Research on the antecedents of early aggressive behavior has burgeoned, due in part to evidence of stability in aggressive-type behaviors over time, beginning as early as age 2–3 years (Crockenberg & Lourie, Reference Crockenberg and Lourie1996; Cummings, Iannotti, & Zahn-Waxler, Reference Cummings, Iannotti and Zahn-Waxler1989; Shaw et al., Reference Shaw, Winslow, Owens, Vondra, Cohn and Bell1998). It has been stimulated further by studies linking infant temperament to early and later aggressive behavior (Belsky, Hsieh, & Crnic, Reference Belsky, Hsieh and Crnic1998; Calkins & Johnson, Reference Calkins and Johnson1998; Moffitt, Caspi, Dickson, Silva, & Stanton, Reference Moffitt, Caspi, Dickson, Silva and Stanton1996), and by evidence that stability in aggressive behavior occurs under conditions of adverse caregiving or other risk conditions, especially for boys (for a review, see Campbell, Shaw, & Gilliom, Reference Campbell, Shaw and Gilliom2000). When circumstances are more auspicious, there is little stability in aggressive behavior. Thus, identification of conditions that moderate effects of infant temperament on later aggression suggests points of intervention into this developmental process during infancy, when behavior may be easier to change.

In this longitudinal study, we investigate the main and interactive effects of infant temperament (negative reactivity to frustrating events) and gender, infant attentional behaviors, and maternal attention-related behaviors on aggressive behavior in the third year of life. One goal is to identify the characteristics and experiences of infants that both contribute to and limit the development of aggressive behavior. Another goal is to test the specificity of infant distress to frustrating events as a predictor of aggressive behavior, by comparing the results to those obtained when infant distress to novel events is used to predict later aggressive behavior.

Negative Reactivity and Regulation: Dual Facets of Infant Temperament

Infant temperament includes both reactive and regulatory components (Cole, Martin, & Dennis, Reference Cole, Martin and Dennis2004; Rothbart & Derryberry, Reference Rothbart, Derryberry, Lamb and Brown1981), and therefore reactivity and regulation likely contribute jointly to behavioral development and adjustment (Calkins & Fox, Reference Calkins, Fox, Bates and Wachs1994; Rothbart & Bates, Reference Rothbart, Bates, Damon and Eisenberg1998). Further, Rothbart and Derryberry (Reference Rothbart, Derryberry, Lamb and Brown1981) identified two types of infant negative reactivity, distress to limits and distress to novelty, that neither correlate highly with each other, nor similarly with child behaviors (Karrass & Braungart-Rieker, Reference Karrass and Braungart-Rieker2004; Mangelsdorf, McHale, Diener, Goldstein, & Lenn, Reference Mangelsdorf, McHale, Diener, Goldstein and Lehn2000). These two types of reactivity are also linked to specific negative emotions (i.e., fear and anger), which are associated, in turn, with different sides of the brain and the synchrony between them (Buss et al., Reference Buss, Schumacher, Dolski, Kalin, Goldsmith and Davidson2003; Davidson & Rickman, Reference Davidson, Rickman, Schmidt and Schulkin1999). This connection provides a basis for expecting discriminant validity of the two types of infant negative reactivity. Nevertheless, a strong case for identifying distress to limits as the type of negative reactivity linked specifically to later aggression requires empirical evidence that distress to novelty shows no comparable effect. We test the specificity of prediction in this study.

Infant regulatory behaviors that begin to develop during the first year of life (Kopp, Reference Kopp1989, Reference Kopp2002; Rothbart, Ziaie, & O'Boyle, Reference Rothbart, Ziaie and O'Boyle1992) may explain why some reactive infants develop in ways that are consistent with their early reactivity and others do not. Central among these behaviors is orienting toward visual locations (the posterior attention system; Posner & Peterson, Reference Posner and Peterson1990) that demonstrates important changes between 3 and 6 months postpartum. These include the increasing ability to disengage gaze from an external stimulus (Johnson, Posner, & Rothbart, Reference Johnson, Posner and Rothbart1991), which correlates with less negative affect in 4-month-old infants, indicating that regulation of attention and expression of negative emotions are linked by this age. Recent studies confirm that by 4–6 months of age infants who regulate attention better in one context show less negative affect in another (Axia, Bonichini, & Benini, Reference Axia, Bonichini and Benini1999; Whitehead & Frick, Reference Whitehead and Frick2004).

