Abstract
In children, the ability to listen to relevant auditory information and suppress distracting information is a foundational skill for learning and educational achievement. Distractibility is supported by multiple cognitive components (voluntary attention orienting, sustained attention, distraction, phasic arousal, as well as impulsivity and motor control) that might get mature at different ages. Here we used the Competitive Attention Test (CAT) to measure these components in 71 4- and 5-year-old children. The goal of this study was to characterize the changes in efficiency of attention during the preschool period, and to explore differences in distractibility in preschool children that could be related to the socio-economic status (SES) background of the child’s family. We found that, sustained attention improves from age 4 to 5, while voluntary attention orienting remains stable. Moreover, irrespective of the age, task-irrelevant sounds induced distraction and phasic arousal, as well as increased impulsivity. Children from lower SES backgrounds showed reduced sustained attention abilities, in particular at 4-year-old, and higher impulsivity compared to their peers from higher SES backgrounds. The present findings suggest that multiple distractibility develops during the preschool age and is likely to vary depending on the SES background of a child’s family.
Introduction
In children, the ability to listen to relevant auditory information and suppress distracting information is a foundational skill for learning and educational achievement (Stevens & Bavelier, 2012). For example, in a typical classroom, a child needs to listen to a teacher’s instructions or focus on a task when there are distractors in the environment, such as other children talking. The common term “paying attention” implies that attention is a unitary phenomenon, but it is in fact a multifaceted construct supported by multiple brain networks that undergo significant and differential development in childhood (Posner et al., 2014; Wetzel & Schröger, 2014). The present study focuses on the different facets of distractibility by examining individual differences in behavioral performance during an attention task in 4- and 5-year-old children and interactions with familial socioeconomic status (SES).
Distractibility
The term ‘distractibility’ has been used to describe an attention state that determines the propensity to have attention captured by irrelevant information and to react behaviorally to this information (Bidet-Caulet et al., 2015; ElShafei et al., 2019, 2020; Hoyer et al., 2021; Masson & Bidet-Caulet, 2019). Distractibility is a result of the balance between voluntary and involuntary attention processes.
Voluntary attention is goal-directed and can enhance the processing of relevant features (Corbetta & Shulman, 2002), locations (Posner, 1980), or modalities (Karns et al. 2009). Two forms of voluntary attention have been shown to directly contribute to distractibility: voluntary orienting and sustained attention. First, voluntary orienting allows one to selectively focus on a location and engenders perceptual enhancement of upcoming targets occurring at this location (Petersen & Posner, 2012; Posner, 1980; Yantis & Jonides, 1990). Behaviorally, voluntary orienting can be measured by contrasting reaction time (RT) to targets preceded by uninformative and informative cues (Petersen & Posner, 2012; Posner, 1980). Second, sustained attention is the ability to maintain attentional focus over time on a given task, and relies strongly on tonic arousal, also called vigilance (Betts et al., 2006; Oken et al., 2006; Parasuraman et al., 1989). Sustained attention can vary over multiple time scales, from rapid (across milliseconds) to slow (e.g., over a day), with attentional fluctuations over time (for review, see Esterman & Rothlein, 2019). Thus, voluntary orienting and sustained attention allow enhanced processing of ongoing task-related relevant stimuli and the effective maintenance of this enhancement over time. Overall, these two forms of voluntary attention allow one to focus.
In contrast to voluntary attention, involuntary attention can be oriented toward unexpected salient stimuli, leading to distraction when the stimulus is irrelevant in the task context (Bidet-Caulet et al., 2015; Escera, Alho, et al., 2000; Näätänen, 1992). ‘Distraction’, distinct from distractibility, describes the deleterious impact of a distractor on ongoing cognitive and behavioral performance. Indeed, when unexpected irrelevant salient stimuli occurs in the surroundings, the reactive allocation of attention and resources toward the irrelevant event is followed by a reallocation of resources toward the task: RT and accuracy can thus, respectively, increase and decrease as a result of distraction (Bidet-Caulet et al., 2015; Escera, Alho, et al., 2000; Näätänen, 1992). In this respect, distractors occurring while one is processing relevant information usually disturb the processing of this information and have a deleterious impact on behavioral performance.
