The present study is part of the Bergen Child Study (BCS). The protocol and population of the stages in the first wave of BCS are described in detail in separate publications 1920, and only a short presentation will be given here. Briefly, the original BCS included three stages: screening for behavior problems and psychiatric disorder of the whole Bergen 7-9-year-old population using the Strengths and Difficulties Questionnaire (SDQ) 21, the Autism Spectrum Screening Questionnaire (ASSQ) 22, and items pertaining DSM-IV symptoms of ADHD and ODD from Swanson, Nolan, and Pelham, version IV (SNAP-IV) 23, supplemented with a number of items designed specifically for use in the BCS (stage 1); Development and Well-Being Assessment (DAWBA) 24 interviews with parents of children defined as screen positive in stage 1 and a sample of screen negative children (stage 2), and in-depth neuropsychiatric/neuropsychological assessment of a subsample of "DAWBA positive" (i.e. children who obtained a diagnosis according to DAWBA) and "DAWBA negative" children from stage 2. Children with a "chronic somatic disorder" reported by parents in stage 1 of the study 25 were included, regardless of participation in stage 2.

The aim of stage 3 was to investigate the neuropsychological function (motor, emotional and cognitive) of children with known mental health problems and normal controls. A total of 329 children met together with their parents for a 6 hours examination procedure at the Neuropsychology outpatient clinic at the University of Bergen. The examination included a diagnostic semi-structured interview of parent and child (Kiddie-Sads-Present and Lifetime Version) 26; the Wechsler's Intelligence Scale for Children, third version (WISC-III) 27, and the AANT 28. Of the 286 children who completed the ANT with an accuracy scores above 50%, all children with a diagnosis in remission (N = 8) and the group of children with Oppositional Defiant Disorder (ODD) without ADHD (N = 11) were excluded. The last group was excluded due to the overlap in symptomatology. The final sample (N = 157) included all children with a definite ADHD diagnosis (N = 45), children with other definite diagnoses (N = 55), and children without any diagnosis or ADHD symptoms (N = 57) (Figure 1). The ADHD group was further divided into a group of children taking central stimulants at the inclusion of the study (N = 9) and children not taking any central stimulants (N = 36). The last group included mainly newly diagnosed children. All children taking central stimulants had an ADHD diagnosis at the entry of the study, and their parents were asked to withhold the medication on the day of clinical examination. Because of the low number of girls in the ADHD group (N = 13), gender differences were not investigated in the present study.

The study was approved by the Regional Committee of Ethics on Medical Research in Western Norway and by the Ombudsman for Privacy in Research, Norwegian Social Science Data Services Ltd.

Kiddie-Sads-Present and Lifetime Version (Kiddie-Sads-PL) 26 was used as a diagnostic instrument. Kiddie-Sads-PL is a reliable semi-structured interview designed to evaluate current and past episodes of psychopathology in children according to the DSM-IV criteria 2629. The diagnosis was ascertained through an interview in two separate sessions on the same day, first with one or both of the parents and then with the child. Diagnoses were scored by the interviewer immediately after the assessment of both informants as either definite, probable (≥75% of symptom criteria met), in remission, or not present 26.

The ANT used in the present study is the original "child version" 30 downloaded from Jin Fan's webpage 28. The test has four cue conditions (no cue, center, double, orienting) and three flanker conditions (congruent, incongruent, neutral), and has been described in detail elsewhere 3031. All combinations of conditions are randomly presented in three blocks of 48 trials each. Overview of the calculations used in the present study is given in Table 1. The calculated measures are based on an Excel macro downloaded from Jin Fans webpage 28, supplemented by measures calculated according to the formulas given by Conners and collaborators 32. This file was imported into SPSS 13.0, which was used for all further analyses.

WISC-III 27 was used to assess intellectual function. The Full Scale IQ (FSIQ) was included in the present study, with scaled scores derived from Swedish norms 33.

The clinical examination was performed in an outpatient clinic at the University of Bergen. Trained psychologists administered the Kiddie-Sads-PL interview. The WISC-III and the ANT were administered by trained test-assistants, in a quiet room designed for testing. It was run on E-Prime software, on a stationary computer with a 17" computer screen. The children sat at a comfortable distance from the screen and used left and right thumb to press the left or the right mouse button, corresponding to a left and right pointing fish. The children were instructed to help feed the hungry fish as fast as they could by pushing the left or right button, according to which direction the fish was pointing. They were told that sometimes the fish would appear alone, and other times it would swim together with other fishes. In all cases, they were told to concentrate on the fish in the middle. They were also asked to keep their eyes on the fixation point during the presentations. The completion time was approximately 25 minutes.

A Pearson correlation analysis with Bonferroni correction was computed, including the FSIQ score, age, and ANT measures. For FSIQ score, age, network measures, error measures, RT, standard error (SE) of RT, variability of SE, and vigilance measures, separate one-way between-groups ANOVAs were calculated. Main effects were further explored with post-hoc tests. The Tukey HSD was used to explore differences between groups where equivalence of variances were assumed, and the Games-Howell test was used if variances were heterogeneous. Bonferroni corrected independent-samples t-test was used to compare the results of children in the ADHD group who regularly took central stimulants and the results of the non-medicated children in the same diagnostic group.

