Attention is a complex cognitive function involving different processes like selectively attending to specific stimuli, focusing for prolonged periods, or regulating and monitoring of actions (Anderson, 2002). It is a basic function that precedes higher-level cognitive abilities (e.g., executive functions or memory). Furthermore, attention underlies our awareness of the world and the voluntary regulation of our thoughts and feelings (Posner and Rothbart, 2007). The evaluation of attention in children is important because of its implications on learning, academic achievement and social functioning (Spira and Fischel, 2005).

Posner and Petersen were the first to propose a framework that divides attention into three functionally and anatomically separate networks: alerting, orienting and executive attention (Posner and Petersen, 1990; Fan et al., 2005). Alerting is the ability to produce and maintain optimal vigilance and performance during tasks (Petersen and Posner, 2012); the brain areas implicated are locus coeruleus, right frontal and parietal cortex. Orienting involves shifting attention to endogenous or exogenous cues (Corbetta et al., 1998; Posner et al., 2006); it involves parietal sites and frontal eye fields. Executive attention involves detecting and resolving conflict among responses, error detection and response inhibition (Bush et al., 2000); anterior cingulate cortex and prefrontal areas are involved in this network.

Following Posner and Petersen's model, Fan et al. developed the attention network test (ANT) to assess the three attentional networks (Fan et al., 2002). The ANT combines the flanker task (Eriksen and Eriksen, 1974), a widely used measure of executive control processes, with the use of cues entailing just alertness or alertness plus orienting information.

Rueda et al. adapted the adult ANT with a format and design more attractive for use with children from 4 years of age (Rueda et al., 2004). It is a rapid and easy-to-apply computerized test that has been validated as a tool for measuring attention processes in large-scale epidemiological studies (Forns et al., 2014) in which efficiency, precision and objectivity of data collection from using a computerized format is a major benefit.

Differences between age-group means in children, adolescents and adults have been encountered in attention processes using the ANT according to previous cross-sectional studies (Mezzacappa, 2004; Rueda et al., 2004; Gupta and Kar, 2009; Federico et al., 2017). Attention starts to develop early in infancy (e.g., exogenous alertness, orienting to external cues). Then, by the end of the first year of life, a more voluntarily controlled attention emerges. During childhood, the three networks continue developing and showing improvement in the endogenous sustained attention, the reorienting of attention, and the inhibitory control. Furthermore, the executive attention network, which depends on the maturation of the prefrontal cortex, shows a longer development into adolescence (Welsh and Pennington, 1988; Rueda et al., 2004, 2015; Konrad et al., 2005; Romine and Reynolds, 2005; Amso and Johnson, 2006). This is in relation to the many neurophysiological changes that occur in the brain (proliferation, migration, differentiation, synaptogenesis, myelination, and apoptosis) from the embryonic period through adolescence (Rice and Barone, 2000). In addition, there are genetic and environmental contributions to these changes on the developing brain, and therefore the attention circuits, during this period (Amso and Scerif, 2015). Gender can also impact attentional performance and although very few studies have focused on gender differences, girls seem to be more-advantaged (Pascualvaca et al., 1997; Klenberg et al., 2001). Furthermore, abnormal attention development is a symptom of attention deficit hyperactivity disorder (ADHD) in relation to structural and functional brain pathological changes (Biederman and Faraone, 2005). Cross-sectional studies that evaluated the three attentional systems in children with ADHD using the ANT, encountered deficits in the alerting and executive attention networks (Johnson et al., 2008; Gupta and Kar, 2009; Mullane et al., 2011; Casagrande et al., 2012). Specifically, children with ADHD showed lower levels of alertness, consistent with previous theories emphasizing difficulties in arousal regulation in relation to deficits in sustained attention (Russell et al., 2006; Johnson et al., 2007). Furthermore, children with ADHD were more vulnerable to the interference demands, and indeed, response inhibition has been described as the primary deficit in ADHD (Barkley, 1997). To our knowledge, there are no previous longitudinal studies quantifying intra-individual changes over time using the ANT. Prospective longitudinal studies on cognitive function are crucial for understanding typical developmental trajectories, confirming theories of neurodevelopment, and detecting the impact of environmental and social risk factors on cognition in epidemiological studies (Amso and Scerif, 2015; Sunyer et al., 2015; López-Vicente et al., 2016). In contrast to cross-sectional designs, longitudinal studies control for intra-individual differences across time as well as cohort effects and age-related differences in rate of change (Sliwinski and Buschke, 1999; Kraemer et al., 2000). In counterpoint, the magnitude of the practice effects due to repeated testing is a main concern (Dikmen et al., 1999). Ishigami et al. explored the stability, isolability, robustness and reliability of the ANT when administered repeatedly to older adults and patients with multiple sclerosis. Despite some practice effects in alerting and executive control, the network scores were robust against practice (Ishigami and Klein, 2011; Ishigami et al., 2013).