Counter-intuitively, a distractor that occurs suddenly in the environment can also have a facilitation effect on RT. Since attention and arousal (i.e., alertness induced by the noradrenergic system) networks are interconnected in the brain (Aston-Jones & Cohen, 2005), this facilitation effect is induced by a transient increase in noradrenaline in response to the occurrence of a distractor and leads one to temporarily react more quickly to any upcoming stimuli (Bidet-Caulet et al., 2015; Masson & Bidet-Caulet, 2019; Max et al., 2015; Näätänen, 1992; Wetzel et al., 2012, SanMiguel et al., 2010). Interestingly, an increased false alarm rate is typically observed in impulsive people and might ensue from an enhanced phasic arousal effect (Eysenck & Eysenck, 1985; Houston & Stanford, 2001; Zhang et al., 2015) and reduced motor control (Booth et al., 2003; Kanaka et al., 2008; van den Wildenberg & Crone, 2005; Wright et al., 2003).
As multiple cognitive components (voluntary attention orienting, sustained attention, distraction, phasic arousal, as well as impulsivity and motor control) contribute to distractibility, it is fair to assume that distractibility is not a unitary function, but rather a cognitive state.
Distractibility in preschool aged children
Evidence suggests that children are more distractible than adults (Hoyer et al., 2021; Wetzel et al., 2006; Wetzel et al., 2016; Wetzel & Schröger, 2014). However, there remains a gap in our understanding of the origins of immature distractibility during early childhood. Below we discuss results from previous studies that separately investigated the different components of distractibility in preschool children.
Voluntary attention orienting and sustained attention
Studies using tasks with endogenous cues that are either informative or uninformative have yielded conflicting results on the efficacy of voluntary attention in preschool children: some studies suggest that the capacity to voluntarily orient attention is mature before the age of four (Colombo, 2001; Johnson et al., 1991; Ross-Sheehy et al., 2015) while others show that the benefit in RT to targets following informative cues increases from four to six years of age (Hrabok et al., 2007; Mezzacappa, 2004; Posner et al., 2014; Rothbart et al., 2011). In preschool-aged children, sustained attention has been investigated using various child-friendly tasks. Taken together, results suggest that sustained attention emerges within the first years of life (Kanaka et al., 2008; Reynolds & Romano, 2016) and these skills have been found to progressively improve between three and five years of age (Graziano et al., 2011; Reynolds & Romano, 2016; Richards & Casey, 1991; Ruff & Capozzoli, 2003). Behaviorally, these changes in sustained attention are reflected in a gradual reduction of false alarm and missed response rates, as well as in RT variability (Mahone et al., 2001). To date, only few studies have investigated the impact of isolated distracting events on sustained attention ability; the small number of available studies suggests that efficient sustained attention abilities in children may shield against the deleterious effect of distraction (Oakes & Tellinghuisen, 1994; Slobodin et al., 2015).
Distraction
To our knowledge, only one study has investigated the distraction effect on behavioral performance in preschool-aged children using an audio-visual oddball task, and results suggest that the deleterious effect of distraction progressively decreases during the preschool period, with a critical developmental step between ages four and five (Wetzel et al., 2018). However, it remains to be determined whether this difference between four and five-year-old children can also be observed when distractors are not deviant or novel, as is the case in oddball paradigms.
Phasic arousal
To date, no study has investigated developmental changes during the preschool years in the beneficial behavioral effect which can be induced by distracting sounds. Phasic arousal, which is hypothesized to be responsible for increased speed in RT to targets preceded by a distractor, has been studied mostly using physiological measurements in the first year of life (Hernes et al., 2002). For example, an increased pupil dilation response to rare, unexpected and complex sounds has been observed in infants from 13 to 16 months of age compared to adults (Wetzel et al., 2015; Max et al., 2015), suggesting that phasic arousal is increased in infants. Behaviorally, phasic arousal has also been found to be reduced in children aged 6 to 13-years-old compared to adults (Hoyer et al., 2021). To that extent, it is possible that phasic arousal undergoes developmental changes during the preschool age but, to our knowledge, no behavioral studies have yet attempted to answer this question.