More compelling evidence that infants modulate negative emotion in frustrating contexts by regulating attention comes from studies in which contingency analyses were used to determine if putative infant regulation behaviors served a regulatory function. Stifter and Braungart (Reference Stifter and Braungart1995) found that self-comforting and orienting away from the frustrating event toward another object or toward mother (i.e., attention shifting) in 5- and 10-month-old infants were more likely to occur when infant distress was decreasing than when it was increasing, suggesting a regulating effect. Buss and Goldsmith (Reference Buss and Goldsmith1998) reported similarly that decreases in anger distress during a barrier task were more frequent than expected following distraction among 6-, 12-, and 18-month-old infants. Distraction was defined as attending to something other than the provoking stimulus. We infer from these findings that infants who shift attention away from frustrating events and toward something else have developed an adaptive way of managing frustration, and therefore should engage in less aggressive behavior as they get older. It seems likely as well that prediction from early negative reactivity to later aggressive behavior should be weaker for negatively reactive infants who develop effective regulatory behaviors than for those who do not (Calkins, Dedmon, Gill, Lomax, & Johnson, Reference Calkins, Dedmon, Gill, Lomax and Johnson2002).

Counter to this line of reasoning is evidence that the ability to inhibit a dominant response in favor of a nondominant response (effortful control) that begins to develop later in the first year (Rothbart, Reference Rothbart, Kohnstamm, Bates and Rothbart1989; Rothbart, Ahadi, & Hershey, Reference Rothbart, Ahadi and Hershey1994). If we assume that the dominant response to frustration is approach, looking away from a frustrating event and toward something else would be a subdominant response, and doing so would require a capacity that may be unavailable to younger infants. Conversely, antecedents of a later behavior can often be detected before the more mature form has fully developed, as demonstrated by evidence that speed of information processing at 6 months predicts more complex forms of cognition 5 years later (Rose & Feldman, Reference Rose and Feldman1997). Thus, the ability to look away from a frustrating event at 6 months could be an early manifestation of effortful control, and demonstrate a similar modulating effect on the development of aggressive behavior. We test this possibility in this study.

Gender and Emotion Regulation

There is evidence that female infants develop the ability to self-regulate more rapidly than male infants. Although males displayed more irritability and crying than females during the neonatal period (Feldman, Brody, & Miller, Reference Feldman, Brody and Miller1980; Korner, Reference Korner1969; Osofsky & O'Connell, Reference Osofsky and O'Connell1977), female newborns showed greater social orientation and maintained eye contact longer than males, possible antecedents of better emotion regulation. Consistent with a gender difference in emotion regulation, Belsky, Fish, and Isabella (Reference Belsky, Fish and Isabella1991) found that infant boys were less likely than girls to exhibit reductions in negative emotionality between 3 and 9 months of age. More specifically, Weinberg, Tronick, Cohn, and Olson (Reference Weinberg, Tronick, Cohn and Olson1999) reported that male infants had greater difficulty than female infants in maintaining emotion regulation during the still-face procedure. Females looked away from mothers and toward other objects more than males, and increased this behavior during the still-face and reunion episodes during which a greater need for regulation is presumed. Stifter and Spinrad (Reference Stifter and Spinrad2002) found similarly that among infants who cried excessively at 6 weeks, only boys displayed lower regulation in a frustrating context at 5 and 10 months. Moreover, looking away from the frustrating event and toward something else was the most frequently used regulating behavior, suggesting that the gender difference in regulation may have reflected a difference in attention shifting. Calkins et al. (Reference Calkins, Dedmon, Gill, Lomax and Johnson2002) reported also that 6-month-old male infants were less able to regulate physiologically than females of the same age.

It appears from these studies that male infants may be deficient in certain self-regulating behaviors during the first year, in particular, regulation involving control of attention. As a consequence, early assessments of attention shifting may be linked more weakly to later aggression for male than female infants, and male infants may rely more on their mothers to regulate negative emotion than do female infants in the middle of the first year of life.