We therefore studied the performance of the ANT in a large longitudinal cohort of 7–11 year old children to detect patterns in attention growth. We focused on this age range for several reasons. First, and as mentioned above, some aspects of attention continue to improve throughout middle and late childhood. Second, the feasibility for the administration of the test to the children in their normal class groups (e.g., understanding instructions, ability to work almost autonomously, assuring a prolonged isolation during the execution of the test). Finally, primary school children from 7 years of age were selected in order to increase accuracy of identification of ADHD symptomatology (Applegate et al., 1997). For the current study, in addition to the original three attention networks based on the Posner and Petersen (1990) model, we also analyzed three other measures based on the Mirsky model of attention (Mirsky et al., 1991) and specifically, the Focus component which is related to attentiveness (Egeland and Kovalik-Gran, 2010). For the calculation of these scores we followed the formulas of the Continuous Performance Test (CPT), a widely used computerized measure of different aspects of attention (Conners and Staff, 2000; Oberlin et al., 2005; Adólfsdóttir et al., 2008). We particularly aimed to study the trajectories of attention in a large longitudinal cohort of children using the ANT in a repeated measures design. Furthermore, we examined potential differences between boys and girls, and the role of ADHD symptoms in attention development. Based on the literature, we expect that: (1) the oldest children will show an improvement in alerting; (2) children will show a continued development in executive attention; (3) girls will show an advantage over boys and (4) children with ADHD symptoms will show a delayed developmental pattern in alerting, executive attention and measures related to attentiveness.

Children were on average 8.6 years old at baseline and 49.6% were girls. Maternal education level was high (58.9% of mothers had a university degree). According to the questionnaires rated by the teachers 10.5% of children presented ADHD symptoms, being the inattentive subtype the most prevalent across age groups (Table 2).

In the multilevel mixed-effects linear regression models we found improvements related to age for all the outcomes, except orienting. Gender interacted with age in executive attention (p for interaction < 0.001) and HRT (p for interaction < 0.001). Teacher-rated ADHD symptoms interacted with age in executive attention (p for interaction = 0.027), HRT (p for interaction = 0.014) and variability (p for interaction = 0.009) (Table 3).

The inclusion of maternal education and socioeconomic status in the models did not change the results (Table S2). Furthermore, the interactions between age and maternal education (p for interaction = 0.263) and socioeconomic status (p for interaction = 0.093) were unrelated to executive attention development (Table S3). Figures 1–6 represent age-associated changes in the ANT networks scores and measures related to attentiveness.

Regarding the network scores, for alerting we detected a quadratic curve indicating a further growth for the oldest ages (Figure 1, Table 3). Different age-related patterns by gender or ADHD symptoms were not found. We obtained no significant age effect for orienting (Figure 1, Table 3). For executive attention, girls performed better at young ages. However, males showed a faster cognitive growth and reached girls performance at around 10 years of age (Figure 1). Children with teacher-rated ADHD symptoms, specifically the inattentive and combined subtypes, performed worse than children with no ADHD symptoms until age 9 in executive attention. No age-related changes were encountered in children with ADHD hyperactive-impulsive symptoms (Figure 2, Table 3).

Regarding measures related to attentiveness, for HRT quadratic curves were found indicating more pronounced increases in performance in the younger groups (Table 3). Although boys were significantly quicker, girls showed a more pronounced improvement during the age range studied and reached the levels of boys at age 11 (Figure 3). Children with ADHD symptoms showed quicker HRT across the entire age range but without reaching the levels of children with no ADHD symptoms (Figure 4, Table 3). For HRT-SE, we found a quadratic curve, indicating a greater improvement in the younger groups that stabilized at around 10 years of age (Figure 5). Different age-related patterns by gender or ADHD symptoms were not detected. Finally, for variability different age-related patterns by gender were not detected. Children with ADHD hyperactive-impulsive and combined symptoms showed higher variability than those without ADHD symptoms across all age groups (Table 3, Figure 6). No age-related changes were encountered in children with ADHD inattention symptoms (Table 3).

