Introduction
Executive functioning (EF) refers to a set of higher order cognitive processes, which are involved in the self-regulation of thought, action and emotion (Séguin and Zelazo
2005). EF processes are necessary for adaptive and goal-oriented behavior, and have been associated with the integrity of neural systems in the prefrontal cortex (Fahie and Symons
2003). A number of different neuropsychological concepts are encompassed within EF, including inhibition, attentional control, working memory, cognitive flexibility or set shifting, goal setting and problem solving (Senn et al.
2004). Impairments in executive functioning (EF) have been linked to an impulsive behavioral style and the regulation of aggressive behavior (Séguin et al.
1995). Despite the fact that the origins of persistent aggressive behavior are to be found in early childhood (Shaw et al.
2005), only a few studies into EF have been conducted in preschoolers with aggressive behavior or conduct problems (e.g., Hughes et al.
2000; Senn et al.
2004). In this introduction, the relation between aggressive behavior and EF in elementary school children and adolescents is discussed first. Second, issues regarding the assessment of EF in young children are addressed. Third, several studies of EF in preschoolers who show aggressive behavior are reviewed and, finally, the aim of the study is presented.
Aggressive Behavior and EF
Aggressive behavior and conduct problems are often defined in different ways, for example in terms of psychiatric disorders, such as disruptive behavior disorders (DBD). The relation between EF and DBD has been studied (e.g., Morgan and Lilienfeld
2000), but to a lesser extent than the relation between EF and attention deficit hyperactivity disorder (ADHD; Willcutt et al.
2005). Consequently, the association between ADHD and impairments in EF has been clearly established in research (Willcutt et al.
2005). In their review, Pennington and Ozonoff (
1996) concluded that impairments in EF found in children with DBD were due to the presence of comorbid ADHD. Since then, this finding has been confirmed in several other studies (e.g., Clark et al.
2000; Nigg et al.
1998; Oosterlaan et al.
2005). However, in several other studies, the relation between aggression and EF impairments was maintained while controlling for ADHD (Déry et al.
1999; Séguin et al.
1999,
2004). Likewise, in a meta-analysis, Oosterlaan et al. (
1998) concluded that deficits in response inhibition in children aged 6 to 12 years were not uniquely associated with ADHD, but also with conduct disorder (CD). In addition, children and adolescents with DBD have been found to show a dysfunction on EF tasks in which motivational processes are involved (Blair et al.
2001; Matthys et al.
1998,
2004; Van Goozen et al.
2004). In conclusion, EF impairments are to be expected in children and adolescents who show aggressive behavior disorders, irrespective of ADHD. It is unclear, however, whether this holds true for preschool children.
Assessment of EF in Young Children
Historically, young children have been assumed to lack executive capacities (Isquith et al.
2005). For this reason, studies into the EF of preschool children are relatively scarce and results are contradictory. Moreover, there is a paucity of measures to assess EF in young children (e.g., Carlson
2005; Espy et al.
2001). Only a few of the tests originally designed for older children, adolescents or adults that were adapted for 4-year-olds are completely developmentally appropriate for preschool children (Gerstadt et al.
1994; Isquith et al.
2005; Kerr and Zelazo
2004; Welsh et al.
1991). Recently, the preschool period has gained attention with respect to the development of EF. During early childhood EF matures substantially, e.g., between 3 and 4 years of age improvements appear in response inhibition and the ability to think flexibly (Espy
1997; Jacques and Zelazo
2001).
However, no consensus exists concerning which EFs can already be distinguished in preschool children. Exploratory and confirmatory factor analyses have often been used to delineate components of EF in both adults and children (e.g. Hughes
1998; Miyake et al.
2000). Welsh et al. (
1991) suggested at least three executive factors: working memory and planning, inhibition of maladaptive prepotent responses, and self-monitoring or attentional flexibility. Senn et al. (
2004) identified working memory, inhibition, and flexibility or shifting as latent executive constructs in a sample of preschoolers, which also have been used in previous studies (e.g., Pennington
1997). A recent review on EF in preschool children shows that the EF components found most often were set shifting, working memory and inhibition (Garon et al.
2008). In confirmatory factor analyses, these three EFs were partially independent but still intercorrelated (Letho et al.
2003; Miyake et al.
2000). Therefore, EF can be viewed as a unitary construct with dissociable components in the preschool years (Miyake et al.
2000).
EF in Preschool Children with Aggressive Behavior
Studies which did investigate EF in young children who show aggressive behavior revealed impairments in inhibition of maladaptive prepotent responses (Hughes et al.