Motor control and impulsivity
Sensory and motor areas of the brain are typically the first to mature (Casey et al., 2005) with structural changes in the sensorimotor cortex which reaches an adult-like functioning between late infancy and the preschool period. However, motor control is dependent on many interconnections between cortical and sub-cortical regions of the brain (e.g., the prefrontal and lateral temporal cortices) which do not appear to reach a complete level of maturity until young adulthood (Gogtay et al., 2004). To that extent, motor control and impulsivity are likely to influence the motor response, which can in turn lead young children to react randomly.
Studies of the development of the different components of distractibility during the preschool years suggest that sustained attention seems to increase during the preschool age but report contradictory results regarding voluntary orienting. Furthermore, the few studies that have examined distraction, phasic arousal, as well as motor control and impulsivity are too few to draw strong conclusions about the behavioral changes associated with the development of these mechanisms during the preschool age. Thus, to better understand how distractibility develops during the preschool age, its components need to be simultaneously assessed in children.
The Competitive Attention Test
The Competitive Attention Test (CAT) is a detection paradigm designed to assess the behavioral and brain correlates of distractibility (Bidet-Caulet et al., 2015). The advantage of this paradigm is that it combines the Posner attention-network task and oddball paradigm principles to provide simultaneous and dissociated measures of voluntary attention, distraction, phasic arousal, motor control and impulsivity in children and adults (Hoyer et al., 2021). To assess voluntary attention orienting, the CAT includes informative and uninformative visual cues toward the spatial location of a forthcoming auditory target. To measure distraction, the CAT includes trials with a task-irrelevant distracting sound preceding the target at several different delays. This change in distractor timing onset allows dissociation of the behavioral effects of facilitating phasic arousal (the difference between median RT in no-distractor and early-distractor conditions) and detrimental distraction (the difference between median RT in late- and early-distractor conditions). Based on previous results (Bidet-Caulet et al., 2015; Masson & Bidet-Caulet, 2019), these differences can be interpreted respectively as good approximations of the facilitation and detrimental distraction effects triggered by so-called distracting sounds (see Fig. S1 Supplemental Material 1). Results from studies in adults typically show that the voluntary orienting effect is manifested by a RT reduction in informative compared to uninformative trials; the distraction effect is manifested by increased RT in late compared to early-distractor condition; and, finally, the phasic arousal effect is indexed by a RT reduction in the early distractor condition compared to no distractor condition.
In order to study the development of these facets of attention with more precision, the CAT was recently adapted for young children and used in a study of a large cohort of participants aged 6 to 25 years (Hoyer et al., 2021). In this study, the behavioral measurement parameters of the CAT were refined compared to those previously used in adults (see Table 2 and Method for a detailed description). Results showed that voluntary orienting is functional at 6 years of age, while the ability to sustain attention gradually develops from 8 to 12 years of age; interestingly, distraction is manifested as omissions (i.e., missed targets) of relevant stimuli in 6-7-year-olds and as impulsivity in 11-12-year-olds. However, the RT distraction measure was not modulated by age, while the RT facilitation effect linked to phasic arousal decreased from 6 to 12 years of age. This attentional imbalance, resulting in increased distractibility in children, may then be more related to reduced sustained attention capacities, enhanced distraction, and increased arousal effects in childhood (6-8-year-olds), but to increased impulsivity in older children and adolescents (10-17-year-olds). Importantly, some measurements (e.g., missed responses, RT variability) show higher variability in younger children than in adults. A part of this variability may be related to environmental factors such as socioeconomic status.