The age range of the participants was 7.9 to 11.9 years, with a mean age of 10. A one-way ANOVA showed a statistically significant difference between the three diagnostic groups in age, F(2, 154) = 3.311, p = .039, but Tukey HSD between group comparison revealed that the difference between the ADHD group and the other groups only bordered on significance (other diagnoses: p = .07; without any diagnosis: p = .06). The mean FSIQ score for all participants was 90.9 (Table 2). A one-way ANOVA revealed a statistically significant difference between the groups on the FSIQ measure, and a Games-Howell post-hoc test showed that the ADHD group obtained significantly lower FSIQ scores than the non-ADHD groups (Table 2 and 3).

A one-way between groups ANOVA revealed no statistically significant differences between the three diagnostic groups on the RT and error measures of the three attention networks (Table 2 and 3).

The overall accuracy on the ANT was 90.7%, and an analysis of the overall number of errors showed that 97.0% of all errors were wrong responses and omissions (Table 2). Outliers and perservations represented only 3.0% of all errors and less than 1% of all responses, and were not included in further analyses. A one-way between groups ANOVA of the remaining error measures showed a statistically significant main effect of group (Table 3). A Games-Howell post-hoc test revealed statistically significant higher overall error scores in the ADHD group than the non-ADHD groups, with significantly more wrong responses than the non-ADHD groups (Table 2 and 3). The number of omission errors was also higher in the ADHD group than in the non-ADHD groups, but the differences were not statistically significant (compared to the group with other diagnoses, p = .09 and the group without any diagnosis, p = .06).

A one-way between groups ANOVA was performed separately for the overall Hit RT and the variability measures (Hit RT SE, variability of SE, Hit RT block change, Hit SE block change). A statistically significant main effect of group was found on the Hit RT SE measure (Table 2 and 3). Between group Tukey HSD comparisons revealed that the ADHD group showed significantly larger Hits RT SE than the group without diagnosis and border on significance when compared to the group with other diagnoses (p = .054). No main effect of group was found on the other measures.

A Bonferroni corrected Pearson's correlation analysis was performed between FSIQ scores, age and ANT measures (Table 4). The analysis revealed a high correlation between age and the overall Hit RT and the variability measures, showing that the older the children, the faster they responded, with a higher level of consistency. Table 4 shows that this only applied to the two non-ADHD groups. Correlations between FSIQ scores and the ANT measures were more widespread, including both error and variability measures. Table 4 shows that this only applied to the ADHD group, showing that the lower the FSIQ scores, the less accurate the children responded and with less consistency.

When FSIQ was included as a covariate in the ANOVA analyses, the main effect for group was no longer statistically significant for the overall error measure (F(2, 153) = 0,689, p = .504) and the Hit RT SE measure (F(2, 153) = 0,316, p = .729).

The group of ADHD children taking central stimulants regularly (N = 9) was compared to the other ADHD children (N = 36) on selected ANT measures (overall errors, Hits RT, Hits RT SE) and FSIQ scores. Bonferroni corrected independent-samples t-test showed no statistically significant difference between the two groups.

As in all research including children with an ADHD diagnosis, the present results are colored by the high degree of heterogeneity within the diagnostic group. In the present study, information about subgroups of ADHD and symptom load was not included. Both Booth 9 and Rodriguez 10 found a difference between the DSM-IV defined diagnostic subgroups of ADHD on the network measures. On the other hand, Seidman 36 argues that there are more similarities between the subgroups of ADHD than dissimilarities when it comes to measures of cognitive functions. We have calculated the within response variability according to Conners 32. According to Russell et al. 41, more extended calculations may give more adequate measures of variability and should be considered in further studies.

The main strength of the study was the case-control selected sample of children with ADHD, and that the results probably are less biased by co-morbid problems than clinical studies. However, no child was excluded due to a low FSIQ score, although some of the BCS participants with very low FSIQ score were excluded because they were unable to perform the ANT. The high correlations between the FSIQ score and error and variability measures in the ADHD group indicate that by excluding children with low total IQ scores, one may have excluded a specific group of ADHD children 39.

Although there have been several studies of the neuropsychological characteristics and the neural basis of ADHD, the deficits of attention in children with this behavioral diagnosis are still poorly understood. To conduct studies of this complex issue, appropriate neurocognitive models that operationalize different aspects of the attention system are necessary. The attention network theory provides one such model and can be used both in group studies and in the clinical evaluation of individual children. In a neuropsychological examination, the range of variables from the ANT may help to characterize the strengths and difficulties of a child. Studies of the attention networks in children with ADHD may contribute to a better understanding of the disorder and to the development of appropriatetraining and treatment methods 42.