1998), deficits in planning and inhibitory control in a group of hard-to-manage preschoolers (Hughes et al.
2000), and impaired performance on a motor-planning and attention task and on a semantic classification and working memory task in a group of children who met the criteria for oppositional defiant disorder (ODD), with and without ADHD (Speltz et al.
1999). In other studies of EF in preschoolers with aggressive behavior, the role of ADHD was taken into account more specifically. Results of these studies showed poor EF in children with DBD and comorbid ADHD, but the association between DBD and EF deficits did not remain significant, when the effect of ADHD was partialled out. For example, Kalff et al. (
2002) showed that children with comorbid DBD and ADHD are more impaired in working memory than children with only DBD. Berlin and Bohlin (
2002) and Sonuga-Barke et al. (
2002) also found a relationship between deficits in inhibition and ADHD. More recently, Thorell and Wåhlstedt (
2006) found that poor EF performance on inhibition, working memory and verbal fluency tasks was associated with symptoms of ADHD, but not with ODD symptoms. In that study, gender differences were also examined but no significant differences between boys and girls were found.
This Study
To sum up, although the association between DBD or aggressive behavior, irrespective of ADHD, and impairments in EF has been established in elementary school children and adolescents, this association is less clear in preschool children. Yet, the question whether EF impairments are related to disruptive behavior in preschool children is important, because of the high risk for antisocial behavior of children who show early onset aggressive behavior (Moffitt
1993). The early detection of EF impairments might shed a light on the role of neuropsychological deficits in the development of behavior disorders. Therefore, the present study aimed to assess EF in a population-based sample of 4-year-old children who show aggressive behavior. We hypothesized that preschool children with aggressive behavior problems display impairments in EF, compared to control children. Specifically, based on EF literature in school-age and preschool children, inhibition and working memory are expected to be most clearly impaired. Due to inconsistency (Klenberg et al.
2001; Overman
2004) and scarcity of studies into EF and gender differences, no hypotheses were formulated regarding possible gender differences.
A battery of tasks adapted for preschoolers was used to assess inhibition, working memory, set shifting and fluency. To identify reliable constructs of EF in this sample of young children a factor analysis was performed. Due to the lack of consensus about the definition of EF, especially concerning young children, and because relatively new EF tasks were used in this study, this factor analysis was needed to explore the relation between the different EF variables in preschoolers. In addition, we investigated whether potential EF deficits were due to aggression or attention problems, by partialling out the influence of attention problems. Based on studies into EF impairments in school-aged children and adolescents with conduct problems or aggressive behavior, we hypothesized that EF deficits are mainly explained by aggressive behavior, irrespective of attention problems.
Results
Data Analysis
Missing data were primarily the result of children who failed to understand task instructions or whose assessment could not be completed due to extreme inattentive or noncompliant behavior. A score of 0 was only given, when the child had attempted the task and then failed. Groups were matched on IQ. The AGGR and CONTR group differed significantly on age (see Table
1); therefore age was used as a covariate. First, a factor analysis was performed on data of all children. Second, ANCOVAs were conducted using a factorscore and other task variables, in which both group and gender were used as between-subject factors. All analyses were performed using SPSS 15.0 (2006).
Factor Analysis
To explore the relation between EF variables, scores from the neuropsychological tasks were submitted to an exploratory factor analysis. Principal axis factoring (PAF) was performed on ten variables from different neuropsychological tasks. For this factor analysis, one variable of each task was chosen to prevent an artificial clustering of variables from the same task. To investigate impairments in EF, the number of incorrect responses of the child was used as main outcome variable. However, not every EF task used was designed to measure the number of incorrect responses, especially not the manually administered tasks. Therefore, we used the number of incorrect responses the children gave on the computerized tasks and for the manually administered tasks, we employed the number of correct responses. Using the number of correct responses of the computerized tasks yielded a comparable factor solution, but factors could less clearly be distinguished. Therefore, we chose to include the number of correct responses on the manually administered task and the number of incorrect responses on the computerized tasks.
