and its hormonal end product testosterone, is associated with aggressive behavior. In particular, acute fluctuations in the release of testosterone during social interactions predict aggressive responding. We hypothesized that rapid testosterone release (measured from saliva samples) would be part of the whole-body defensive mindset response pattern. Carré et al. brought into the laboratory the Durham participants in the Fast Track study and asked participants to play a computer game with a peer who was in another room. There was no actual peer; rather, Carré et al. simulated the peer’s responses. The goal of the game was to earn points by pressing a button as rapidly as possible or by “stealing” points from the peer by pressing a different button. Throughout the game, points were “stolen” from the participant by the peer at random intervals. Carré et al. collected saliva samples from the participants before they began the game and after each of three 10-minute intervals. As expected, following the initial provocation by the fictitious peer, testosterone levels increased by 7% (among participants who had not experienced clinical intervention). Second, a higher magnitude of testosterone increase predicted the participants’ higher rate of reactive aggressive behavior (stealing points) toward the peer. Third, the group of participants who had been randomly assigned to receive the Fast Track intervention showed a pattern of less release of testosterone in response to provocation and less reactive aggressive behavior toward the peer (see Section 6 on intervention).

3.2.4 Step 4: Response Generation

Once children interpret social cues and experience emotions and goals, one or more possible behavioral responses is called to mind from a memory bin of behaviors. Investigators have called this process “response generation” when it occurs outside of consciousness and “social problem solving” when it is a conscious act, and they have measured both the number of responses children generate and the quality of those responses. Hundreds of studies have shown that children who generate fewer assertive and more aggressive solutions than do average peers are relatively likely to develop aggressive behavior problems.

Reference Dodge, Pettit, McClaskey and BrownDodge et al. (1986) assessed children’s response generation patterns in several kinds of situations. They began with a non-conflict situational task, trying to initiate entry into a group of peers already at play, by creating video vignettes of unacquainted peers at play and asking each child, “Tell me all the ways you could get other kids to let you join their group.” Later, they brought each child to a different room where two strange peers were already at play on a puzzle and asked the child to try to join the group, and they coded children’s peer group entry strategy behaviors. Dodge et al. were able to predict children’s actual likelihood of behavioral success in joining the peer group based on the proportion of responses to the problem-solving task that were aggressive and scored as incompetent, as well as the total number of solutions they generated. Next, Dodge et al. presented hypothetical provocations to children and asked them to generate possible responses to being provoked. In a separate session, they brought the same children into the laboratory to play with an unacquainted peer (actually, a child actor) who provoked them, and they observed children’s actual behavioral response. They found that the more reactive aggressive behavior responses children generated in the problem-solving task, the greater the likelihood that children reacted aggressively when actually provoked.

Reference Bookhout, Hubbard, Zajac, Mlawer and MooreBookhout et al. (2021) expanded on these assessments with novel stimuli and research methods. They asked children to imagine being provoked by a peer and to generate possible responses. They found a positive correlation between the number of aggressive responses a child generated and directly observed rates of aggressive rule-breaking behavior on the playground.

3.2.5 Step 5: Response Evaluation and Decision Making

Not every behavioral response that is called up from memory gets enacted. Thankfully, children learn the ability to withhold an impulse. Children might feel anger and have an impulse to hit back (or grab a candy bar off the shelf of a store or spit on a baby brother), but people develop the capability to hold in mind a possible response and evaluate it before making a decision to act upon it. We learn impulse control. People also develop the ability to observe themselves before responding, called “executive function.” This step of processing is a major component of emotion regulation.

Reference DodgeFontaine and Dodge (2006) articulated a formal model of response evaluation and decision making (RED) in real time. The model proposes the idea of minimal acceptability threshold; that is, a possible behavior that is called to mind (such as hitting a peer in the head) is quickly evaluated by the brain for acceptability; if the response is above a minimum threshold, it is selected for enactment. If it is below threshold, the person seeks an alternative response. Numerous factors may push this threshold downward, such as high blood alcohol level, fatigue, angry mood, or a general dispositional propensity to act quickly. Other factors may push the threshold upward, such as authority figures who are ready to pounce on one’s mistake. Fontaine and Dodge also asserted that the threshold is higher at the beginning of a social interchange, and if no generated response reaches the minimally acceptable threshold, a child might “lower the standard” and resort to a response that had previously been evaluated as unacceptable.

How does a child decide whether a response is above threshold of acceptability in expected outcome? It is a risk-reward calculation. Fontaine identifies many domains of outcomes that children could consider in making a decision, including instrumental consequences (“Will I get my ball back?”), social consequences (“Will they like me?”), intrapersonal consequences (“Will I feel good inside myself?”), and moral consequences (“Is this what I believe is right?”). Not only are the expected consequences evaluated, but each consequence is given a valuation. Children might consider that a behavior will cause others to dislike them but might not care much about that consequence.

In real time, children consider the likely evaluated and valuated consequences of each generated possible response in each of multiple relevant domains, do some mental calculus, and select the best response for implementation. If no response exceeds a minimal threshold of acceptability, children might go back to the memory bin for more possible solutions, experienced as “thinking” and problem solving. Fontaine hypothesized a mathematical formula to guide decision making, in which children (not consciously, maybe not even in reality but heuristically) compute an acceptability score for a possible behavioral response that is the sum of the various evaluated consequences, weighted by the value given to each consequence.

