Introduction

Recently, two studies have found preliminary support for an elevated prevalence of overweight in children with an Attention Deficit/Hyperactivity Disorder (ADHD) and vice versa [2, 25]. In addition, just like overweight children [6, 26], children with ADHD have difficulties with delay of gratification facing food cues [4]. Holtkamp and colleagues [25] suggested it could be that comparable mechanisms underlie overweight and ADHD, with impulsivity or dysfunctional inhibitory processes as a binding factor.

In everyday parlance the term impulsivity typically refers to behaviour that incorporates a component of rashness, lack of foresight, or planning or to behaviour that occurs without reflection or careful deliberation. We all engage in such impulsive acts from time to time, some of us more so than others. And so, the term is best understood to reflect a continuum [13]. Although impulsivity is widely used, research has traditionally encountered several problems. Firstly, there is neither a widely accepted definition nor a ‘golden standard’ to measure impulsivity. However, there is broad consensus on the fact that impulsivity should not be seen as a one-dimensional construct, but as a construct consisting of a number of related dimensions [41] like the capacity to tolerate delay for reward, to inhibit a response already initiated, and to estimate the passage of time [16]. Secondly, different measures of impulsivity are poorly related to each other. This is particularly relevant when comparing self-reported and behavioural measures of impulsivity [60].

Probably the most widely used behavioural measure of reflection-impulsivity is the Matching Familiar Figures Task (MFFT) [28]. This task requires subjects to search a number of similar pictures for one that matches a criterion picture exactly. An impulsive approximation is associated with short latencies and many incorrect answers [29]. Although the MFFT evoked some controversy and cannot be considered as a valid diagnostic measure of impulsivity, it is still one of the primary instruments used in research to investigate impulse control. Moreover, at group level it appears to discriminate between children with ADHD and their normal counterparts [18].

The present study addresses impulsivity in overweight children thereby taking into account the measurement problems in capturing the complexity of construct. Next to MFFT, we will use child self-reports and an interviewer-based measure when looking for impulsivity symptoms. Moreover, we will interview the parents not only in our search for detecting impulsivity symptoms but also for understanding impulsivity as a personality trait.

Personality traits can be defined as stable characteristics, influencing our thoughts and behaviour in a variety of situations [11]. According to this definition, personality may, at least partly, be associated with lifestyle and eating patterns. Problem behaviour is repeatedly shown to be associated with certain personality profiles predicting the increased risk of alcohol consumption, drug use, smoking and high risk sexual encounters [27], and bulimic overeating in adults [31, 58, 61]. Within the field of obesity, Fassino and colleagues [20] found obese women’s temperament to be more impulsive.

In children, a maternally rated difficult temperament, measured as hyperactivity, unpredictability, and low attention span, has been associated with excess weight gain in middle childhood [9]. Although preliminary evidence suggests that in early life a child’s temperament may contribute to later obesity, and recent meta-analyses have provided stability estimates from 3 years to adulthood [48], prospective studies on impulsivity as a personality trait are rare [39, 43].

Few studies have analysed gender differences, despite research indicating that gender may moderate the relationship between psychological characteristics and Body Mass Index (BMI) [10, 46]. Rydén et al. [52] found that differences with respect to impulsivity in women were small but moderately higher in obese men. Furthermore, a meta-analysis on gender differences in delay of gratification tasks [54] suggests that girls are better able than boys to control their impulses in choice situations. Unfortunately, no comparisons between overweight and normal weight samples were made.

Given the limited amount of evidence in children, the central aim of this study is to investigate whether or not obese children have a stronger tendency to act on impulse than normal weight children. Both parent and child based measures, traditional questionnaires and performance based and interview methodologies are used in this study. Three subsequent research questions are discussed. Firstly, we will explore whether or not obese children score higher on the MFFT. Secondly, we will investigate whether or not obese children show symptoms of impulsivity as diagnosed in ADHD. Thirdly, we will investigate whether or not obese children have a specific personality profile. In addition, we will include gender as a factor in the analyses.