The present study explored, for the first time, the attention development using the ANT in a large sample (n = 2,835) of primary-school children in a prospective longitudinal study using repeated measures. The pathways of the three attention networks and measures related to attentiveness showed different developmental trajectories. Specifically, the attention development in the age range studied was more pronounced in executive attention, HRT, HRT-SE, and variability, particularly from 7 to 9 years of age. We detected gender differences in executive attention and HRT up to age 10. Children with teacher-rated ADHD symptoms showed a developmental delay in executive attention, HRT and variability.

The trajectories of attention growth observed confirm previous cross-sectional studies that used the ANT, which described an ongoing development of some aspects of attention during early and late childhood (i.e., ages 5–12) (Mezzacappa, 2004; Rueda et al., 2004; Mullane et al., 2016; Federico et al., 2017). This is consistent with neuroimaging studies reporting continued myelinization of the neural circuitry involved in attention processes until adolescence (Hudspeth and Pribram, 1992; Konrad et al., 2005). Furthermore, larger improvements in the younger groups correlate with the fourth rapid brain growth stage (i.e., ages between 6 and 8 years; Epstein, 1986).

Regarding the developmental trajectories of the attention networks, for alerting we found a slow progression toward the end of the investigated age range (Ridderinkhof et al., 1997; Gupta and Kar, 2009; Mullane et al., 2016; Federico et al., 2017). Alertness starts to develop in the first few postnatal weeks and by the third month babies are able to maintain an alert state in relation to the external sensory stimulation (Rueda et al., 2015). Despite the presence in infancy, the alerting network undergoes a significant improvement during late childhood in relation to continued development of frontal regions during this period (Rueda et al., 2004). The orienting network starts developing very early in infancy; newborns show head an eye movements toward a peripheral cue and from age 4 months there is great stability (Clohessy et al., 2001; Rueda et al., 2004; Gupta and Kar, 2009). For orienting we obtained no significant age changes in children from 7 to 11 years old, consistent with the formerly described early maturation of this network. The executive attention network follows a more prolonged development during childhood and into early adolescence (Ridderinkhof et al., 1997; Rueda et al., 2004; van Meel et al., 2012; Loher and Roebers, 2013; Mullane et al., 2016). Executive attention involves detecting and resolving conflict among responses, error detection and response inhibition. Its development is related to the maturation of the anterior cingulate cortex and lateral prefrontal cortex (Bush et al., 2000) which are not fully matured until adolescence (Romine and Reynolds, 2005). Furthermore, we encountered that the developmental course of the executive attention network differed by gender. Girls were superior compared to boys in the younger groups, probably reflecting variations in the maturation rate between males and females up to ages of 9 and 11 years (Pascualvaca et al., 1997). A more rapid biological, cognitive, and social-emotional development of girls is in fact well-known (Keenan and Shaw, 1997; Cahill, 2006). Finally, in our study the executive attention trajectory was unrelated to maternal education or socioeconomic status, in contrast to previous literature documenting an association between different indicators of socioeconomic status and executive function development (Hackman and Farah, 2009).

Regarding the measures related to attentiveness, significant improvements in reaction time (HRT), response speed consistency (HRT-SE) and intra-individual variability were found over the age range studied, and particularly between ages 7 and 9 years (Levy, 1980; Rebok et al., 1997). Boys had faster reaction times compared to girls (Pascualvaca et al., 1997; Klenberg et al., 2001), although the progression was further for females. The lack of significant differences in the trajectories of response speed consistency and intra-individual variability by gender, suggests similar age trends for males and females for these aspects of attention (Pascualvaca et al., 1997).