Fontaine and Dodge represented the evaluation process as:

where OS_i is the overall acceptability score for outcome i, and is a function of Oe_i which is the expectancy score for outcome i multiplied by Ov_i which is the valuation score for outcome i.

In a study of 259 adolescents, Reference Fontaine, Burks and DodgeFontaine et al. (2002) found that children’s positive evaluations of aggressive responses, negative evaluations of nonaggressive responses, and high valuations of self-serving consequences predicted heightened levels of aggressive behavior problems. In a replication with the CDP sample, Reference Fontaine, Yang, Dodge, Bates and PettitFontaine et al. (2008) found that these aggressogenic RED processes predict the growth of aggressive behavior across adolescence.

One general decision-making heuristic that has been hypothesized to characterize aggressive children is the overvaluation of immediate outcomes to the neglect of long-term outcomes. One may be familiar with Mischel’s “marshmallow test” in which young children are presented with a choice to grab one marshmallow immediately or multiple marshmallows later if they can wait a short period of time. So, too, during certain social encounters, some children inordinately value immediate consequences (e.g., hitting a peer who is playing with a child’s toy) over longer-term consequences (e.g., staying out of trouble with the teacher). Economists call the tendency to prioritize immediate consequences “delay discounting.”

Delay discounting is not always a conscious, reasoned decision by a rational child. Perhaps we all can gain some degree of empathy or at least insight if we see a similarity between children’s impulsive hitting and adults’ problems with impulse control in certain situations. Even the best of us sometimes caves into the temptation of a piece of chocolate cake. We know we should not take that bite, but in one fleeting moment the allure of the taste is so powerful we ignore consideration of later costs. Alcoholics face this problem each time they see a drink. Cigarette smokers know they should quit, but when they smell smoke or are stressed, they tell lies to themselves that one more cigarette will not kill them. These are otherwise responsible adults who fall down in certain situations. Why should aggressive children be any different?

Economists may have labeled the phenomenon of delay discounting, but psychologists have identified life experiences and eras that control it. Reference BelskyBelsky (2014) hypothesized that early life experiences of severe and constant threat predispose children to accelerate pubertal timing and to discount long-term consequences, partly because their “long-term” may never come to pass if they die before growing up. Reference KotlowitzKotlowitz (1991) recounted an interview with children on the south side of Chicago who told him, “If I grow up, I’d like to be a bus driver.” Kotlowitz recounted the narrative as: “If, not when.” No doubt this schema about life influences the child’s mental calculus about whether to invest in homework or after-school play. In a study of many cultures around the world, Reference Steinberg, Icenogle and ShulmanSteinberg et al. (2018) identified an era during biological adolescence when teenagers’ immediate reward calculations go berserk: “I want, I want, I want.”

Impulsive responding without considering consequences at all and over-consideration of rewards (and the complementary neglect of consideration of punishment) are two additional components of decision making that predict children’s aggressive behavior problems. In our lab, Reference Yechiam, Goodnight and BatesYechiam et al. (2006) proposed and tested a formal model of responding to laboratory gambling tasks (known as “go no-go” or passive avoidance tasks). In one task, the child is presented a series of numbers on a screen and is asked each time to press a button if the number is “good.” If correct, the child receives 25 cents. If the child presses a button for a “bad” number, the child loses 25 cents. The child receives neither punishment nor reward for not pressing a button. Across 90 trials, Reference Yechiam, Goodnight and BatesYechiam et al. (2006) scored the child’s responsiveness to rewards (button pressing for good numbers to receive 25 cents) and impulsivity (errors of commission when the child should have withheld button-pressing in response to bad numbers) and found that each factor predicted both teacher- and self-reports of externalizing behavior problems, even controlling for intelligence and family income.

3.2.6 Step 6: Behavioral Enactment

Ultimately, a child selects a behavioral response. The final step in processing is to transform a selected behavioral response into behavioral actions. This step might seem trivial, but important motor and verbal skills are involved and factors such as anxiety can impair performance. Children might decide on an assertive behavioral response (e.g., in response to a bullying peer who took the child’s bicycle for a ride, the child might decide to ask for the bike back), but meek performance could reduce the likelihood the behavior reaches its goal. Modeling of behavior, rehearsal, and practice can improve the ease of behavioral enactment and lead to more rapid and relaxed performance. Individual differences in enactment skill may be large and could determine success.

Relatively few studies have examined this step. Reference Dodge, Pettit, McClaskey and BrownDodge et al. (1986) brought aggressive and nonaggressive children into the laboratory for interviews, during which the interviewer asked children to act out different ways of responding. For example, for each of four hypothetical peer group entry situations, the interviewer said,

For each of four provocation situations, children were asked to enact a competent-assertive response to being provoked, such as asking a peer to return a toy they had taken from the child. The idea was to control all decisions about what behavior to enact and to focus solely on behavioral and verbal skills. Children’s competence in enacting each response was scored by the interviewer on a 5-point scale. The aggressive group was rated as less competent at enacting competent-assertive responses than the nonaggressive group.

5.1.1 Physical Maltreatment Leads to Defensive Mindset

The strongest findings for the antecedents of defensive mindset come from research on children who had experienced physical abuse in the first years of life.