Method

Participants

Overweight children starting treatment in a paediatric centre between the beginning of July 2003 and the beginning of September 2003 and their parents were invited to participate. Children younger than 10 or older than 18, mentally retarded children, and children with an adjusted BMI of <120% were excluded from the study. All families agreed. The overweight group consists of 56 children, 25 boys, and 31 girls. The control children in the study were selected from ordinary primary and secondary schools and were matched with the overweight group according to age and gender (25 boys and 31 girls). Control children with an adjusted BMI fewer than 80% or above 120% were excluded from the study. Three of the 56 selected control children dropped out.

Measures

Body weight. The BMI (weight/height2) was calculated for each child. In order to make BMI comparisons between children of different ages, the adjusted BMI was used in this study. The formula is [Actual BMI/Percentile 50 of BMI for age and gender] × 100. The percentiles 50 of the BMI were based on European normative data [21].

The Matching Familiar Figures Task [26] is administered individually and comprises two training items and 12 measure items. Each item consists of a model drawing and 6 (5 to 12-year-old version) or 8 (12 to 18-year-old version) highly similar drawings, one of which is identical to the model drawing. The child has to point out the picture that is identical to the model drawing. If mistaken, the child tries again until the correct picture is identified. Performance is assessed by the time taken to give the first response (latency) and the total number of errors across items. These scores were then converted into normalised standard scores for each subject and then transformed to obtain an Impulsivity score (I: errors minus latency) and an Efficiency score (E: errors plus latency) [53]. Large positive I scores are indicative of impulsivity, and large negative I scores indicate reflectivity. Highly positive E scores indicate inefficiency and highly negative E scores indicate efficiency. In addition, we computed the variance of raw latencies and errors for each subject. The MFFT has satisfactory reliability and construct validity [19]. Alpha coefficients were .95 for the latency scores and .63 for the error scores in this study.

There is still discussion on the impact of IQ as a potential confounder when performing the MFFT [38]. Therefore, correlations between IQ and Impulsivity and Efficiency were explored in a pilot study (n = 56) and analyses revealed no significant findings, indicating that response patterns on the MFFT cannot be predominantly explained in terms of cognitive ability in children.

The Structured Clinical Interview for DSM-IV, Childhood version (KID-SCID) [23] similar to the adult version, the KID-SCID is a semi-structured instrument designed to generate childhood DSM-IV symptoms and diagnoses. In the current study, only the ADHD module, which is part of the disruptive behaviour disorders section was assessed based on an interview with the child. The first part of the ADHD-module assesses nine criteria referring to attention deficit problems. The second part of the module assesses hyperactivity (6 items) and impulsivity symptoms (3 items). Each criterion is scored as present (‘1’ score) or absent (‘0’ score). In case of doubt, the criterion in question is assigned a ‘0’ score. In the study of Timbremont et al. [57], child–parent agreement rates varied between 88.7% and 100%, which indicates high reliability. In addition, preliminary results of a study by Matzner et al. [34] showed fair to excellent test-retest reliability for the disruptive behaviour disorders module (alphas between .63 and .84). Pilot data also indicated excellent inter-rater reliability for the disruptive behaviour module (α = 1.0 for ADHD) [33]. In this study, Cronbach alpha internal consistency coefficients were 0.82 for the attention deficit subscale, 0.74 for the hyperactivity, and 0.63 for the impulsivity subscale respectively.

The Attention Control Scale for Children (ACS-CH) [14] assesses two types of attention control in children, scored on a 4-point Likert-scale. The attention focusing subscale (7 items) measures the ability of the child to stay focused on relevant information. The attention shifting subscale (10 items) measures the ability to shift attention from one source of information to another. Higher scores on the ACS-CH are indicative of better attention control. In the present study, Cronbach alpha internal consistency coefficients were 0.65 for attention shifting and 0.70 for attention focusing respectively.