A PAF followed by varimax rotation was performed and three factors with eigenvalues greater than 1 were extracted. This solution accounted for 33.13% of the variance. Orthogonal and oblique rotation resulted in the same factor solution. Item loadings of 0.40 or higher are depicted in Table
2. The first factor accounted for 16.67% of the variance, with factor loadings pointing to errors of commission on the inhibition trials in several computerized tasks. Therefore, this factor, measuring impairments in inhibition, was labeled ‘Inhibition’. The internal consistency of this factor was 0.69. The second factor accounted for 8.87% of the variance and consisted of only one item, measuring both working memory and set shifting. This task variable, the number of incorrect responses of the Shape School Both condition, was analyzed separately. The third factor accounted for 7.66% of the variance and consisted of variables from three manually administered tasks. The internal consistency of this factor was 0.39. Factor loadings are depicted in Table
2. Due to the low percentage of explained variance of the factor analysis and the low internal consistency of the third factor, we decided to include only the Inhibition factor in the analysis. All other tasks variables were analyzed separately.
Table 2
Item Loadings of the Rotated Factor Matrix
SS Inh: incorrect inhibition | 0.783 | | |
SS Inh: incorrect | 0.558 | | |
GNG: incorrect inhibition | 0.537 | | |
SS Both: incorrect inhibition | 0.488 | | |
SS Both: incorrect | | 0.701 | |
OCTC correct | | | 0.488 |
DS correct | | | 0.481 |
VF correct | | | 0.413 |
ANCOVAs
Significant correlations were only found between variables included in the Inhibition factor. Therefore, one-way ANCOVAs with age as a covariate were carried out to compare performance of the AGGR group on the Inhibition factor and the other task variables to the CONTR group (see Table
3). To conduct the ANCOVA on the Inhibition factor, a factorscore was constructed by transforming the inhibition variables from different tasks into standardized scores. This
z-score transformation enhances the comparability among the variously scored tasks. Next, the Inhibition factorscore was computed by dividing the total standardized score of the four variables by the number of variables included in the factor. Group and gender were entered in the analysis as between-subject-factors. Means, standard deviations and the results of the ANCOVAs of the Inhibition factor and of the other task variables are displayed in Table
3, for group and gender. Effect sizes were also calculated (see Table
3), using Cohen’s
d; 0.2 indicates a small effect, 0.5 a medium effect and 0.8 a large effect size (Cohen
1992).
Table 3
Mean Scores and Standard Deviations by Group and Gender
Factorscore Inhibition | 0.34 | 0.93 | –0.09 | 0.71 | –0.85 | 0.52 | –0.36 | 0.29 | 9.90 | 0.00 | 0.57 | 10.31 | 0.00 | 0.57 |
SS Inh: incorrect | 1.41 | 2.13 | 0.70 | 1.49 | 0.73 | 1.00 | 0.49 | 1.04 | 4.40 | 0.04 | 0.35 | 3.49 | 0.06 | 0.33 |
SS Inh: incorrect inhibition | 2.25 | 3.43 | 1.39 | 2.86 | 0.95 | 1.67 | 0.46 | 1.25 | 6.50 | 0.01 | 0.48 | 2.92 | 0.09 | 0.31 |
SS Both: incorrect inhibition | 2.47 | 3.31 | 0.70 | 1.02 | 1.75 | 3.27 | 0.37 | 0.88 | 2.10 | 0.15 | 0.25 | 11.95 | 0.00 | 0.66 |
GNG: incorrect inhibition | 3.00 | 3.16 | 2.22 | 3.15 | 1.66 | 2.21 | 0.91 | 0.92 | 7.26 | 0.01 | 0.54 | 3.78 | 0.05 | 0.35 |
VF correct | 6.58 | 2.62 | 8.04 | 3.61 | 7.27 | 3.32 | 8.54 | 3.44 | 0.33 | 0.57 | 0.23 | 7.80 | 0.01 | 0.43 |
DS correct | 4.88 | 1.85 | 5.17 | 2.19 | 5.47 | 2.03 | 5.69 | 2.14 | 0.84 | 0.36 | 0.29 | 0.87 | 0.35 | 0.14 |
DN correct | 10.12 | 4.26 | 11.74 | 3.41 | 11.91 | 4.23 | 12.20 | 3.64 | 2.25 | 0.14 | 0.36 | 2.21 | 0.14 | 0.25 |
OCTC correct | 6.02 | 1.68 | 6.96 | 1.94 | 6.39 | 1.44 | 6.46 | 1.50 | 0.14 | 0.71 | 0.08 | 3.85 | 0.05 | 0.28 |
SS Both: incorrect | 3.69 | 2.65 | 2.83 | 2.73 | 3.80 | 2.97 | 3.77 | 3.18 | 0.89 | 0.35 | 0.12 | 0.87 | 0.35 | 0.12 |
SS Switch: incorrect | 6.76 | 4.58 | 2.39 | 3.19 | 3.80 | 3.65 | 3.49 | 4.04 | 1.53 | 0.22 | 0.44 | 13.18 | 0.00 | 0.53 |
The factorscore for Inhibition was included in the first ANCOVA. The four variables that were included in the factorscore were examined in subsequent analyses. Effects for both group and gender were found to be significant on the Inhibition factor and showed medium effect sizes. For group, other variables that yielded significant effects were only the variables included in the Inhibition factorscore. Age as a covariate did not affect the results on the Inhibition factor. Moreover, inspection of the data pointed out that the difference in inhibition scores was not affected by the distribution of age over the groups. The group effects on the Inhibition factor, and on variables included in this factor, pointed in the expected direction, with the AGGR group demonstrating more incorrect inhibition-trials than the CONTR group. This implies that the AGGR group was significantly more impaired in inhibition than the CONTR group. In addition, the Pearson product-moment correlation between the Aggressive Behavior score of the CBCL and the Inhibition factor was computed (r = 0.287, p = 0.00), which implies that a higher aggressive behavior score was associated with more impairments in inhibition. Gender effects were also found on the Inhibition factor, and on the task variables included in this factor, with girls outperforming boys.