Negative Reactivity, Infant Regulation, and the Development of Aggression

There have been few tests of the association between emotion regulation in infancy and later aggressive behavior, although in studies with older children this association is fairly well established, both concurrently and predictively, especially for children high in negative emotionality (e.g., Eisenberg et al., Reference Eisenberg, Fabes, Guthrie, Murphy, Maszk and Holmgren1996). Kochanska, Murray, and Harlan (Reference Kochanska, Murray and Harlan2000) extended the main effect of emotion regulation downward in age, reporting that effortful control at 22 months predicted more regulated anger at 33 months of age. This is noteworthy because effortful control is thought to include an attentional component, and Kochanska et al. (Reference Kochanska, Murray and Harlan2000) reported further that focused attention at 9 months predicted later effortful control, confirming this link. It is uncertain, however, how their measure of focused attention relates to the infant's ability to shift attention in the service of emotion regulation, the focus of our investigation.

To date, moderating effects of emotion regulation on high infant negative reactivity in relation to later aggression-related behavior have been tested in two studies. Stifter, Spinrad, and Braungart-Reiker (Reference Stifter, Spinrad and Braungart-Rieker1999) reported that 5-month-old infants who were highly reactive and highly regulated during a frustration task were less defiant as toddlers than highly reactive infants who lacked regulation. Similarly, Belsky, Friedman, and Hsieh (Reference Belsky, Friedman and Hsieh2001) reported that high negative emotionality at 15 months was associated with lower social competence at 36 months only when attentional persistence was poor. Taken together, these findings indicate that reactive and regulatory components of infant temperament interact to predict differences in later aggression-related behaviors. However, in Stifter et al. (Reference Stifter, Spinrad and Braungart-Rieker1999), regulation behaviors were combined, leaving unanswered whether differences in attention shifting explain the moderating effect. In Belsky et al. (Reference Belsky, Friedman and Hsieh2001), attentional persistence included both attention focusing and attention shifting, obscuring the specific process by which attention may have regulated infant negative reactivity. We extend these findings by testing the moderating effect of infant attention shifting and attention focusing on the trajectory between infant reactivity to frustrating events and later aggressive behavior.

Negative Reactivity, Maternal Behavior, and the Development of Aggression

Thomas and Chess's (Reference Thomas and Chess1977) thesis that parents' ability to match their behavior with their infants' temperament influences the development of negatively reactive infants has been elaborated further (Lerner, Nitz, Talwar, & Lerner, Reference Lerner, Nitz, Talwar, Lerner, Kohnstamm, Bates and Rothbart1989; Wachs & Kohnstamm, Reference Wachs, Kohnstamm, Wachs and Kohnstamm2001) and linked with the development of aggressive behavior (Campbell et al., Reference Campbell, Shaw and Gilliom2000). Caregiver behavior has also been implicated in the process by which infants learn to regulate negative arousal, and is a likely moderator of infant negative reactivity for that reason as well.

Maternal sensitivity/responsiveness has been shown to modulate infant negative affect concurrently (Haley & Stansbury, Reference Haley and Stansbury2003), as have specific maternal behaviors (Crockenberg & Leerkes, Reference Crockenberg and Leerkes2004; Jahromi, Putnam, & Stifter, Reference Jahromi, Putnam and Stifter2004). Thus, it is reasonable to expect mothers to alter associations between early negative reactivity and later aggression by fostering emotion regulation. Support for such a moderating effect of maternal behavior is accumulating. As reported by van den Boom (Reference van den Boom1994, Reference van den Boom1995), mothers of irritable infants who participated in an intervention designed to foster responsiveness to infant cues were more sensitive when their infants cried, and at age 4 their children exhibited fewer externalizing behaviors. Support for a beneficial effect of maternal behavior on infant negative reactivity derives from Feldman, Greenbaum, and Yirmiya's (Reference Feldman, Greenbaum and Yirimiya1999) finding that mutual synchrony between mothers and infants at 9 months predicted self-control at age 2 among those with “difficult temperaments” as infants.