Reference Pollak and KistlerPollak and Kistler (2002) hypothesized that early experiences of child abuse would lead children to become hypervigilant to anger cues in facial expressions. They created novel stimuli using morphed images of faces depicting emotions of “happy,” “sad,” “fearful,” and “anger,” paired emotions, and morphed the faces across each of 10 computer-generated images so that a clearly angry face gradually morphed into an ambiguous expression and then into a fearful face (or a sad face or a happy face). They presented the morphing images to abused children who had been identified through the child welfare system, as well as matched non-abused children, and asked them to identify each new facial expression as quickly as they could. Children who had experienced physical abuse overidentified morphed faces as depicting anger.

In a series of studies, Pollak and Cicchetti showed that children exposed to physical abuse come to display an array of defensively oriented patterns of social information processing. Abused children show enhanced perceptual processing as indexed by an early- to mid-latency event-related potential (ERP) component (P260) while watching angry faces (Reference Curtis and CicchettiCurtis & Cicchetti, 2011) and heightened P3b ERP in response to angry faces compared to other emotions (Reference Pollak, Cicchetti, Klorman and BrumaghimPollak et al., 1997). Abused children quickly identify angry facial expressions with less perceptual information (Reference Pollak and SinhaPollak & Sinha, 2002) and fewer expressive cues (Reference Pollak, Messner, Kistler and CohnPollak et al., 2009), and they allocate more attention to vocal expressions of anger (Reference Shackman and PollakShackman & Pollak, 2005). Reference Hosseini-Kamkar, Farahani and NikolicHosseini-Kamkar et al. (2023) showed through meta-analysis of 83 functional magnetic resonance imaging (fMRI) studies of 5,242 participants that exposure to adverse life experiences, including maltreatment, is reliably associated with higher amygdala reactivity and lower prefrontal cortical reactivity across a range of task domains, suggesting how early abuse might have an enduring impact on future behavior. Reference Keil and PriceKeil and Price (2009) assessed hostile attributional bias through hypothetical provocation scenarios in abused children identified through the child welfare system and matched non-abused children. The abused group responded with higher levels of hostile attributional bias.

Most studies of abused children identify them through the child protective services (CPS) division of county or state governments. The advantage of this sampling method is assurance that an independent professional (almost always, a social worker) has investigated and verified the fact of physical abuse or neglect. The disadvantages, however, are that this might be a select group of abused children who have come to the attention of officials because of racial bias, particularly egregious abusive behavior, or capriciousness of reporters; if so, these children might not represent the population of abused children. Also, it is plausible that the experience of being “processed” in the child welfare system has its own deleterious effects (e.g., separation from supportive family members, change of schools, blame for having been reported) that are at least partially responsible for children’s later outcomes.

Our CDP research team adopted a complementary approach by sampling a community-representative group of 4-year-old children from three communities (Nashville, TN; Knoxville, TN; Bloomington, IN) and conducting interviews to identify children who had been abused in the past but had not necessarily come to the attention of public officials. We conducted conversational, open-ended, interviews with mothers and fathers privately about past experiences the children had. We were careful to warn parents that if they revealed current abuse by themselves or their partner, we would be obliged to report the exchange to the local CPS division. Nonetheless, we did hear about several cases of current possible maltreatment (usually, a mother wanting to report the behavior of her partner) and followed through with reporting. Most cases, however, were in the past and did not qualify as reportable events according to ethical self-scrutiny, local CPS authorities, and the prevailing law at the time (NB: laws vary across states and many have changed since 1986).

Trained CDP interviewers reliably coded children as having been maltreated or not. Altogether, 11.9% of the sample (69 out of 585) met criteria as having been abused in the first 5 years of life, a high figure but one that is probably an underestimate of the true experiences of children in the United States (Reference Dodge, Bai and GodwinDodge et al., 2022). Independent interviewers assessed social information processing patterns in these children as they made their way through school, at ages 5, 6, 7, 8, 13, and 16, and created an overall defensive mindset latent score across measures and ages. The third and fourth columns of Figure 11 show the mean standardized defensive mindset construct score for the maltreated group was two-thirds of a standard deviation higher than the mean score for the non-maltreated group.

Figure 11 The effect of early physical abuse and chronic peer social rejection on the development of defensive mindset

(adapted from Reference Dodge, Bai and GodwinDodge et al., 2022.)

We then replicated the entire set of analyses with the Fast Track sample (Reference Dodge, Bai and GodwinDodge et al., 2022). History of maltreatment was assessed through parent interviews at age 5, and 12.8% of the sample met criteria as having been abused. Social information processing patterns were assessed at ages 6, 7, 8, 9, 10, 11, and 14, to create a robust latent score for defensive mindedness. Replicating the CDP, one can see in the left-most columns of Figure 11 that the maltreated group had a defensive mindset score that was one-third of a standard deviation larger than the mean for the non-maltreated group. Maltreated children are more likely than others to develop defensive mindset with a replicated effect size somewhere between one- and two-thirds standard deviations.

5.1.2 Abused Children Grown up

What happens to abused children as they grow up? A committee of the National Research Council (Reference Petersen, Joseph and FeitPetersen et al., 2014) adopted the terminology of “cascading effects” to describe how the early experience of physical maltreatment has an enduring effect on children’s social cognition, which cascades into long-term effects on children’s maladaptation to school, peers, and, eventually, the adult world.