The Disruptive Behaviour Rating Scale (DBD) [42, 45] was developed to measure externalising behaviour by interviewing the parents. The DBD contains 42 items, which are scored on a 4-point Likert scale. The questionnaire contains 4 scales composed of the DSM-IV-TR items for ADHD inattentive subtype, ADHD hyperactive/impulsive subtype, oppositional defiant disorder, and conduct disorder. The higher the score on the different scales, the more the child is impaired. Adequate psychometric properties have been reported [42]. In this study, only the Hyperactivity/Impulsivity subscale (9 items) was administered to assess symptoms of hyperactivity and impulsivity in the children. In this study, the Cronbach alpha coefficient for the Hyperactivity/Impulsivity subscale was 0.85.

The Hierarchical Personality Inventory for Children (HiPIC) [36] assesses the Big Five personality dimensions in children, i.e. Extraversion, Benevolence (Agreeableness), Conscientiousness, Emotional Stability (Neuroticism), and Imagination, and 18 facets hierarchically organised under the primary traits. All 144 HiPIC items refer to overt behaviour of children and adolescents between 6 years and 17 years of age, have a similar grammatical format, and are scored on a 5-point Likert type scale. Given the limited vocabulary of both children and young adolescents, HiPIC ratings are usually provided by parents. The factor structure proves to be highly replicable across both childhood and adolescence, with Cronbach alphas for domains and facets in adult observer samples usually exceeding 0.80 [36].

Procedure

The local research ethics committee approved the protocol of this study. Families of both overweight and control children were invited to participate in the study by mail. After informed consent was obtained, child questionnaires were administered in small groups for both the overweight and normal weight children. The order in which the different instruments were administered was randomised. Each child was taken separately for the ADHD module of the KID-SCID interview and the MFFT. All children were tested by the same person, a trained bachelor in clinical psychology. The interviewer had no prior knowledge of the hypotheses of the study. Parent questionnaires were posted to the parents with the explicit request to send them back as quickly as possible. We assessed the socio-economic backgrounds of individuals using the Hollingshead Index of Social Position [24].

A General Linear Model procedure of SPSS 12.0 was used to conduct MANOVA-analyses with group (normal weight versus overweight) and gender as the between-subject factors, and the different impulsivity measures as the set of dependent variables (a) different measures of the MFFT (after transformation into z-scores), (b) different subscales of the ACS-CH, (c) The Hyperactivity/Impulsivity scale of the DBD-Q (d) five personality dimensions of the HiPIC. The term significant refers to P ≤ 0.05. To minimise the probability of Type-I errors, only when the multivariate F was significant, follow-up univariate analysis (ANOVAs) were conducted. The highly skewed count variables on the KID-SCID did not met the distributional assumptions of most classical approaches to data analysis such as (M)ANOVA. Therefore non-parametric Mann-Whitney Tests were conducted here.

Results

Missing values

Despite several reminders, response rates for the parent questionnaires in the overweight group were rather low (54%). This was mainly due to the fact that these parents were difficult to reach, a problem frequently encountered in clinical groups. With a response rate of 77%, parents of control children seemed less reluctant to participate. ANOVAs did not reveal significant differences between children from parents that did or did not return the two parent questionnaires for all MFFT scores, the subscales from the KID-SCID, and the ACS-CH. Therefore, we consider the completed parent questionnaires to be sufficiently representative of the total sample. The cases with missing values were deleted listwise in the analysis of the parent questionnaires.

Description of the sample characteristics

The mean age of the overweight children was 13.50 years (SD = 2.33). Mean height and weight were 161.64 cm (SD = 9.87) and 72.66 kg (SD = 14.97) respectively, with a mean adjusted BMI of 149.32% (SD = 20.41) (see Table 1). According to the CDC growth charts, [40], overweight children displayed BMI z-scores between 0.98 and 2.37 and could be categorised as overweight (120–140% overweight, 37.5%), mildly obese (140–160% overweight, 32.1%) or obese (>160% overweight, 30.4%). About 30% of children were in primary school and 70% went to secondary school. Subjects were from families spanning all socio-economic strata (upper and upper-middle class: 7.3%; middle class: 49.1%; lower-middle and lower class: 43.7%).