Socioeconomic Status
Socio-economic status (SES) background is a proxy variable for variability in the early environment, typically assessed by a combination of parental education, occupational status, and/or household income. Disparities as a function of SES have been documented in a wide range of neurocognitive outcomes, and one of the neurocognitive systems most consistently associated with SES is self-regulation, including specific aspects of attention (Hackman et al., 2010; Pakulak et al., 2018; Ursache & Noble, 2016). Altered functioning of executive attention has been linked to reduced voluntary attention abilities (Diamond, 2013; Posner, 1980, 2012), and has been consistently reported in children and adolescents from lower SES (LSES) backgrounds (Farah, 2017; Farah et al., 2006; Noble et al., 2013). Event-related brain potential studies have also found differences in selective attention as a function of SES in adolescents (D’Angiulli et al., 2008) and in preschool-aged children (Giuliano et al., 2018; Hampton Wray et al., 2017; Stevens et al., 2009). These studies have found that children from LSES backgrounds show an increased brain response to stimuli they are instructed to ignore, relative to children from higher SES (HSES) backgrounds. Moreover, the relationship between SES and brain response to distracting sounds is mediated by the sympathetic nervous system activity: the larger the sympathetic activity the better distractor suppression, suggesting a biological cost, for children from LSES backgrounds, to achieve similar cognitive performance than HSES children (Giuliano et al., 2018).
Increasingly, evidence suggests that the effects of socioeconomic inequality begin early and persist during development. For example, behavioral signs of impulsivity in the first year of life persist into first grade only in children from LSES backgrounds (Meade, 1981); in this population, impulsivity in adolescence has also been linked to increased risk taking in early adulthood (Auger et al., 2010). Importantly, this pattern of results may represent a functional adaptation to environmental demands (e.g., increased sensitivity to potential threats) that may have deleterious effects in other environments (e.g., a classroom). In addition, these systems are amenable to evidence-based interventions targeting self-regulation and attention in preschool-aged children (Neville et al., 2013). In order to inform the refinement of such approaches and the development of novel approaches, it is crucial to improve our understanding of specific aspects of attention that contribute to distractibility in preschool-aged children from different SES backgrounds.
The Current Study
As the development of distractibility during the preschool years is still an open question, the goals of the present study are (i) to determine whether the CAT is an appropriate task for children from 3 to 5 years of age, (ii) to characterize the changes in efficiency of multiple aspects of attention during the preschool period, and (iii) to explore differences in distractibility in preschool children that could be related to the SES background of the child’s family. Specifically, we examine the development of voluntary orienting and sustained attention, distraction, phasic arousal, impulsivity and motor control, as well as potential differences in this development as a function of SES. Based on the literature, we hypothesized that distraction would decrease during early childhood, accompanied by an improvement in motor control and by a maturation of voluntary orienting and sustained attention, with a different profile in LSES children.
Methods
Participants
Preschool children aged three (3YO), four (4YO) and five (5YO) years participated at their usual childcare site. All children had normal or corrected-to-normal vision. Parents were informed and provided signed informed consent on-site, and children provided informed verbal assent prior to participation. With their permission, parents at each site were entered to win a $50 gift card in a lottery regardless of whether they opted for their child to participate in the study. Children chose a small educational prize for participating in the study. Recruitment and study procedures were approved by the University of [blinded] Research Compliance Services (protocol 03172011.022 entitled “[blinded]”) and by participating preschools.
Socioeconomic status was operationalized based on demographic characteristics of participating preschool sites, which were selected specifically for this purpose. While individual-level assessment of SES was beyond the scope of this study, parental income and education are considered a valuable proxy for the wider range of characteristics that tend to vary as a function of SES, even at an aggregate group level (Hackman et al., 2010; Ursache & Noble, 2016). Children were recruited from either preschools associated with a university campus (higher SES; HSES) or from Head Start preschool sites (lower SES; LSES). Note that, for readability, we will use ‘HSES and LSES children’ in the Method and Results sections of this manuscript to refer to ‘children from higher SES and lower SES backgrounds’ respectively.