Unlike the effects of group, the effects of gender were not limited to tasks measuring inhibition. Significant gender effects were also reported on the Verbal Fluency task and the number of errors in the Shape School Switch condition, with medium effect sizes. Girls performed better than boys on these task variables. In addition, the effect of gender was marginally significant for the OCTC, but the effect size was smaller than for the other significant effects of gender. Again, girls were outperforming boys. Only one significant interaction effect for group and gender was manifested, i.e., the number of errors on the Shape School Switch condition [group × gender, F(1, 181) = 9.91, p = 0.00]. Covarying for age did not yield significant effects on any task variable, except for the Digit Span (words) [age, F(1, 181) = 5.61, p = 0.02]. No significant results for group or gender were found on this variable.
To control for the influence of Attention Problems, ANCOVAs were also carried out with Attention Problems as an additional covariate. The Inhibition score remained significant for group [group, F(1, 181) = 5.27, p = 0.02] and gender [gender, F(1, 181) = 10.66, p = 0.00]. The significant effect of group for the number of errors on the Shape School Inhibition condition was also maintained [group, F(1, 181) = 5.95, p = 0.02]. The effect of group for the variable Shape School Inhibition incorrect responses disappeared and the effect of the Go/No go did not yield significant results for group anymore.
Discussion
The aim of the current study was to examine EF in preschool children with aggressive behavior in a population-based sample. Our findings demonstrated that preschoolers who show primarily aggressive behavior displayed impairments in inhibition. In this group of aggressive preschoolers these impairments in inhibition were maintained after controlling for attention problems. In addition, gender differences in EF were found, with boys exhibiting more EF deficits than girls.
We first examined which EF could be distinguished in this non-clinical sample of preschool children by factor analysis. Although the neuropsychological tests in this study were aimed at assessing working memory, inhibition, fluency, and set shifting, the only EF factor, which could clearly be distinguished was Inhibition. The other EF could not be distinguished as separate constructs. This implies that the differentiation of EF at this young age remains a complicated issue (Senn et al.
2004). EF is still maturing in the preschool period and will to develop into more specific functions. Considering the complexity of the construct of EF in the preschool years, the finding of inhibition as the only EF factor in the present study indicates that inhibition is a robust concept of EF at four years of age.
An explanation for identifying only inhibition as an EF factor is that inhibition is one of the first EF to emerge (Barkley
1997; Brocki and Bohlin
2004). The ability to inhibit prepotent responses generally increases significantly over the preschool period, which is necessary for the exertion of control over one’s behavior (Espy
1997). Therefore, it appears to be a prominent feature in this period of rapid cognitive development, whereas other EF cannot yet be clearly detected at this young age and develop over time (Korkman et al.
2001). Aggressive behavior was only found to be associated with impairments in inhibition and not in other EF. In the meta-analysis of Oosterlaan et al. (
1998) and in the study of Hughes et al. (
1998) deficits in inhibition were also related to aggressive behavior problems. Inhibition problems constitute a key characteristic of aggressive behavior problems and are found to be persistent over time (Brophy et al.
2002). Although the relation between inhibition problems and DBD is well established in school-aged children, adolescents and adults (Morgan and Lilienfeld
2000), finding this relation in a sample of young children who show aggressive behavior is important, because it might be that inhibition plays a crucial role in the developmental trajectories of aggression.