Our team (Reference Dodge, Price, Bachorwski and NewmanDodge et al., 1990; Reference Weiss, Dodge, Bates and PettitWeiss et al., 1992) followed the 4-year-old boys and girls of the CDP and Fast Track across childhood to track outcomes for abused children. We took into account prior factors (including socioeconomic status, single parent household, child temperament, and family ecology) that might be correlated with outcomes in order to isolate the effect of maltreatment. At the first time point at age 5, direct observers found that abused children (relative to non-abused peers) displayed more frequent aggressive behaviors toward peers on the playground, teachers rated them as more aggressive, and peers rated them as more aggressive. In late adolescence, Reference Lansford, Miller-Johnson and BerlinLansford et al. (2007) gathered court records and interviews to find that 34% of abused children had been arrested, in contrast with just 20% of non-abused children, with higher rates of arrest for both violent crimes and nonviolent crimes. At age 21, structured interviews indicated that the abused group was much more likely to have engaged in violence in their romantic relationships (both as perpetrator and as victim; Reference Pettit, Lansford, Malone, Dodge and BatesPettit et al., 2010), serious violent delinquency (e.g., physical assault, use of weapon, sexual assault, gang fight), and to have been fired from a job (Reference Lansford, Miller-Johnson and BerlinLansford et al., 2007). Only 14% of the abused group graduated from college, in contrast with 30% of the non-abused group. By age 24, among girls, the abused group was more likely than the non-abused group to engage in substance abuse, and, among boys, the abused group was more likely than the non-abused group to engage in violent crime (Reference Lansford, Malone and DodgeLansford et al., 2010).

By age 26, the life course outcomes for the formerly abused children looked grim. At this age, Reference Lansford, Godwin and McMahonLansford et al. (2021) relied on self-report and administrative records and interviewed a peer or partner to get a fresh objective perspective. Only 53% of the abused children had become employed full time, in contrast with 69% of the non-abused group; and 20% of the abused group was receiving public assistance, in contrast with 10% for the non-abused group. The abused group had worse overall health and more frequent risky sexual behaviors. They were twice as likely to have been convicted of a crime in the past 12 months. Diagnosed psychopathology among the abused group continued to be high, with higher rates (compared with the non-abused group) of disorders for antisocial personality, attention deficit hyperactivity, and anxiety. As adults at ages 32–34, the outcomes had not gotten any better (Reference Dodge, Bai and GodwinDodge et al., 2022). The formerly abused group was more likely than the non-abused group to have been incarcerated and for longer periods, less likely to have graduated from college, less likely to be employed, less likely to be married, more likely to be on government assistance, and more likely to have been victimized by others.

The replicated findings are clear. Children who had experienced physical abuse in early life are at higher risk than others to engage in violent behavior and to experience a diverse array of tragic outcomes including substance abuse, failure in establishing and keeping social relationships, and dependency on government and others. Not all abused children drifted into lives of violent crime; some never became aggressive but instead failed to get “on track” in school, employment, and life. Early abuse derailed them, disorganized their relationships, and debilitated them. These associations are probabilistic. Not all abused children have poor outcomes, so the search for mediators and moderators is crucial in the quest for preventive intervention.

5.1.3 Defensive Mindset Mediates the Effect of Maltreatment on Adult Outcomes

How does early childhood maltreatment wreak such awful lifelong outcomes? The development of a defensive mindset, depicted in Figure 12, may be a major mediating process to account for how early adversity leads to long-term dysfunction.

Figure 12 A model of how early adverse experience leads to maladaptive outcomes through the acquisition of a defensive mindset.

To demonstrate mediation statistically, Reference Baron and KennyBaron and Kenny (1986) proposed that four empirical conditions must be met. First, the initial driver (here, the experience of physical abuse as rated by a trained interviewer of mothers and fathers at child age 4) must predict the hypothesized outcome (here, children’s frequency of aggressive behavior at age 5 as measured by direct observations, peer nomination, and teacher ratings). Second, the driver must predict the hypothesized mediator (here, defensive processing measured by hypervigilance to threat, hostile attributional bias, retaliatory response generation, and positive evaluation of retaliation). Third, the mediator must predict the outcome. Fourth, in a regression model, the magnitude of the relation between the driver and the outcome must be reduced when the mediator is entered into the model.

In 1990, the CDP team (Reference Dodge, Price, Bachorwski and NewmanDodge et al., 1990) hypothesized that the experience of physical abuse in the first 5 years of life would drive children to become aggressive toward their peers and the effect would be mediated through the development of defensive patterns of social information processing. All measured processing steps were empirically supported as mediators.

What about the long-term? The CDP team continued to test whether childhood defensive mindset mediates the effect of early abuse on children’s maladaptive outcomes as they get older. Reference Dodge, Pettit, Bates and ValenteDodge et al. (1995) found that defensive processing partially mediates the effect of early abuse on aggression in later elementary school. Reference Dodge, Bai and GodwinDodge et al. (2022) found that a comprehensive and highly reliable construct measure of defensive mindset as assessed during childhood partially mediates the effect of early abuse on a range of dysfunctional outcomes at age 33, including a large number of days being incarcerated, a high rate of externalizing psychopathology, low current income, a low level of ultimate education, being victimized by others in adulthood, and not being married. These mediational tests for all of the same outcomes were replicated with participants in the Fast Track project at age 32 (Reference Dodge, Bai and GodwinDodge et al., 2022). Although these tests show replicated and robust mediational processes, it must be noted that the magnitude of mediation was modest. Typically, about a quarter of the variance in adult outcomes was accounted for by defensive mindedness, indicating that other processes must also be involved.

