Abstract
Low autonomic nervous system activity is claimed to be a biomarker for aggressive and antisocial behavior. Although there is evidence that low skin conductance activity (SCA) accounts for variation in the severity of antisocial behavior and predicts the onset of aggression in children and adults, it is unknown whether SCA measured in infancy can predict the development of aggression. We measured SCA in 70 typically developing 1-year-old infants at baseline, during an orienting habituation paradigm, and during a fear challenge. We also observed the infants’ fear behavior, and each mother rated her infant’s temperament and her attachment to her child. At follow-up, mothers rated the children at 3 years old for aggressive and nonaggressive behavior problems. Low infant SCA predicted aggressive behavior, but there was no association between SCA and nonaggressive behavior problems. Mothers’ ratings of the infants’ temperament and their maternal attachment and the infants’ observed fearlessness did not predict later aggression. These results suggest that SCA is a specific biomarker for aggression in low-risk samples of infants.
Early aggression poses challenges to families, educators, and health professionals, and timely prevention efforts are therefore important. Elevated aggression at age 2 years is a risk factor for the persistence of these problems through childhood (Côté, Vaillancourt, LeBlanc, Nagin, & Tremblay, 2006; NICHD Early Child Care Research Network, 2004). Despite decades of research, little is known about the early antecedents of aggression (Hay et al., 2012). Children with persistent antisocial behavior have neurobiological problems (Moffitt & Caspi, 2001; van Goozen, Fairchild, Snoek, & Harold, 2007), yet most studies investigating early aggression have focused on psychosocial influences, and only a few have examined neurobiological influences (Gao, Raine, Venables, Dawson, & Mednick, 2010a, 2010b; Raine, Venables, & Mednick, 1997). The research reported here examined whether neurobiological factors measured in infancy predict later aggression.
Low autonomic nervous system (ANS) arousal is a well-replicated marker for aggression (Raine, 2002). There are three prevalent theories to explain the relationship between reduced ANS arousal and the development of aggression. One theory (Zuckerman, 1979) focuses on stimulation seeking and posits that low physiological arousal motivates individuals to seek out stimulation to raise arousal to an optimal level. Aggressive acts increase arousal. The fearlessness theory (Raine, 2002) proposes that fearless children are more likely to engage in aggression to obtain reward and social status because they are relatively insensitive to the negative consequences of their actions. Fear conditioning is the mechanism by which children learn to link antisocial acts with negative consequences, such as punishment. The fearlessness account, which focuses on the consequences of aggression, and the stimulation-seeking explanation, which focuses on the antecedents to getting into risky situations, are not incompatible.
A third explanation incorporates the previous two: Low ANS arousal in young children may be a risk factor for externalizing behavior problems because individuals with low ANS arousal have difficulty attending and reacting to environmental stimulation (Wilson & Gottman, 1996). Individual differences in prefrontal cortex functioning may be involved in this dynamic: The prefrontal cortex is involved in the allocation of attention, emotion regulation, and stress reactivity (Damasio, Tranel, & Damasio, 1990). Damage to this region leads to psychophysiological abnormalities that predispose individuals to antisocial behavior (Wilson & Gottman, 1996).
A range of measures indexing low ANS arousal have been linked to poor developmental outcomes. Raine et al. (1997) found that low resting heart rate in typically developing children predicted aggression at age 11 years. A large literature has focused on skin conductance activity (SCA) and aggression. Low SCA has been found in children with conduct disorder (Herpertz et al., 2005; van Goozen, Matthys, Cohen-Kettenis, Buitelaar, & van Engeland, 2000), and it predicts persistence of conduct disorder from childhood into adolescence (van Bokhoven, Matthys, van Goozen, & van Engeland, 2005).
Although SCA can explain variation in aggression, to conclude that SCA is a biomarker for aggression it needs to be shown that low SCA precedes the onset of aggression (Sterzer, 2010). Recently, Gao and her colleagues (2010a, 2010b) found associations between low SCA conditioning in typically developing 3-year-olds, aggression at age 8 years, and criminal behavior 20 years later: Low SCA at age 3 was found to be a biological marker for aggression and criminality in later life. What is unclear is whether SCA measured even earlier can predict which children will become aggressive. This is what we sought to determine in our study.