Head Start preschools (LSES preschool sites) are part of a U.S. program serving families living at or below the poverty line, which is determined by the U.S. Federal Poverty Guidelines (https://aspe.hhs.gov/poverty-guidelines); e.g., for the year in which data were acquired the poverty line for a family of four was an annual salary of $25,100 (see Supplemental Material 2 for guidelines for different family sizes). University sites (HSES preschool sites) serve children whose parents are faculty, students, or staff at the university. According to the information provided by the university sites, among children from the HSES preschool sites, 60% had parents who were faculty members, 11% had parents who were students at the university and 29% had parents who were staff at the university (e.g., Office of Administration). Among them, less than 5% could benefit from the federal “reduced lunch program” which eligibility criteria are much higher than poverty guidelines, e.g., for the year in which data were acquired the poverty line for a family of four was an annual salary of $46,435 (see Supplemental Material 2 for guidelines for different family sizes). While this operationalization was necessarily broad and imprecise, it is likely that the two groups were qualitatively different in ways that reflect to some degree differences captured by more systematic assessments of SES.
Table 1 summarizes characteristics of the final sample. Although we tested 3YO children, experimenter field notes and visual inspection of the data indicated that the youngest children did not reliably complete the task, so they were excluded from further analysis (n=14). Data from an additional 21 children were excluded due to non-compliance with instructions, either because they abandoned the task or because they did not look at the screen during the task: specifically, 11 4YO (LSES = 8, and HSES = 3) and 10 4YO (LSES = 8, and HSES = 2) children were excluded. A total of 71 subjects (51% female, 4 to 5 years of age) were included in the analysis. Within SES groups, participants were matched in age and gender.
Characteristics of the population sample.
Behavioral measurements, parameters for calculations and measured construct.
Stimuli and task
Detailed task methods have been previously published (see Hoyer et al., 2021 and Table S2 Supplemental Material 3 for more details). In brief, children were instructed to keep their eyes fixated on a cross during the task intervals between trial events. In 50% of the trials, a visual cue (200-ms duration) was followed, after a 940-ms delay, by a 200-ms duration target sound (Fig. 1a). The directional cue was a dog facing left, right, or to the front. The target sound (a dog bark) was monaurally presented in headphones. For the other 50% of trials, a binaural distracting sound (300-ms duration) was played during the 940 ms delay period (Fig. 1b). The distracting sound could be played at three different times during the delay: 200 ms (Dis1), 400 ms (Dis2) and 600 ms (Dis3) after cue offset, distributed equiprobably. Target sounds played at 75 dBA and distracting sounds at 85 dBA.
Note. a) In uninformative trials, a facing-front dog was used as visual cue (200 ms duration), indicating that the target sound would be played in either the left or right ear. In informative trials, a dog visual cue facing left or right (200 ms duration) indicated in which ear (left or right, respectively) the target sound will be played (200 ms duration) after a delay (940 ms). If the participant gave a correct answer within the 3300 ms post target offset, a feedback (800 ms duration) was displayed. b) In trials with distractors the task was similar, but a binaural distracting sound (300 ms duration) - such as a phone ring - was played during the delay between cue and target. The distracting sound could equiprobably onset at three different times: 200 ms, 400 ms, or 600 ms after the cue offset.
Cue categories (informative, uninformative) and target categories (NoDis, Dis) were equiprobable for trials with and without distracting sounds. The directional cue was informative for 75% of the trials. Participants pressed a key as fast as possible when they heard the target sound. They were asked to focus their attention to the cued side. Feedback was given when participants detected the target within 3300 ms after onset (800-ms duration, 500 ms after the response), followed a fixation period (1700 ms to 1900 ms). If the participant did not respond in time, the fixation cross was displayed for an additional randomized delay (100 ms to 300 ms).
Procedure
Participants were tested in groups of two in a quiet room. The composition of each duo was determined by the classroom teachers as two children who were considered to have good interactions at school but did not regularly play together during playtimes. This was done to minimize both potential stress (children were all tested by two testers they only had met once during the presentation of the study to the class, so we reasoned that being with a classmate during the experiment would make the situation more comfortable for children) and distraction (children are often tempted to talk to close friends even when asked to not do so) during the experimental session. Prior to the task, children were shown a treasure map (see Fig. S3.1 Supplemental Material 3) for a game to perform three blocks of the task. In each block they would help a dog find his bone by pressing the button when they heard the dog bark and were told they should not press the button if they heard any other sounds. At the end of each block, children took a break to sing a nursery song with experimenters while pretending to row to the next island. If children chose not to complete all three blocks they still received a prize.