Table 1 Descrpition of the sample characteristics
Full size table

Children in the control group had a mean age of 13.34 years (SD = 2.32). All controls were within the normal weight range; their mean height and weight were 161.38 cm (SD = 13.79) and 49.24 kg (SD = 13.08) respectively, with a mean adjusted BMI of 100.32% (SD = 8.42). According to the CDC growth charts [40], BMI z-scores ranged from −1.73 to 0.91. About 26% of control children were in primary school and 74% attended secondary school. Again, all socio-economic strata were represented, with most families situated in the middle classes (upper-middle class: 14.3%; middle class: 68.6%; lower-middle and lower class: 17.1%).

No differences were found between the overweight children and the controls with respect to age, F(1,108) = .13, P > .10, or gender χ²(1) = .07, P > .10. Analyses on the socio-economic background as measured with the Hollingshead ISP revealed that more children in the overweight group belonged to the category ‘lower-middle and lower class’, LR(2) = 7.37, P < .05.

Answering patterns on the MFFT in overweight versus normal weight children

A significant multivariate main effect for group was found, F(2,104) = 6.55, P < .01, η² = .11, (power of .90) indicating a significant difference between the overweight and the normal weight group as regards response behaviour on the MFFT. No significant multivariate main effect for gender, F(2,104) = 2.11, P > .10, η² = .04, nor a multivariate interaction effect between group and gender, F(2,104) = 1.77, P > .10, η² = .03, were found. Univariate analysis further showed a significant main effect of group for Impulsivity, F(1,108) = 10.18, P < .01, η² = .09 (see Table 2). The positive mean for I in the overweight group indicated greater impulsivity in these children, whereas the negative mean for I in the controls indicated that control children were less impulsive and showed more reflectivity. There was a trend effect of gender for Efficiency, F(1,108) = 2.90, P = .09, η² = .03, girls tend to be more efficient than boys in both groups.

Table 2 MFFT scores and child reported attention control according to group and gender
Full size table

Next, a MANOVA with group and gender as the between-subject factors and latency variance and error variance as the dependent variables was conducted. Only the multivariate main effect of group was significant, F(2,104) = 5.50, P < .01, η² = .10, (power of .84), indicating a difference between normal weight children and overweight children with respect to variation in responding across MFFT items. Follow-up univariate analyses revealed that latency variance was significantly smaller in overweight children than in control children, F(1,108) = 5.84, P < .05, η² = .05, whereas error variance was significantly larger in overweight children than in controls, F(1,108) = 9.84, P < .01, η² = .09.

Do overweight children show elevated levels of attention deficit problems, impulsivity, and hyperactivity symptoms compared to normal weight children?

ACS-CH. Looking at the Attention Control Scale for Children (see Table 2), a MANOVA revealed a significant multivariate main effect for group, F(2,102) = 4.06, P < .05, η² = .07, (power of .71) and a significant multivariate interaction between group and gender, F(2,102) = 3.15, P < .05, η² = .06, (power of .59). The multivariate main effect of gender was not significant, F(2,102) = .88, P > .10, η² = .02. Further univariate analysis showed a main effect of group, F(1,106) = 7.12, P < .01, as well as a significant interaction effect between group and gender, F(1,106) = 5.16, P < .05, for attention focusing. Overweight boys were worse at focusing their attention than control boys, while no significant differences were found between overweight girls and control girls. The effect of group was also significant for attention shifting: overweight children were worse at shifting their attention than their normal weight counterparts, F(1,106) = 4.90, P < .05. The effect of neither gender nor the interaction between group and gender was significant for attention shifting.

KID-SCID. The response variables in the present analysis were counts of impulsivity, hyperactivity, and attention deficit symptoms. As is typical of such symptom data, each distribution was highly skewed, with modal and median values being at or around zero. When looking at both groups separately, the range of impulsivity symptoms (range 0–3), was between 0 and 3 in both groups, with 88.7% of counts being 0 in the control group and 76.8% of counts being 0 in the overweight group respectively. The number of hyperactivity symptoms (range 0–6) varied between 0 and 4 in controls, and between 0 and 6 in the overweight group, with 86.8% of counts being 0 in controls and 64.3% of counts being 0 in overweight children respectively. The attention deficit symptoms (range 0–9) varied between 0 and 5 in controls and between 0 and 7 in the overweight children, with 81.1% of counts in controls and 55.4% of counts in overweight children being 0.