Researchers have used different ways of assessing SCA as a marker for aggression. Although most studies of children have measured resting SCA (e.g., Kruesi et al., 1992), others have measured SCA during emotionally challenging paradigms (e.g., van Goozen et al., 2000), and a few have measured SCA during fear conditioning (Gao et al., 2010a, 2010b). Very few studies have measured SCA in infants (Hernes et al., 2002; Storm, 2000) or toddlers (Posthumus, Böcker, Raaijmakers, van Engeland, & Matthys, 2009), and none have tested the relationship of SCA with aggression before the age of 3 years. Researchers therefore know surprisingly little about infants’ biological risk factors for aggression. The aim of this study was to assess SCA in a low-risk sample of infants using a multimeasure approach to establish whether SCA predicts aggressive behavior at age 3 years.
Method
Participants
One hundred infants took part in this study near their first birthday (mean age = 10.01 months, SD = 1.76, range = 7–14 months). Mothers of 70 of these infants (36 boys, 34 girls) filled out the Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2000) 2 years later, when the children were 3 years old (mothers’ mean age = 35.66 years, SD = 3.95, range = 22–43 years). This retained sample did not differ from the participants who dropped out with respect to demographic characteristics or any dependent measures. Participants were recruited from day-care nurseries. By the mothers’ reports, 63 participants were Caucasian, 1 was mixed-race Caribbean, and 6 were South Asian; 97% described themselves as British. Ninety percent of the mothers had completed a university education, 3% had some postsecondary education but no university degree, and 7% had completed secondary education only. Because our sample was relatively highly educated, we qualified it as a low-risk sample. Ninety-four percent of the mothers described themselves as married or living with a partner; 6% were single or separated. The participating child was the first or only child for 51% of the mothers. The study was approved by Cardiff University’s School of Psychology research ethics committee.
Procedure
Each child’s SCA was measured at baseline, during an orienting habituation paradigm (OHP), and during a fear challenge. Mothers accompanied their child to the laboratory. Once the child had settled, SCA electrodes were placed on his or her foot, and baseline SCA was recorded for 3.5 min. The OHP (Hernes et al., 2002) took 3 min, and the fear challenge (Goldsmith & Rothbart, 1999) took 3.5 min.
Infant measures
Equipment for measuring SCA
SCA recording and analysis programs were custom-made using PsychLab software (Contact Precision Instruments, Cambridge, MA). Two Ag/AgCl electrodes (diameter = 8 mm) with an applied voltage of 50-mV root-mean-square were attached on the left foot: one on the medial side over the abductor hallucis muscle, and the other midway between the phalanges and midpoint of the heel. Electrodes were secured using adhesive collars. Abrasive electrolyte gel (made of V14: Lectron II) improved conductivity. SCA was sampled with a frequency of 50 Hz with a 16-bit resolution.
Resting SCA
To obtain baseline SCA, we measured skin conductance level continuously while the mother and child were quietly playing for 3.5 min. Data were measured in seven 30-s epochs. Mean baseline SCA in micro-Siemens was calculated by taking an average of the averages from the seven epochs.
OHP
During the OHP, we presented an auditory stimulus (1-s burst of white noise at 75 dB) 10 times, with a mean randomized intertrial interval of 18 s (range = 13–28 s; Hernes et al., 2002). Skin conductance response was defined as an increase in conductivity occurring within a latency window 1 to 3 s poststimulus. Amplitudes exceeding 0.01 µS indicated that a skin conductance response had been elicited. The dependent variable was the number of times (SCA frequency) the child showed a skin conductance response to the noise across the 10 stimuli (with scores varying between 0 and 10).
Inducing fear with the Laboratory Temperament Assessment Battery (Lab-TAB)
Fear was induced using the unpredictable-mechanical-toy component of the Lab-TAB (Goldsmith & Rothbart, 1999). For this study, a remote-controlled robot was introduced to the room while the mother monitored the child from an observation booth; she could go to the child if she wanted to. The experimenter placed the robot approximately 1.5 m away from the child, who was strapped into a children’s car seat. The experimenter made the robot approach the child, stopping approximately 15 cm away, while making movements with its arms and emitting noise. The robot then walked backward and stopped at the back of the room for about 10 s before moving forward again. This trial was repeated three times, in line with the Lab-TAB protocol. Video records were made so that the infants’ fear behavior could be coded. Both SCA and behavior data were analyzed in seven 30-s epochs.
SCA during fear
During the Lab-TAB procedure, SCA was measured continuously for 3.5 min, and an average was calculated for each of the seven epochs. An overall mean in micro-Siemens was calculated by taking the average of these averages (fear SCA). Because the fear SCA data were skewed, a square-root transformation was applied.