During the task, participants were seated in front of a laptop (approximately 50 cm from the screen) that presented pictures and sounds and recorded behavioral responses using Presentation software (Neurobehavioral Systems, Albany, CA, USA). Auditory stimuli were played in headphones. Participants performed three 4-minute blocks of 48 pseudo-randomized trials each. Verbal instructions accompanied visual illustrations of the task (see Fig. S2.2 Supplemental Material 3). Children pressed a button on the keyboard with their dominant hand (the hand they preferentially use to draw). Each child verbally confirmed to the experimenter that they were able to hear the dog bark during the task. An experimental session lasted around 30 minutes.
Measurement parameters
We used a custom MATLAB program to preprocess behavioral data. The shortest RT for a correct response (RT lower limit) was calculated for each age range (see Fig. S4 Supplemental Material 4). For each participant, the longest RT for a correct response (RT upper limit) was calculated from all RT > 0 ms using the Tukey method of leveraging the interquartile range. Correct response rate corresponds to the percentage of responses with a RT (relative to target onset) superior or equal to RT lower limit and inferior or equal to RT upper limit. Eight behavioral measures were extracted for each child (Tab. 2. see also Fig. S5 Supplemental Material 5).
Statistical analysis
Sample characteristics
Bayesian statistics were used to test the probabilistic certainty that the multivariate measures are related. Bayesian statistics were performed using JASP® software (JASP Team (2018), JASP (Version 0.9) [Computer software]).
To confirm that our sample population was similarly distributed across age ranges in block order, gender, and SES, we performed Bayesian contingency table tests. We employed Bayes Factor (BF10) as a relative measure of evidence in favor of the null model, a uniform distribution (BF between 0.33 and 1, weak evidence; 0.1 to 0.33, positive evidence; a BF 0.01 to 0.1 strong evidence; BF lower than 0.01, decisive evidence) and the strength of evidence against the null model (BF of 1 to 3, weak evidence; BF 3 to 10, positive evidence; BF 10 to 100, strong evidence; BF higher than 100, decisive evidence (Lee & Wagenmakers, 2013).
Behavioral data analysis
We expected large inter-individual variability in RT and response type rates as a function of condition, which limits the comparison of data between conditions and means that data cannot simply be pooled for analysis. Generalized Linear Mixed Models (GLMM) are preferred, as they allow for correction of systematic variability (Bates et al., 2015). The heterogeneity of performance between subjects and experimental conditions was considered by defining them as effects with random intercepts and slopes, thus instructing the model to correct for any systematic differences in variability between the subjects (between-individual variability) and condition (between-condition variability). We used the Akaike Information Criterion and the Bayesian Information Criterion as estimators of the quality of the statistical models generated (Matuschek et al., 2017).
To assess the impact of the manipulated task parameters (cue information and distractor type) and participant demographic characteristics (age and SES), on each type of behavioral measure (RT, RT SD, LateRep, MissRep, CueRep, DisRep, AntRep, RandRep), we analyzed the influence of four possible fixed effects (unless specified in Table 2): the between-subject factor AGE: 2 levels (4YO and 5YO); the between-subject factor SES: 2 levels (LSES and HSES); the within-subject factor CUE: 2 levels (CUE: informative vs. uninformative); the within-subject factor DISTRACTOR: 4 levels (NoDis, Dis1, Dis2 and Dis3), except for DisRep: 3 levels (Dis1, Dis2 and Dis3). A summary of the data and factors used in statistical modeling can be found in Table 2. Note that for response types cumulating less than a mean of 10 observations across subjects (CueRep, DisRep and LateRep), we did not consider the within-subject factor DISTRACTOR in the analysis.