Mann-Whitney tests revealed a significant main effect of group for the number of impulsivity symptoms, U = 1298.00, P < .05; the number of hyperactivity symptoms, U = 1119.50, P < .01, as well as the number of attention deficit problems, U = 1058.00, P < .01, as measured with the KID-SCID clinical interview (see also Table 3). There was a significant effect of gender for all three behavioural symptoms too, U = 1283.00, P < .05, U = 1107.50, P < .01 and U = 897.50, P < .001 respectively. When overweight boys were compared with control boys, overweight boys showed more impulsivity, U = 236.50, P < .05, hyperactivity, U = 183.00, P < .01, and attention deficit symptoms, U = 127.50, P < .001, than control boys. There were no significant differences between overweight girls and control girls.

Table 3 Counts of impulsivity, hyperactivity and attention deficit symptoms according to group and gender on the KID-SCID
Full size table

Six overweight children (10.7%) obtained a clinical interview-based diagnosis of AD(H)D (3 inattentive types, 2 hyperactive types, and 1 combined type). Six control children (11%) obtained a clinical interview-based diagnosis of AD(H)D (1 inattentive type, 3 hyperactive types and 2 combined types).

Parent DBD-Q. With respect to the ANOVA for the Hyperactivity/Impulsivity scale of the DBD-Q, neither a main effect of group, nor a main effect of gender, nor a significant interaction between group and gender could be found (see Table 4).

Table 4 Parental reports on the DBD Hyperactivity/impulsivity scale and on the HiPIC personality measure, according to group and gender
Full size table

Personality in overweight versus normal weight children

Differences between groups with respect to personality dimensions as reported by the parent were explored using a MANOVA with group and gender as the between subject factors and the five personality dimensions as the dependent variables (see Table 4). A significant multivariate main effect of group was found, F(5,52) = 3.11, P < .05, η² = .23, (power of .83) indicating a significant difference between the overweight and the normal weight group regarding personality dimensions. No significant multivariate main effects for gender, F(5,52) = 1.22, P > .10, η² = .10, nor a multivariate interaction effect between group and gender, F(5,52) = .39, P > .10, η² = .04, were found. Follow-up univariate analyses showed a main effect of group for Conscientiousness, F(1,59) = 7.99, P < .01, η² = .13: the overweight group scored significantly lower on Conscientiousness than the control group. There were no significant main effects of group or gender or any interaction effects for the other domains.

Because differences between the study sample and the control group were found for SES, analyses were repeated introducing the total ISP-score for SES as a covariate in the multivariate models. However, no main effects for SES were detected and SES did not affect the findings significantly. Although overweight and normal weight children were comparable on other potentially confounding variables (age, diagnosis), we also controlled for their impact on the analyses. Analyses were repeated whereby subjects with an ADHD diagnosis were left out. This revealed the same results and no loss of significance. It was not completely unexpected that main effects of age were found for the scores on the MFFT and the HiPIC but not on the ACS-CH and the DBD. Except for the analysis on the MFFT, all significant and non-significant findings on the different measures remained. Although the multivariate main effect of group on the MFFT was still significant, F(2,103) = 7.74, P < .001, follow-up ANOVAs revealed that the trend significant finding for Efficiency disappeared. The findings for Impulsivity, latency variance, and error variance remained.

Discussion

This study is one of the first to investigate impulse control in overweight children thereby taking into account the complexity of the impulsivity construct. As hypothesised, results on the MFFT revealed that overweight children respond more impulsively (faster and more inaccurately). Differences between overweight and normal weight children were also found for symptoms of ADHD and on personality characteristics, thereby also confirming hypotheses two and three.