Marker SCA
The three SCA measures (baseline SCA, SCA frequency, and fear SCA) were standardized and combined by taking their averages to create a new variable, marker SCA. The reliability of this composite measure was high (α = .72).
Behavioral coding
Lab-TAB’s guidelines were followed for coding the video recordings of the fear episode (Goldsmith & Rothbart, 1999). Infants’ responses in each epoch were scored for intensity of facial fear (0–3), distress vocalization (0–2), and bodily fear (0–3), and the scores within each category were averaged across the seven epochs. A composite fear score (Lab-TAB fear) was derived by summing these three averages; scores ranged between 0 and 8, with higher scores indicating more fear and lower scores indicating more fearlessness. Four coders scored the episodes independently for 22% of the sample; intraclass correlation coefficients ranged between .70 and .99.
Mother-reported behavior
Maternal Attachment Inventory
The Maternal Attachment Inventory (MAI; Muller, 1994) is a 26-item self-report instrument assessing maternal activities and feelings indicating a mother’s affection toward her child. Items are scored from 1 (almost never) to 4 (almost always), with higher scores indicating stronger attachment (α = .84). The MAI has been shown to have good internal and test-retest reliability, as well as good validity compared with other indicators of maternal attachment (Muller, 1994).
Revised Infant Behavior Questionnaire
The Revised Infant Behavior Questionnaire (IBQ-R; Gartstein & Rothbart, 2003) assesses 14 domains of infant temperament. We focused on 2 domains, fear (Fear subscale) and anger (Distress to Limitations subscale), that are relevant for the development of aggression. Each of these subscales of the IBQ-R consists of 16 items (αs = .87 and .82, respectively). Higher scores indicate greater levels of fear and anger.
Aggression and antisocial behavior at age 3 years
At follow-up, we used the CBCL to assess aggression and antisocial behavior. The CBCL for ages 1.5 to 5 years is widely used, is standardized, and has good psychometric properties. On the CBCL, parents rate whether 99 problem items apply to their child (0 = not true, 1 = somewhat/sometimes true, and 2 = very/often true). The items are organized into syndromes, and the scores for two syndromes (Attention Problems, 5 items, and Aggressive Behavior, 19 items) are combined to create an Externalizing Behavior Problems score. Following procedures detailed by Raine et al. (1997) and Raine, Reynolds, Venables, Mednick, and Farrington (1998) and suggestions by Burt (2012), we divided the 19-item CBCL Aggressive Behavior syndrome scale into aggressive- antisociality and nonaggressive-antisociality subscales. The aggressive-antisociality subscale consists of 8 items (with total scores ranging between 0 and 16) measuring physical and verbal aggression (e.g., gets into fights, screams a lot; α = .73). The nonaggressive-antisociality scale consists of 11 items (with total scores ranging between 0 and 22) measuring oppositional and hard-to-manage behavior (e.g., demands must be met immediately, punishment does not change behavior; α = .74). Higher scores on these subscales indicate higher antisociality. For comparison purposes and following guidelines (Achenbach & Rescorla, 2000), we used the individual standardized t scores on the broadband Externalizing Behavior Problems scale (α = .89) to classify children as clinically externalized (t score ≥ 64) or normal.
Results
Means and standard deviations are shown in Table 1. Correlations between variables are shown in Table 2. Aggressive-antisociality and nonaggressive-antisociality scores were moderately correlated (r = .59, p = .001). Of the 70 children, 9% had CBCL Externalizing Behavior Problems scores in the clinical range at age 3 years. This is slightly higher than the rate in a sample of 300 low-income boys (6%), who represented a group of “early starters” who showed persistent conduct problems beginning in early childhood (Shaw, Gilliom, & Giovannelli, 2000).
Descriptive Statistics
Note: SCA = skin conductance activity; MAI = Maternal Attachment Inventory (Muller, 1994); IBQ-R = Revised Infant Behavior Questionnaire (Gartstein & Rothbart, 2003); Lab-TAB = Laboratory Temperament Assessment Battery (Goldsmith & Rothbart, 1999).
Intercorrelations Between Study Variables
Note: SCA = skin conductance activity; MAI = Maternal Attachment Inventory (Muller, 1994); IBQ-R = Revised Infant Behavior Questionnaire (Gartstein & Rothbart, 2003); Lab-TAB = Laboratory Temperament Assessment Battery (Goldsmith & Rothbart, 1999).
p < .05. **p < .01.