Frequentist models and statistics were performed in R® 3.4.1 using the lme4 (Bates et al., 2015) and car (Fox & Weisberg, 2018) packages. Because both fixed and random factors were taken into account in statistical modelling, we ran a type II analysis of variance. Wald chisquare tests were used for fixed effects in linear mixed-effects models (Fox & Weisberg, 2018). We only considered the explanatory variables. The fixed effect represents the mean effect across all subjects after correction for variability. We considered results of main analyses significant at p < .05. When we found a significant main effect or interaction, Post-hoc Honest Significant Difference (HSD) tests were systematically performed using the emmeans package (emmeans version 1.3.2). P-values were considered as significant at p < .05 and were adjusted for the number of comparisons performed. To determine to what extent observed significant effects would be replicable, we used the simr R package (Green & MacLeod, 2016) to perform retrospective power analysis of GLMM fixed effects. We simulated the models using the powerSim function (Monte Carlo simulation method) and calculated (over 1000 simulations) the probability that they would have detected the effects of interest. In the following, power for interactions is reported in % with confidence intervals (CI). The power represents the probability that one would observe the same significant effects if the study is replicated: this probability is considered adequate when up to 80% (although this threshold is still a matter of debate, see Field et al., 2007 for more details). In the Results section, we report the SEM as the estimator of the distribution dispersion of the measures of interest, when not specified.
To ensure that analyses were performed on a sufficient number of trials per condition, participants with fewer than 12 trials with positive RT in each of the distractor conditions (N = 6, all LSES) were excluded from the median RT analysis (resulting in a total average of trials with positive RT of 49 ± 1.8 in NoDis, 16 ± 0.7 in Dis1, 15 ± 0.6 in Dis2 and 16 ± 0.6 in Dis3 conditions across the overall sample). Revised sample sizes for median RT analysis are: 4-year-olds, n = 23 and 5-year-olds, n = 43. The percentage of missing data over the total sample of included subjects in analyses is shown in Table 3. Raw RT were log-transformed at the single trial scale in order for RT and RT SD analyses to be better fit to a linear model with Gaussian family distribution; response types were re-fit to a linear model with binomial distribution without transformation (see Table 3 and Supplemental Material 6 for additional details).
Main statistical analyses according to behavioral response types.
Results
Sample characteristics
We first tested whether the distribution of participants was similar between the levels of a given factor according to age. Using Bayesian contingency table tests, we found positive evidence for a similar distribution in block order (BF10 = 0.133) and gender (BF10 = 0.323) across age ranges. We observed weak evidence for a similar distribution in SES (BF10 = 0.746) across age, indicating that there were more LSES participants in the 5YO compared to 4YO group (4YO: 13 LSES and 13 HSES; 5YO: 30 LSES and 15 HSES). Second, we tested whether the distribution of participants was similar between the levels of the gender factor within the SES groups. We found no evidence for different distributions in gender in the LSES (BF10 = 0.395) and HSES (BF10 = 0.764) groups, indicating that there were as many boys than girls within the LSES and HSES groups.
Behavioral Data
For each type of behavioral measurement, we analyzed the influence of AGE, GENDER, SES, CUE, and DISTRACTOR factors (unless specified otherwise in the Table 3). In the following, when a factor is involved in a main effect and a higher order interaction, only the post-hoc analysis related to the interaction is described.
Reaction Times
As expected, a main effect of AGE (χ2 (1) = 6.45; p < .05; power = 100.0%, CI [99.5, 100.0]) on median RT indicated that 4YO children (985.0 ± 73.0 ms) were overall slower to respond than 5YO (768.6 ± 28.7 ms). Irrespective of age, children were faster when the cue was informative (833.6 ± 32.7 ms) rather than uninformative (872.6 ± 38.9 ms; main effect of CUE: (χ2 (1) = 4.32; p < .05; power = 100.0%, CI [99.6, 100.0]).
Consistent with our previous work in older children and adults, we observed a main effect of the DISTRACTOR (χ2 (3) = 180.4; p < .001; power = 100.0%, CI [99.6, 100.0]; see Fig. 2) on median RT. Post-hoc pairwise comparisons showed that distractors speeded the response to targets. RTs were slower in NoDis (898.5 ± 35.0 ms) than in Dis1 (766.1 ± 39.0 ms), Dis2 (801.5 ± 49.2 ms) and Dis3 (848.2 ± 42.6 ms) conditions. As shown in Figure 2, slower median RTs were observed in Dis3 compared to Dis1 and Dis2 conditions, whereas no difference was found between Dis1 and Dis2 conditions (p = .345).