In contrast to evidence suggesting that EF deficits in preschoolers with aggressive behavior are mainly explained by symptoms of ADHD (e.g. Sonuga-Barke et al.
2002; Thorell and Wåhlstedt
2006), this study reported results which demonstrated that inhibition deficits were significantly related to aggressive behavior, irrespective of attention problems. Attention problems did not influence the effect of aggression on the Inhibition factor. It should, however, be specified that the level of attention problems was relatively low in the present sample. These findings are consistent with the notion that when problems concerning the inhibition of behavior arise, the risk of the development of aggressive behavior is increased (Kochanska et al.
2000). However, even in young children who primarily show aggressive behavior, attention problems remain important because of deficits in working memory and set shifting.
Gender differences in inhibition, verbal fluency, working memory and set shifting were found irrespective of aggression or attention problems. These findings contrast with studies of Overman (
2004) and Thorell and Wåhlstedt (
2006), in which no differences in EF of boys and girls at this age were manifested. However, gender differences in effortful control, i.e., the ability to inhibit a dominant response to perform a non-dominant response have often been reported. Kochanska et al. (
2000) found gender differences with girls outperforming boys in inhibiting impulsive responding, at ages as young as of 22 and 33 months. Olson et al. (
2005) reported a significant effect for gender; girls showed higher levels of effortful control than boys at age three. The more rapid developmental maturation of girls (Keenan and Shaw
1997) might be responsible for their higher level of inhibitory skills and other EF in the preschool period and could explain why preschool boys make more mistakes on EF measures. Especially the relative delay in inhibition of preschool boys makes them more prone to the development of aggressive behavior. In further research it is important to study the inhibitory skills of large samples of boys and girls separately in order to more clearly detect gender differences. We also found a difference between boys and girls in verbal fluency, with girls being more verbally fluent than boys. This is relevant as poor verbal skills compose a risk factor for the development of aggressive behavior problems (Loeber et al.
1998). Children at high risk for DBD are children who experience difficulties regarding both executive and verbal cognitive skills (Moffitt
1993). In addition, socialization practices might contribute to gender differences in inhibition. In general, girls are encouraged to exert more control over their behavior than boys (Keenan and Shaw
1997). As a result, girls learn to show more inhibited behavior than boys from an early age on, which might be reflected in our findings. Finally, the marginally significant effects of gender on other EF might be due to a lack of statistical power, because the number of girls in this sample was relatively small. Therefore, EF such as working memory and set shifting also require to be analyzed separately for boys and girls in future studies.
This study has a number of limitations that need to be considered. First, the majority of the parents of the children in our sample showed a high educational level. Therefore, our findings have limited generalizability to children from less educated parents. Second, we employed extreme inclusion criteria. We compared a group of children who showed a low level of aggressive behavior to a group of children who showed a relatively high level of aggressive behavior; these groups represent the extreme ends of the behavioral spectrum. The finding of EF impairments in the aggressive group can only be interpreted within this context. Third, we used an experimental battery of tasks, consisting of six neuropsychological measures. The use of other EF measures might have yielded different EF factors, and might thus have captured other EF deficits present in this group of young aggressive preschoolers. An additional limitation is that the measures used did not assess ‘pure’ EFs; most EF tasks measured more than one EF. Tapping pure EFs is conceptually not feasible, because almost every task requires a subject to keep rules in mind and thus addresses working memory next to the EF which was aimed to be measured. By conducting a factor analysis, we deducted the common variance between the variables from the measures used, resulting in a latent Inhibition factor that represents a more pure measure of EF. The use of tasks which assess the more motivational aspects of EF are of interest, considering that school-aged children and adolescents with DBD show dysfunctions on these kinds of tasks (Blair et al.
2001; Matthys et al.
1998,
2004; Van Goozen et al.
2004). Future studies are needed to clarify the role of reward and the more affective aspects in the EF of children who show aggressive behavior in this young age group.
The relevance of assessing EF in the preschool years is clearly supported by the current findings. Results of this study show that impairment in inhibition is a correlate of aggressive behavior in preschool children, regardless of attention problems. This study adds to the growing body of literature on the role of neuropsychological deficits in the development of behavioral disorders. Since inhibition deficits may contribute to the development of aggressive behavior and DBD, future research should assess EF in aggressive preschoolers longitudinally to gain insight in the role of EF deficits as precursors or risk factors for the development and persistence of DBD.