6.1.1 The Fast Track Intervention Components

The Fast Track intervention is described in detail in a volume by the Conduct Problems Prevention Research Group (2019). During planning, the Fast Track team addressed two questions: 1) what do we want aggressive children to learn? and 2) through what modalities could we accomplish that learning?

1. What do we want aggressive children to learn? Based on the body of research and clinical experience, we know a crucial goal of intervention with highly aggressive six-year-old children must be to teach them to slow down and to apply executive control over their emotional reactions before they respond with impulsive aggression. How could we teach slow processing? The team adapted a metaphor from Weissberg, the stop light, and its three steps of Stop, Think, and Go, as depicted in Figure 13, and developed a curriculum around that metaphor. The curriculum was implemented by bachelor-level trained Educational Coordinators (ECs).

Figure 13 The stop light metaphor to teach self-control and problem solving.

(adapted from Reference Weissberg, Caplan, Bennetto and JacksonWeissberg et al., 1990)

The first part of the curriculum is to help children learn that when they encounter a problem, any kind of problem, they should first STOP, slow down, and calm down. ECs played games and role-played stories to teach children what a “problem” is; that is, a problem is any time they feel unhappy, stuck, hurt by someone else, or in trouble, or any time someone else seems to have a problem with them. ECs helped aggressive children identify strategies for slowing down, such as counting to ten, taking a deep breath, feeling their heart beat more slowly, and looking away. They played games to help the children practice, they observed them in free play with peers, and they seized on naturally occurring problem events to practice slowing down. Although there are many components to competent social information processing that would take many years to master, ECs declared a success if during the first year aggressive children could simply slow down.

Once ECs got children to slow down, they directed them to go to the yellow light (depicted in the figure in gray shade), where they were encouraged children to engage in other mental processes, starting with recognizing one’s own and others’ emotions. They began to teach children to “see things differently,” that is, to consider multiple divergent interpretations of their problem. Here is where hostile attributional biases come front and center: Could it be that the other child did not see you before he ran into you? Might it be that the teacher was not trying to make fun of you but was simply trying to get all children to be quiet?

As children learned to consider multiple interpretations of their situation and to understand (and regulate) their emotions, ECs encouraged them to begin a problem-solving process. What might they be able to do if that particular problem occurred to them? ECs encouraged expansive thinking to get children out of their old habits. They taught them the difference between thinking about what they could do and actually doing it – impulse control. They taught children to consider more than one way of reacting before deciding to react. And they taught children to make the decision they believe is best for them.

When children learned the process of making a behavioral decision, ECs taught them to go to the green light (depicted in the figure as dark gray) to try it out. The process does not end there. While at the green light, ECs taught children to observe themselves and others. Did that decision solve their problem? Are they calmer and happier? Have others stopped bothering them? If not, they would teach them to start over by going back to the red light.

2. Through what modalities can we intervene to help aggressive children? The intervention team was realistic enough to realize that helping aggressive children become less defensive minded would require a multi-pronged, all-out effort over a long time. It had taken years for children to develop a defensive mindset; it would take years to change it. The team committed to intervening with these children for 10 years, through high school. The intervention began in first grade (age 6) on a weekly basis, moved to biweekly sessions the next year, and monthly sessions thereafter. The components were as follows:

Friendship Groups. Children were invited to attend Friendship Groups of four to eight children held after school or on weekends that targeted children’s social-cognitive skills of accurate awareness of others’ intentions, emotional expression and emotional understanding, emotion regulation, anger control, and social problem-solving skills. EC leaders were assisted by para-professionals to use role-playing, videotaping, and coaching methods that included direct instruction and modeling to build skill concepts, behavioral rehearsal, practice, and feedback. As the children moved into middle school (ages 11–14), group sessions populated only by high-risk children were discontinued because of the emerging literature on the dangers of deviant peer influence that could occur in volatile peer groups and replaced them with mentoring sessions and individual planning meetings between the EC and the children.

PATHS®. Classroom teachers taught a grade-level version of the PATHS® (Promoting Alternative Thinking Strategies) Curriculum (Reference Greenberg and KuschéGreenberg et al., 2011), 2–3 times per week in Grades 1 to 5 (ages 6–10) to all classmates of aggressive children. This curriculum reinforced skills taught in the Friendship Groups and brought the skills to the entire peer group. The curriculum targeted prosocial skills (helping, sharing, cooperating), self-control, emotional awareness and understanding, and social problem solving.

Peer pairing. Prior research had suggested that including classmates in social skills training plans could give aggressive children a chance to practice what they had learned in Friendship Groups with other children and promote positive changes in peers’ behavior and attitudes toward aggressive children (Reference Bierman and FurmanBierman & Furman, 1984). ECs provided a supervised half-hour play session with classroom peers each week.

Academic tutoring. Although social skills were the centerpiece of the Fast Track intervention, the team knew that aggressive children need help in other domains as well, so they developed a phonics-based tutoring program to teach reading skills based on work by Reference Coie and KrehbielCoie and Krehbiel (1984). Aggressive children received three tutoring sessions per week in the first year and extra tutoring in subsequent years based on need.

Parenting groups with home visits. The team knew they needed parents’ support, so they delivered Parent Groups that were based on ideas by Reference Forehand and McMahonForehand and McMahon (1981) and which were delivered while the children were attending Friendship Groups. These groups were led by Family Coordinators (FCs) who focused on improving positive parent–child interactions, reducing harsh and punitive discipline, and increasing consistent limit setting. FCs told parents that in the children’s Friendship Groups, they were teaching them to slow down through the stoplight metaphor; inevitably, at least one parent would comment, “Well, we could all use that help.” So the FCs incorporated the stoplight tool in teaching parents to respond to their children in more deliberate, mindful ways; this practice not only helped parents in their discipline behavior, it enabled parents to use the same language and terms that the team was teaching their children.