Mothers also engage in behaviors that increase the likelihood of later aggression, especially in negatively reactive infants. Calkins and Johnson (Reference Calkins and Johnson1998) reported that the correlation between distress to limits and aggression at 18 months was greater when mothers acted in ways that precluded children from doing an activity themselves. Similarly, Belsky et al. (Reference Belsky, Hsieh and Crnic1998) found that the extent to which parents imposed their own agendas on a child was more strongly linked with externalizing if the child had been a negatively reactive infant. Rubin, Hastings, Chen, Stewart, and McNichol (Reference Rubin, Hastings, Chen, Stewart and McNichol1998) reported similar results for 2-year-old boys: maternal negative dominance (intrusive, overcontrolling behavior) was associated with more aggression and externalizing behavior for boys with less regulated temperaments. For girls, aggression was linked only with a lack of regulatory skills.

It appears from these studies that mothers exacerbate early negative reactivity, increasing the likelihood of later aggressive behavior, by depriving infants of opportunities to self-regulate. In contrast, sensitivity to infant cues modulates negative reactivity, reducing the risk of later aggression. We propose that it does so in part by supporting infant efforts to shift attention away from the source of frustration. For young infants whose mobility is limited and whose ability to shift attention is just developing, parents produce diverting events that attract their infants' attention. They use animated facial and vocal cues to engage their infants, introduce visual stimuli that interest them, and maintain infant attention through these behaviors. From this perspective, parents who engage in such behaviors when infants are exposed to frustrating events foster the redirection of infant attention in the service of emotion regulation. Parents who encourage infant attention to the events should have the opposite effect because infants who focus on the source of frustration should be more distressed, especially when the source cannot be overcome through direct action. Moreover, if males develop the ability to regulate attention more slowly than females, the impact of maternal behavior should be greater for them than for females because they must depend more on their mothers to regulate negative affect.

The Current Study

We test the main and interactive effects of infant distress to frustrating events and attention shifting and focusing at 6 months on aggressive behavior 2 years later, using maternal reports of infant reactivity and observations of infant attention. Second, we attempt to replicate previous gender differences in infant reactivity and regulation, and test gender-linked patterns of prediction from infancy to later aggression. Third, we test associations between infant attention behaviors and maternal encouragement of those behaviors to determine if they are consistent with the expected pattern of influence. Fourth, we examine moderating effects of maternal behaviors expected to increase or reduce the risk of aggressive behavior among easily frustrated infants. Fifth, we test the discriminant validity of distress to limits relative to distress to novelty in relation to later aggressive behavior. The hypotheses are listed in the order in which they are tested.

Method

Measures

Infant Behavior Questionnaire (IBQ)

Two IBQ (Rothbart, Reference Rothbart1981) subscales were administered to assess mothers' perceptions of their infant's temperament, distress to limits (distress to frustrating events), and distress and latency to approach sudden or novel stimuli (distress to novelty). Mothers indicate on a 7-point scale how frequently their infants respond to specific events by fussing, crying, or no reaction during the previous week (e.g., when placed in a car seat—limits, or when exposed to a loud noise—novelty). At 6 months, subscales have good internal reliability (.75–.81), moderate interrater reliability (.54–.66), and concurrent validity with home observations of infant temperament (mean r = .40), and the negative emotionality and approach–sociability subscales of the Revised Infant Temperament Questionnaire and the Infant Characteristics Questionnaire (rs = .61 to .73; Rothbart, Reference Rothbart1981; Rothbart & Goldsmith, Reference Rothbart and Goldsmith1985). Mean ratings on the 20-item Distress to Limits Scale (α = .78) and on the 17-item Distress to Novelty Scale (α = .68) served as emotion-specific measures of infant reactivity; the latter was used to test the discriminant validity of the distress to limits scale as a predictor of later aggressive behavior. Descriptive data are included in Table 1.

Table 1.

Descriptive statistics

Note: IBQ, Infant Behavior Questionnaire.

6-Month behavioral observation

Following a 5-min warm-up, mothers placed their infants in a car seat (attached firmly to a chair), and then sat 3 feet away, situated so that by turning their heads to the right infants could see them. Two novel events were introduced first, followed by a 5-min break, and then two frustrating events: arm restraint and toy removal. The order of frustrating events was counterbalanced to control for event-specific effects on infant and maternal behavior. Only data obtained from the frustrating events are included in this report.