When impulsivity is studied, it seems that gender differences should be taken into account. The present findings are consistent with previous research demonstrating better impulse control in girls than in boys [53]. However, apart from the reported main effects of gender, comparing overweight and normal weight children on the different measures used in this study revealed some interesting gender specific findings. Not surprisingly, the findings indicate that especially overweight boys show more impulsivity, hyperactivity, and attention deficits as measured via the KID-SCID-interview as well as problems with focussing attention on the attention control scale compared with their normal weight controls. The phenomenon of boys being more vulnerable to impulsivity problems is not new. More research is needed to clear out the robustness of the gender differences found and to explore whether a gender-specific pathway leads to impulsivity in boys.

Despite the fact that the results from the child behavioural test, the questionnaire, and the child interview revealed significant results, parents of overweight children report neither a heightened impulsivity nor a heightened hyperactivity in their children compared to parents of controls as measured with the DBQ. There are several possible explanations for these findings. Firstly, we had to recognise that not all parents were willing to participate. Since ANOVAs revealed no significant differences for the child measures between children of parents who did and did not return the parent questionnaires in either group, selection biases are possible, but not very plausible. Secondly, the type of responses required on the MFFT may show minimal correspondence with children’s impulsive and inattentive behaviour exhibited at home [5]. Children’s impulsive and inattentive behaviours may differ as a result of the greater novelty and higher frequency of performance feedback inherent to the test setting [3]. Nonetheless, the MFFT provides information about inattention and impulsivity that contributes, but is not identical, to the exhibition of impulse control difficulties in natural settings [22]. As Achenbach et al. [1] have proposed, parent and child can be equally informative because they can highlight variations in judgements of the child’s functioning across interaction partners and situations. Hence, assessment strategies can’t solely rely on parental reports and adopting a multi-informant approach in future research endeavours remains essential.

With respect to personality as reported by the parents, overweight children display lower Conscientiousness scores compared to controls, indicating that they find it more difficult to concentrate, persevere and be well organised and that they are less motivated to perform than normal weight children. Low Conscientiousness is a trait that has been frequently associated with the emergence of problematic impulsive behaviours. Although personality traits are indisputably consistent across time and age [48], modest changes had to be acknowledged either because of environmental influences, or by biologically based intrinsic maturation [35]. In a recent study by De Fruyt et al. [15], structural continuity was found for impulsivity traits, as measured with the HiPIC, with an additive genetic variance varying between .36 (time 1) and .38 (time 2) in a twin and sibling sample (explained variance: 40%) with smaller (near zero) estimates for shared environment but a greater impact of non-shared environment (.64 at time 1 and .62 at time 2; explained variance: 54%). These data suggest that genes determine continuity of personality across time but environmental factors, unique to an individual, operate across time also and both were important to consider. Given the moderately high levels of personality consistency from childhood to young adulthood [11], the question still remains as to whether a certain personality profile can explain the overweight problem in children. Nonetheless, we can hypothesise about the relation between personality and obesity in three different ways [30]: (a) an impulsivity prone personality profile predisposes to overeating in genetically vulnerable persons, especially in unhealthy food environments [59]; (b) being obese in itself may have an effect on personality; and (c) a combination of the first two mechanisms is possible. Given that temperament traits resembling hyperactivity have been shown to contribute to the accumulation of body fat in childhood [8, 9], one should bear in mind the possibility of a common underlying temperamental factor which explains both an impulsive prone personality and obesity.

An important limitation of the study is that most of the variables have very large standard deviations. The finding that latency variance is smaller for the overweight group was unexpected as variability tends to be increased in clinical groups including those with ADHD. However, error scores are more widely correlated with other conventional test scores than are the latencies, this difference may result from the fact that error scores are determined more by stable, culturally valued abilities and attitudes, and by the developmental status of the child. Latency or speed, however, may be more influenced by such context-specific variables as set, expectation, and task instructions [53]. In addition, symptom counts obtained via the KID-SCID had a skewed distribution and were therefore analysed using non-parametric tests. Given that these tests are more susceptible to Type-II errors, the fact that we do find significant differences for the KID-SCID data suggests that our findings are quite robust. Moreover, this study is also susceptible for type I errors, because of multiple testing. Although our findings were confirmed by different informants and with different measures, it seems that the findings of this study should be replicated before general conclusions can be made.