The developmental trajectories for executive attention, response speed and intra-individual variability were significantly different in children with teacher-rated ADHD symptoms compared to typically developing children. Neuroimaging studies have reported a delay of 2–3 years in brain maturation in ADHD, and not a complete deviation from typical development (Castellanos et al., 2002; Shaw et al., 2007). Difficulties in executive attention and alerting networks have been described as deficits that underlie ADHD (Berger and Posner, 2000). As expected, typically developing children outperformed children with ADHD symptoms in the executive attention network which involves response inhibition, indeed the essential impairment in this disorder (Barkley, 1997). However, we encountered no differences in the developmental trajectory of alerting in the age range studied between the two groups. The late further development of the alerting network (e.g., from 10 years of age) (Ridderinkhof et al., 1997; Rueda et al., 2004) may partly explain these results. Furthermore, in the published literature there are also discrepancies related to a weaker alertness in ADHD (Booth et al., 2007; Adólfsdóttir et al., 2008; Johnson et al., 2008; Gupta and Kar, 2009; Mullane et al., 2011; Casagrande et al., 2012). Regarding reaction time in responding to a target, slower response speed has been reported previously for children with ADHD (Gupta and Kar, 2009; Mullane et al., 2011), and that is in fact what we observed in our study. Children with ADHD symptoms had a developmental trajectory for response speed that paralleled the growth curve for typically developing children but on a lower track. Finally, in relation to intra-individual variability, we encountered more variability in response speed consistency in children with ADHD symptoms compared to children without ADHD symptoms. The moment-to-moment fluctuations in attention is in fact the most remarkable symptom of ADHD (Castellanos and Tannock, 2002). Indeed, intra-individual response variability measures, defined as short-term fluctuations in performance of an individual over a time-scales of seconds, have been described as the best to discriminate between ADHD and control groups (Russell et al., 2006). Finally, in relation to ADHD subtypes, no clear differential patterns of attention trajectories were observed despite the well-described clinical differences between them (Cantwell and Baker, 1992).

This study has some strengths and limitations that need to be considered. The present findings are dependent on the version of the ANT used, the age range of the participants and the teacher-rated ADHD symptoms—not a medical diagnosis of ADHD. However, we found that the child ANT (Rueda et al., 2004) was able to detect attention trajectories in childhood, including gender differences and ADHD symptoms. In relation to the age effects, the late development of the alerting network or the earlier executive attention development in females, for instance, still needs further investigation. Furthermore, the normalization of the performance in executive attention at around 10 years of age in children with teacher-rated ADHD symptoms needs future confirmation. Further studies using the original ANT (Fan et al., 2002) are warranted in order to discard any ceiling effects of the child ANT in late childhood. Finally, the cognitive change observed with repeated test application over a 1-year period may still include some practice effects. However, in previous studies, despite short test-retest intervals (i.e., hours or days) in attentional function in children or in adults the magnitude of the practice effects was moderate to small and the ANT networks showed robustness against practice (Mollica et al., 2005; Ishigami and Klein, 2011). Furthermore, the exact age in each assessment and the time intervals between the neuropsychological testing varied among the children and that contributed to minimizing practice effects. Strengths of this population-based study include the large sample size and the longitudinal design for the detection of cognitive change. Repeated measurements within-participants provided the prospective data required to define developmental trajectories. In addition to the attention networks of the ANT, three measures related to attentiveness were also calculated and allowed us to explore the growth patterns of more aspects of attention. Furthermore, we analyzed the role of gender and ADHD symptoms in attention development. Finally, the study of the attention development in grade-schoolers can be used to identify students who deviate from normality and the identification of possible environmental or social risk factors. This would contribute to the implementation of school accommodations or other interventions needed for these children. The implications of attentional deficits in the school setting are non-negligible since they are associated with behavioral difficulties and poor academic achievement (e.g., lack of self-control, failure to complete tasks, and commission of procedural mistakes) (Pascualvaca et al., 1997; Anderson, 2002).

In summary, we observed an ongoing development of some aspects of attention in primary school aged children. Nevertheless, the developmental changes were more evident in executive attention and measures related to attentiveness, and in the younger groups. Furthermore, girls were more advantaged at younger ages and children with teacher-rated ADHD symptoms showed a delayed development in some attention processes. Our findings support the ANT for assessing attention in children in large epidemiological studies.

ES, JF, and JS conceptualized and designed the study and drafted the initial manuscript. RG, ME, and XB supported and supervised the statistical analyses and revised the manuscript. MA coordinated and supervised data collections and critically reviewed the manuscript. ML, AL, JJ, and AC supervised the interpretation of the results and critically reviewed the manuscript.

The research leading to these results received funding from the European Research Council under the ERC Grant Agreement number 268479 – the BREATHE project. JJ holds a Miguel Servet contract (MS14/00108) awarded by the Spanish Institute of Health Carlos III (Ministry of Economy and Competitiveness).