Prior to the toy retraction, the experimenter moved the chair holding the infant seat into a table and then sat to the infant's left. She showed the infant two attractive toys (a teething ring with multicolored plastic keys and a brightly colored plastic face with a mirror on the back). When the infant displayed interest in one toy, the experimenter placed that toy on the table directly in front of the infant for 15 s and placed the unselected toy out of view. After 15 s, she placed the toy just beyond the infant's reach. This sequence was repeated 12 times. The experimenter did not interact with the infant during this task. During arm restraint, the infant seat was pushed away from the table. The experimenter knelt in front of the seat and gently held the infant's forearms immobile for 3 min. Her head was bowed so that she did not interact with the infant.

During the first frustrating event (mother uninvolved), mothers remained neutral so that we could observe infants' self-regulatory behaviors. During the second event (mother involved), mothers interacted with their infants as they liked, but were asked not to intervene directly (e.g., give the removed toy to the infant) or to remove their infant from the seat unless they wished to end the activity.Footnote ² Measures of infant behavior were obtained when mothers were not involved; measures of maternal behavior were derived from the mother involved condition.

Infant and mother behavior coding

Infant and maternal behaviors were coded continuously from videotapes, using a computerized, event-based coding system. Trained students coded in pairs to maintain accuracy while viewing tapes, operating the VCR, and entering codes. Different pairs coded infant and maternal behaviors to avoid bias, and were blind to all other data and to the specific hypotheses of the study. Pairings varied to prevent pair-linked coder drift. The authors coded 25 videotapes independently, at the beginning and midway through the process, to assess reliability and to prevent coder drift for each type of coding.

Fourteen mutually exclusive behavioral codes, 12 of which were adapted from Rothbart et al. (Reference Rothbart, Ziaie and O'Boyle1992), were used to code infant behavior. Two infant behaviors, resists and looks at experimenter, were added based on our observations of infant behavior during the frustrating events. Brief definitions are included in Table 2. Coding instructions are available from the first author. Thirteen other codes were created during coder training to identify instances in which infants engaged in two or more behaviors simultaneously (e.g., self-soothe and look at mother). Intercoder reliability for all codes within a 1-s interval ranged from .69 to .97 (mean κ = .78).

Table 2.

Definitions of infant and maternal behaviors

Note: The infant and maternal behaviors used in this table are in bold.

To maintain an adequate subject to variable ratio, infant behaviors were combined based on both conceptualizations of the behaviors and their simple correlations. To control for time differences, each variable was defined as the percent time the infant engaged in the behavior. Based on the conceptualization of attention shifting presented above and prior research identifying attention shifting as an effective regulator in infancy and a likely moderator of negative reactivity, two infant behaviors were identified a priori as predictors of aggressive behavior: attend to the frustrating event; and look away from the frustrating event. Attend to was the sum of inspects the frustrating event (κ = .75) and looks at experimenter (κ = .77), combined because both involved attending to the frustrating event and correlated significantly, r (84) = .50, p < .01. Look away was the sum of visual regard of another object (κ = .85) and look at mom (κ = .79), combined because both involved looking away from the frustrating event and toward something else, and correlated significantly, r (84) = .57, p < .01. These variables were positively skewed, and therefore underwent a logarithmic transformation to reduce skewness (Tabachnick & Fidell, Reference Tabachnick and Fidell1996).

Twelve maternal behavior codes were created based on existing schemes (Farran, Kasari, Comfort, & Jay, Reference Farran, Kasari, Comfort and Jay1986; van den Boom, Reference van den Boom1994.) Intercoder reliability (κ) within a 1-s interval ranged from .65 to .85 for the 12 codes (mean κ = .75) using procedures described above. These are defined in Table 2. Coding instructions are available from the first author. Variables representing the percent of time mothers engaged in each of the 12 behaviors were created. Because of technical difficulties, data were missing on one infant and on two mothers; mean scores were substituted in those cases.