When studying impulsivity in children, different covariates can be included in the model. Arguments can put forward to include post hoc analyses on subgroups, thereby taking diagnosis, age, gender, and SES into account. However, given the exploratory nature of the study, the sample size was limited and testing a full model including all factors was not within the scope of the present study. Based on the literature, we found most arguments to include gender. However, the study also revealed differences between overweight and normal weight children on SES and we checked therefore for the impact of this variable in the analyses. Future research can overcome this limitation, by including more children representative of the different SES groups, preferable in larger samples. Differences in IQ between the two groups could also account for the variance of the relevant items. Given its high correlation with IQ, especially in younger children, future research should also control for IQ.

Other confounders can be explored in the future as well. Overweight children watch TV frequently [49] and early television exposure has been associated with inattention problems at age 7 [12]. However, others revealed that television exposure is only a weak predictor of later ADHD symptoms: effect sizes for the relationship between television exposure and symptoms of ADHD were close to zero and not statistically significant [55]. No studies have explored the association between television exposure and inattention problems in overweight children. And so, it remains to be seen whether or not and how inattention problems are related to obesity.

This study is one of the first to demonstrate evidence for impulsivity proneness in overweight children thereby showing that difficulties in controlling one’s impulses go together with attention deficits. Future research should focus on the implications of these characteristics. There is evidence in infants, adolescents, and adults that those individuals who are well able to focus and shift attention are better at tolerating frustration and exercising self-control [50, 51]. Self-control, the opposite of impulsive behaviour, seems correlated with skills in directing one’s attention [56]. Perhaps, this is because one way of resisting temptation is directing attention away from it. Metcalfe and Mischel [37] posit that people can use ‘internal strategies’, i.e. direct attention, to activate those types of brain circuits that are responsible for rationality and calculation as opposed to impulsivity.

Although the present study was designed to focus on general measures of impulsivity, further research and replication of the present study findings is indicated in overweight children with respect to the link with eating behaviour. Preliminary findings indeed suggest that overweight children eat faster, take more bites per minute, chew less, interrupt their meals less, and play less with their food than normal weight children do [17, 32]. Furthermore, two studies have found that obese children were less able to delay gratification for food than children of normal weight, according to direct choice measures [6, 26]. However, the link with impulsivity still has to be explored.

The study focused on the assessment of impulse control in referred children starting an obesity treatment programme. Although the overweight group consisted of children from all overweight categories, findings need to be replicated in non-referred overweight samples. Based on previous research [7] it is reasonable to assume that overweight children seeking treatment display comorbid psychopathology and were different from those not seeking help. Nevertheless, the comorbidity of ADHD and obesity in referred groups was only recently detected and will lead to new challenges when planning treatment programs for children with comorbid problems [2, 25].

Impulsivity proves an obstacle to effective treatment [39, 47]. Given that a person’s level of impulsivity is enduring but not immutable, and biologically linked but susceptible to learning-based influences [56], high impulsive overweight children would definitely benefit from self-control skills. Since self-control is correlated with skills of focusing attention, learning self-instructions can be helpful as well [56]. It can also be useful to explicitly focus on parenting skills because it was previously shown that impulsive children have greater difficulties understanding the consequences of their behaviour and therefore need more consequent parenting, a more transparent communication of rules and more reinforcement than other children [44].

To conclude, overweight children in this study display a stronger tendency to act on impulse than normal weight children as measured with well-evaluated measures of impulsivity. However, taking account of the complexity of the construct, further research is needed to explore whether or not the significant and non-significant findings can be replicated and whether or not the gender differences indicated different pathways leading to obesity. Specific in referred children, screening for comorbid impulsivity symptoms may enhance treatment practices.