All measures of SCA were positively associated with each other (rs = .24–.78); however, infant behavioral fear and anger (the IBQ-R results) were not significantly associated with SCA. There were no significant sex differences in the magnitude of these associations. Forty-four infants cried during the fear challenge, and 12 mothers entered the room while the challenge was carried out. There were no differences in SCA (baseline, fear, OHP frequency, or marker) between children who cried and who did not cry, nor between those whose mothers entered the room and those whose mothers did not.
Negative associations were found between all measures of SCA and aggressive antisociality: baseline SCA (r = −.34, p = .005), fear SCA (r = −.36, p = .002), and SCA frequency (r = −.21, p < .083). Nonaggressive antisociality was not significantly associated with any measures of SCA.
To examine whether SCA, as a general marker, predicted later aggression, we used the composite score, marker SCA. Marker SCA was inversely correlated with aggressive antisociality (r = −.38, p = .001): Lower SCA in infancy predicted higher levels of aggression 2 years later. A nonsignificant association was found between marker SCA and nonaggressive antisociality (r = −.12, p = .324).
Discussion
Aggression peaks during toddlerhood, and more children will desist from early aggression than will persist to disorder (Tremblay, 2006). However, evidence also shows that aggression is moderately or highly stable in the most aggressive children (Shaw et al., 2000), and that clinically significant disruptive behavior disorder can be diagnosed in preschoolers (Keenan et al., 2011). Although research on contextual factors has made progress toward identifying children who are at risk for later disorder, researchers know much less about the role of organic factors, including psychophysiological ones, during this time of considerable developmental transition.
We examined whether low SCA in infancy is a biomarker for later aggression. In finding that psychophysiological differences in infancy are predictive of later aggression, we showed that the first 3 years of life are a period when important systems involved in the experience and expression of emotions are being established. Not only did low SCA in infancy predict aggressive behavior at age 3, but it specifically predicted physical and verbal aggression, as opposed to a broader spectrum of difficult behavior. By presenting a novel toy in the context of a brief separation, we provided a challenge that infants had to meet on their own. Low autonomic arousal under these conditions appears to be a biomarker for aggression. These findings enhance current understanding of the early precursors of aggression and can inform prevention efforts by highlighting the role of infant psychophysiology. It is noteworthy that the other measures taken in infancy did not predict later aggression. Mothers’ reports of their infants’ temperament, which correlated with observed fear, were not correlated with SCA and did not predict aggression.
This study used a range of SCA measures. SCA during baseline and the fear challenge were correlated (r = .78, p < .01), and equally associated with later aggression (rs = −.34, and −.36, respectively). These findings support both the fearlessness and the sensation-seeking accounts of aggression. As noted earlier, these accounts are compatible with each other. The prefrontal cortex is important in emotion and personality development. Damage to this region in young children can lead to deficits in psychophysiological responding, such as reduced SCA orienting and arousal (Williams et al., 2000), which could adversely affect emotion regulation and prosocial development and thereby cause aggression in later life.
Our sample was relatively middle-class. The findings, therefore, highlight a link between individual psychophysiological differences and later emotion development in a low-risk sample. Psychophysiological factors show stronger relationships to aggression among people from benign social backgrounds than among those from difficult ones (Raine, 2002). It is therefore possible that the link between SCA and later aggression identified in this sample would not be replicated in higher-risk groups. Nevertheless, psychosocial factors interact with psychophysiological risk factors, and antisocial behavior increases exponentially when social and biological risk factors combine (Raine, 2002). Further research should examine whether the observed effects generalize to vulnerable populations. Future research in the same early developmental time frame should also incorporate features of the child’s environment (including parenting behavior and maternal stress) into the study of SCA, early emotion regulation, and risk pathways.
The study reported here provides evidence for a biological marker that may help to identify subgroups with a distinct neurobiological profile early in life. The scope for changing behavior is greatest in the early years because of the greater plasticity of the brain at that time (Sterzer, 2010). Identifying precursors of disorder in the context of typical development can inform the implementation of effective prevention programs and ultimately reduce the psychological and economic costs of antisocial behavior to society. Although caution is needed in conducting this type of research given concerns about the possible consequences of early labeling of young children (Brotman et al., 2009), investigations into risk factors for later psychopathology should not be delayed until middle childhood.
Footnotes
Acknowledgements
We are grateful to the mothers and children who participated in the study.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.
Funding
The research was supported through studentships from the School of Psychology, Cardiff University, and a program grant from the Medical Research Council (Grant GO400086).