Based on the argument that mothers foster attention shifting by inviting their infants to look away from frustrating events and engaging positively with the infants when they do so, and discourage it when they encourage their infants to attend to the event, twoFootnote ³ maternal behaviors were included as potential moderators of infant reactivity to frustrating events: mother encourages infant attention away from the frustrating event (nontask engagement, κ = .80) and mother encourages infant attention to the frustrating event (task engagement, κ = .73). The latter variable was skewed, and therefore corrected using a logarithmic transformation. The untransformed descriptive data for infant and maternal behaviors are included in Table 1.

Child Behavior Checklist (CBCL)/2-3

This 100-item checklist (Achenbach, Reference Achenbach1992) was administered to mothers to assess children's behavioral and emotional symptoms. Mothers indicate whether specific behaviors are not true, somewhat/sometimes true, or very/often true for their child within the last 2 months. The T score for the 15-item aggression subscale (e.g., hits others, temper, disobedient, defiant) was the operational measure of aggressive behavior used in this study. The distribution of T score aggression was skewed, and thus corrected using a log transformation; an outlier was adjusted using guidelines from Tabachnick and Fidell (Reference Tabachnick and Fidell1996). Only one child had a score in the clinical range of the aggression subscale. Descriptive data using the untransformed T scores are included in Table 1.

The scale has excellent test–retest reliability for maternal ratings of nonreferred 2- and 3-year-olds 1 week apart, r (60) = .86, p < .001, stability over a 1-year interval, r (74) = .62, p < .001, and good interparent agreement for 2 year olds, r (63) = .71, p < .001 (Achenbach, Reference Achenbach1992); it also distinguishes clinically referred and nonreferred children (p < .001), controlling for demographic differences.

State-Trait Anger Expression Inventory (STAXI)

Mothers completed the STAXI (Spielberger, Reference Spielberger1988) at 2.5 years to provide a brief, objectively scored measure of their tendency to become angry. The scale consists of 44 items, administered in three subsections, and distributed across five main scales. Of these, only the 10-item Trait Anger scale (Cronbach α = .73) was included in this study as a possible covariate of maternal reports of infant frustration and aggressive behavior. We reasoned that mothers who were prone to anger might view their infants as easily frustrated and their children as aggressive, or behave in ways that would increase aggressive behavior. Items on this scale are rated on a 4-point scale in relation to “how I generally feel,” with 1 = almost never, and 4 = almost always. Examples are “I have a fiery temper,” “I fly off the handle,” “It makes me furious when I am criticized in front of others.” Higher scores indicate greater and more persistent trait anger. Descriptive data are provided in Table 1.

The test–retest reliability for the Trait Anger scale of the STAXI over a 2-week interval was .77 for adult females (Jacobs, Latham, & Brown, Reference Jacobs, Latham and Brown1988). The scale exhibits convergent validity with the Buss–Durkee Hostility Inventory and the Hostility and Overt Hostility Scales of the Minnesota Multiphasic Personality Inventory (Spielberger & Sydeman, Reference Spielberger, Sydeman and Maruish1999).

Conclusions and Implications

In this study, we demonstrate that an infant's tendency to look at a frustrating event is associated with a stronger tendency to engage in aggressive behavior 2 years later, and further that the infant's shifting of attention away from a frustrating event and toward something else is adaptive over time, as demonstrated by its association with less aggressive behavior, albeit only for girls. In addition, the exacerbating effect of maternal behavior provides meaningful information about the conditions under which we can expect both continuity and lawful discontinuity between early temperamental reactivity to frustrating events and later aggressive behavior. Taken together, these findings are congruent with a contextual approach to temperament that emphasizes the fit between infant characteristics and the social environment in predicting developmental outcomes (Lerner et al., Reference Lerner, Nitz, Talwar, Lerner, Kohnstamm, Bates and Rothbart1989; Thomas & Chess, Reference Thomas and Chess1977; Wachs & Kohnstamm, Reference Wachs, Kohnstamm, Wachs and Kohnstamm2001). They suggest that altering specific caregiver behaviors toward easily frustrated infants may reduce aggressive behavior later in development, but also that attempts to change other maternal behaviors (e.g., encouragement of looking away) may have a greater impact on infant emotion regulation and later aggression if it occurs later in the second year of life. If these findings are replicated, they will identify strategies useful to clinicians and other practitioners engaged in preventive and therapeutic interventions with infant–mother dyads.