The team supplemented Parent Groups with individualized home visits, which provided the FCs an opportunity to understand each family’s unique situation and to build a strong working alliance with each primary caregiver.

Parent–child sharing. The team followed Parent Group and Friendship Group sessions with a half-hour parent–child sharing session created by Lochman that had two goals: 1) to foster positive parent–child relationships; and 2) to provide parents a chance to practice the skills introduced in Parent Groups, with staff guidance.

Mentoring. The team added a mentoring program in the later years to provide a same-sex, same-race, community volunteer mentor for children who lacked an adult role model in their lives. Mentors were briefed in the goals of the intervention and encouraged to support the interventions, including the stop light model.

6.1.2 The Fast Track Randomized Controlled Trial

How could the team evaluate the impact of this massive intervention? We matched the complexity of the intervention with an equally complex randomized controlled trial. The intervention and the evaluation trial were implemented in each of three years in each of four communities (Durham, NC; Nashville, TN; rural central PA; and Seattle, WA), which afforded a diverse sample of children and multiple opportunities to hone and test the program. The team recruited 55 high-risk schools based on demographic characteristics about the crime and poverty rates of the neighborhoods they served. From the population of 9,594 kindergarteners in these schools, the team used a multiple-gating procedure to ask teachers (and then parents) to screen and select 891 (top 9% of total) highly aggressive children. These children were randomly assigned (based on the school cluster they attended) to receive the Fast Track intervention (n=445) or serve as a control (n=446). The targeted children were a high-risk group, mostly from backgrounds of disadvantage or trauma. Sixty-nine percent of the children were boys. Fifty percent were African American, 47% were European American, and 2% came from other groups. Fifty-eight percent lived with single parents, and 40% were in the lowest socioeconomic class. Proof of their high risk would come later when the team followed the control group into adulthood and found that, at age 26, without the Fast Track intervention, 69% ended up meeting diagnostic criteria as having a psychiatric disorder.

The team implemented the intervention with high fidelity and strong participation (CPPRG, 1999). In the first year, over 99% of the families assigned to intervention consented to participate in the intervention and received at least one session, and 72% attended more than half of each intervention component. In later years, participation declined modestly, but most families remained engaged.

To evaluate the impact of Fast Track on children, we appointed an independent research team to follow the children assigned to intervention (whether or not they actually participated) and the control children across time by interviewing children and their parents each summer and their teachers and peers each school year (for five years). We also retrieved school records and, later, criminal records. We successfully tracked 85% of the original sample into adulthood and continue to follow the original sample today as they reach their mid-30s and have children of their own.

6.1.3 The Impact of Fast Track on Aggressive Children

Impact on defensive mindset

The team’s initial focus was to determine whether the intervention accomplished its proximal goal to improve children’s social-cognitive skills, particularly the reduction of defensive mindset. The assessment protocol followed the social information processing model to test the effect of Fast Track on children’s processing patterns. Although not every outcome yielded a statistically significant effect, in general, the intervention was successful, although it took some time. Figure 14 depicts these effects.

Figure 14 Summary of significant effect sizes (in standard deviations) for the positive impact of random assignment to the Fast Track intervention on five steps of defensive processing.

(adapted from CPPRG, 1999, 2002)

At the end of the first grade year (CPPRG, 1999), the children randomly assigned to receive the Fast Track intervention demonstrated more skillful and less defensive processing than control children in four of the five steps of processing that had been assessed. Intervention-assigned children performed better at recognizing emotions in others (Step 1), coping with difficult emotions (Step 3), generating less aggressive solutions in social problem solving (Step 4), and decision making in response to social problems (Step 5), with effect sizes ranging from .12 to .36 standard deviations. In the first year, the intervention did not have any effect on children’s hostile attributional biases. The team realized this central component of children’s defensive mindset would take more time to break. After three years of intervention, the team found a positive impact: Fast Track-assigned children displayed lower (better) scores for hostile attributional bias than control children with a significant effect size of .14 (CPPRG, 2002). The team also found sustained positive impact of random assignment to the Fast Track intervention on reduction of components of defensive mindset and improvement of social competence in Grades 4 and 5 (ages 9 and 10) (CPPRG, 2004; Reference Dodge and GodwinDodge et al., 2013).

Impact on aggressive behavior in elementary school

Next, the team evaluated the effect of Fast Track on children’s aggressive behavior problems. Again, not all tests proved statistically significant, but, in general, the intervention was successful in reducing children’s aggressive behavior and improving social adjustment and well-being across the first five years, as shown in Figure 15.

Figure 15 Summary of significant effect sizes (in standard deviations) for the positive impact of random assignment to the Fast Track intervention on children’s aggressive behavior, social adjustment, and well-being during childhood.

(adapted from CPPRG, 1999, 2002)

At the end of the first year, objective playground observers coded more time in positive peer interaction and fewer aggressive-and-noncompliant behaviors for the intervention-assigned children than control children, with effect sizes of .27 and .31, respectively. Peer sociometric interviews revealed that intervention-assigned children received higher social preference scores and more nominations for prosocial behavior (effect sizes = .28 and .21, respectively). Classroom teachers rated the behavior change of intervention children much more positively than for control children (effect size = .53). School records indicated fewer special education diagnoses (mostly for behavior problems) for intervention than control children (effect size = .26). At home, objective observers coded greater warmth and less noncompliance/aggression in parent–child dyads of intervention families than control families (effect sizes = .23 and .22, respectively). The Parent Daily Report (based on four days of nightly telephone calls to parents) revealed fewer instances of misbehavior for intervention children than control children (effect size = .17), and parents rated the behavior change of their intervention children more positively than did parents of control children (effect size = .50).

At the end of third grade, the pattern persisted. Classroom teachers continued to rate the aggressive behavior problems as lower and behavior change across the year as more positive for intervention than control children (effect sizes = .19 and .27, respectively). Special education diagnoses continued to be lower for intervention children (effect size = .14). At home, the Parent Daily Report revealed fewer instances of misbehavior (effect size = .15), and parents reported more positive child behavior change (effect size = .20).

Impact on antisocial behavior in adolescence

As the children moved into adolescence, we all thought disaster had struck. It seemed that the intervention children (and their classmates) had “lost their minds” and become obsessed with deviance, independence, and misbehavior. Analyses of children’s skills and behaviors produced significant effects of the Fast Track intervention for only two (age 13 hyperactivity and delinquency behaviors) of the 17 variables tested (CPPRG, 2010).

When the children entered high school, however, significant positive impact of random assignment to the intervention had returned. Primary measures were the Self-Reported Delinquency (SRD) scale, as well as other measures such as diagnoses of psychiatric disorder. This instrument documents the number of times each of 25 delinquent acts, including property damage, theft, physical assault, and substance use, were committed in the past year. The intervention group had a mean delinquency score that was 26% lower than the mean score for the control group (CPPRG, 2007).

Next, we tested the primary hypothesis guiding the intervention that improvement in children’s social information processing patterns would mediate an indirect impact on adolescent antisocial behavior. The model and findings are depicted in Figure 16. As hypothesized, random assignment to Fast Track had significant positive impact on reducing children’s hostile attributional bias, aggressive response generation, and favorable evaluations of the consequences of aggressing. In turn, these processing patterns predicted antisocial behavior at age 14 (after grade 9) and significantly accounted for about 25% of the effect of Fast Track on antisocial behavior (Reference Dodge and GodwinDodge et al., 2013).

Figure 16 The Fast Track intervention targets the long-term prevention of delinquency by changing children’s defensive mindset.

(adapted from Reference Dodge and GodwinDodge et al., 2013). Figures represent path coefficients, with “*” indicating a significant path at p < .05

Impact on outcomes in adulthood

The Fast Track intervention continued for 10 years, through age 16. At that point, the program “graduated” children and wished them well. The research team has followed participants for another 20 years and continues following them today as participants have reached full adulthood, become employed and community citizens, and have families of their own. Intervention staff members have attended numerous weddings of “Fast Trackers,” celebrated their life successes, consulted when crises ensued, and even mourned several deaths. Some participants have spent time in prison, and some have become contributing citizens. Systematic analysis indicates the overall stability of positive impacts of the early intervention on adult life outcomes across a wide array of domains, as depicted in Figure 17 for crime and social competence and in Figure 18 for health and mental health.

Figure 17 The impact of Fast Track on crime and social competence.

(adapted from CPPRG, 2015)

Figure 18 The impact of Fast Track on health and mental health.

(adapted from CPPRG, 2015)

Generalizability of impact across groups

What is particularly striking about the effect of Fast Track is the consistency of its impact across the different groups of participants: The investigators observed a positive effect of Fast Track among each of the 13 sub-groups they studied: male and female; European American and African American; Cohorts 1, 2, and 3; Durham, Nashville, rural PA, and Seattle sites; and highest and moderate risk children. Figure 19 depicts the consistent impact on the most general measure in adulthood, being diagnosed with any psychiatric disorder (CPPRG, 2015).

Figure 19 The impact of Fast Track on psychiatric disorder in adulthood, by subgroups.

(adapted from CPPRG, 2015)

6.1.4 Complementary Interventions to Change Defensive Mindset

It is essential to replicate findings and to see whether other investigators replicate and extend findings. Reference Bierman, Welsh and HallBierman et al. (2022) conducted a randomized controlled trial of the Friendship Group component of the Fast Track intervention with 224 young children and found positive impact of assignment to intervention on improving social-cognitive skills, reducing externalizing behavior reduction, and improving peer relationships.

Other investigators have tested their own interventions to address defensive mindset with diverse populations of aggressive children, with encouraging results. Reference Hudley and GrahamHudley and Graham (1993) tested an intervention to alter the attributions made by African-American boys about peer provocations and found positive impact on reducing peer-directed aggressive behavior. Reference Fraser, Galinsky and SmokowskiFraser and colleagues (2005) created and tested the Making Choices: Social Problems Solving Skills for Children (MC) program with 8-year-old children and found that the program improves children’s social information-processing skills and reduces their aggressive behavior. Reference Van Bockstaele, van der Molen, van Niewenhuijzen and SaleminkVan Bockstaele and colleagues (2020) focused specifically on “hostile attribution bias modification training,” for aggressive adolescents in the Netherlands. They created a five-session group intervention to train adolescents to make more benign interpretations of ambiguous provocations. Random assignment to the intervention reduced hostile attribution bias and decreased levels of reactive aggressive behavior. There was no effect on adolescents’ proactive aggressive behavior.

Defensive mindset is one of the core constructs targeted in a broader set of interventions to improve children’s social emotional learning (SEL). These interventions include classroom curricula designed to teach essential social-cognitive skills such as emotion recognition and social problem solving. Meta-analyses of the effectiveness of SEL interventions have repeatedly shown that curricular interventions are effective in improving these skills and have positive impact on children’s social behavior that is sustained, sometimes, over many years (Reference Cipriano, Strambler and NaplesCipriano et al., 2023). SEL interventions are particularly effective as universal approaches to reach the full population of children and can be effective as adjuncts to more comprehensive interventions for high-risk children, which is how the Fast Track program used the PATHS® curriculum.

6.2.1 Young aggressive children are not necessarily destined to a life of violent crime.

A popular notion in the 1990s was that young aggressive behavior-problem children who disrupted their classrooms were “super-predators” (Reference Bennett, Dilulio and WaltersBennett et al., 1996), who were genetically and biologically defective and should be segregated into special education classrooms, thrown out of school into “juvie” detention centers, and then imprisoned for long sentences as teenagers and adults (Reference Bennett, Dilulio and WaltersBennett et al., 1996; Reference Herrnstein and MurrayHerrnstein & Murray, 1994). Intervention was thought to be futile, and society could be saved only by ridding these super-predators from schools and neighborhoods. It also happened that many of these children were African American. This idea was widespread in society and held by scholars and policy makers such as Reference Herrnstein and MurrayHerrnstein and Murray (1994). Standing in stark contrast, the Fast Track intervention challenges the notion of inevitable destiny, and the findings obtained through the randomized controlled trials show definitively that young aggressive children can be guided toward a trajectory of productive nonviolent adult lives.

6.2.2 The key psychological process that drives children to aggressive behavior is a defensive mindset, and the key to successful intervention is to address this mindset

If the large body of research summarized in this Element were not sufficient to drive home the idea that the key to understanding children’s aggressive behavior is to understand defensive mindset, then the Fast Track intervention findings should close the case. A primary target of change in Fast Track was to help aggressive children view the world in a different way and from a different perspective than their defensive mindset. Even though these children may have come to be defensive as an understandable and adaptive response to a cruel world, the key to changing their life course would be to develop social-cognitive skills and reject perspectives that are captured in the notion of a defensive mindset. The evaluation of Fast Track shows this intervention can work: Aggressive children can learn these valuable skills and a new mindset. Furthermore, changing defensive mindset to a more open mindset is a key ingredient to the success of Fast Track.

6.2.3 Fast Track is not (yet) a viable model for at-scale implementation.

Fast Track achieved enduring positive change for aggressive children, and a cursory benefit–cost analysis shows the investment in Fast Track would likely yield a very favorable return on that investment. The Fast Track intervention cost about $58,000 per child over a ten-year period. According to Reference CohenCohen (2005), the average life-persistent aggressive child grows up to cost society over $2,000,000 in judicial costs, incarceration, and cost to victims. Simple arithmetic shows a decrease in the prevalence of 2-million-dollar predators by just 3 percentage points would bring a positive rate of return on an investment of $58,000. Assume intervention with 100 children at $58,000 each, totaling $5.8 million. If just 3 of the 100 children were “saved” from otherwise becoming a 2-million-dollar predator, the savings would be $6 million, and the intervention would produce a positive return on investment. Of course, a more sophisticated analysis would bring in discounting due to inflation and other factors. The findings from the randomized trial show a much higher success rate from Fast Track: The rate of adult externalizing psychopathology was reduced by 10 percentage points, which is over 3 times the rate needed to break even. This is a 3-to-1 return on investment that would hold up to rigorous scrutiny.

So why is Fast Track not currently a viable model for scaling up to achieve population impact? Even with this success, the Fast Track intervention as implemented should not be scaled up in any large community for four reasons. First, the “success” achieved still left too many children growing up to have recurring problems. Although Fast Track reduced the rate of adult psychopathology in this high-risk group from 69% to 59%, that outcome still leaves too many children facing unsuccessful lives. An intervention would need to almost “wipe out” the problem (rather than modestly reduce it) for society to “buy” this approach. Second, the cost is too high. Even if the return is great, communities cannot, or will not, allocate such a large amount to “help” disruptive, aggressive children. The political will is not there, and the discretionary budget may not be there. Third, the intervention required a great deal of cooperation from teachers and school principals that is not likely to be forthcoming if the intervention is imposed on them. This problem became evident when children in Fast Track moved to new schools in remote districts. Principals at those schools could not be convinced to invest in school-wide curriculum changes for just one child who had been assigned to their school. Fourth, and perhaps most importantly, we learned that age 5 is relatively late to begin intervention for a process that starts earlier in life. We know the devastating impact of early adverse events. Intervention should start sooner.

6.2.4 To achieve greater success, the intervention needs to start early and reach universally.

The problems of chronically aggressive children are rooted in early threat and adversity, whether it is neglect due to poverty or trauma due to physical abuse or related exposures in the first five years of life. Threat leads to a defensive mindset. Once a pattern of defensive processing begins, it is self-reinforcing and becomes difficult to change. The problem needs to be attacked earlier in the life cycle, before abuse occurs and before neural pathways are established that cannot easily be “undone.” Because early adverse events can occur in any sector of the population, the entire population of children and their families need to be reached in early life.