Note. Mean of median reaction time as a function of the distractor condition [NoDis, Dis1, Dis2 and Dis3] (p < .05 *, p < .001 ***; Error bars represent 1 SEM).
Standard deviation of reaction times
Although we observed a main effect of AGE (χ2 (1) = 5.5; p < .05; power = 70.3%, CI [67.4, 73.1]) on response time variability (RT SD), this should be interpreted in light of a significant three-way AGE by BLOCK by SES interaction (χ2 (2) = 7.3; p < .05; power = 68.9%, CI [65.9, 71.7]); Fig. 3). Irrespective of SES, post-hoc analysis revealed that 4YO (634.8 ± 61.7 ms) had higher RT SD compared to their older peers (462.2 ± 35.5 ms; p <. 01). Only LSES 4YO showed higher RT SD during the third block (565.7 ± 62.6 ms) compared to the first block (840.1 ± 89.7 ms). Moreover, on the third block only, LSES 4YO showed higher RT SD than HSES 4YO (463.7 ± 44.2 ms).
Note. Mean of median reaction time as a function of the block (1st, 2nd and 3rd), by SES (HSES = higher SES, LSES = Lower SES) and age (4-year-olds on the left and 5-year-olds on the right). (p < .05 *, p < .01 **; Error bars represent 1 SEM).
Global accuracy
The proportion of the different types of behavioral responses according to age is shown in Fig. 4. The average correct response rate was 50.7 ± 2.2%. No main effect of AGE or SES, nor interaction with AGE or SES, was found for RandRep (total average: 6.0 ± 0.7%; see Fig. S7b Supplemental Material 7), DisRep (total average: 5.9 ± 0.9%; Fig. S7c Supplemental Material 7), AntRep (total average: 13.8 ± 1.1%; Fig. S7d Supplemental Materials 7) or LateRep (total average: 9.9 ± 0.6%; Fig. S8a supplemental material 8). Significant effects of AGE and SES on the other response types are detailed in the following sections.
Note. Response type proportions for a) LSES and b) HSES groups. (HSES = higher SES, LSES = Lower SES).
Erroneous responses to cue
The rate of cue responses (CueRep, 2.9 ± 0.4% on average; Fig. S7a Supplemental Material 7) was modulated by SES (χ2 (1) = 7.0; p < .01; power = 71.6%, CI [68.7, 74.4]). A CUE by SES interaction was also significant (χ2 (1) = 6.8; p < .01; power = 70.7%, CI [67.8, 73.5]); Fig. 5). Post-hoc analysis revealed that LSES children (3.7 ± 0.5%) made more CueRep than HSES children (1.6 ± 0.5%) irrespective of the validity content of the cue. Additionally, children from HSES backgrounds made more CueRep in the informative condition (2.2 ± 0.6%) compared to the uninformative one (0.9 ± 0.4%).
Note. Mean percentage of cue response as a function of cue type (informative and uninformative) and (HSES = higher SES, LSES = Lower SES). (p < .05 *, p < .001 ***; Error bars represent 1 SEM).
Erroneous responses in anticipation of target
The rate of anticipated responses (AntRep, 13.8 ± 1.1% on average; Fig. S7d Supplemental Material 7) was modulated by DISTRACTOR (χ2 (1) = 46.0; p < .001; power = 52.3%, CI [49.2, 55.4]). Children had higher proportions of AntRep in Dis1 (18.7 ± 1.8%) than in NoDis condition (8.9 ± 0.9%) regardless of age and SES.
Missed responses
The rate of missed responses (MissRep, 15.1 ± 1.5% on average) was modulated by AGE (χ2 (1) = 15.4; p < .001; power = 99.2%, CI [98.4, 99.7]): 4YO (21.4 ± 3.2%) made more MissRep than 5YO (1.3 ± 1.3%) (Fig. S8b Supplemental Material 8).
We also observed significant main effects of SES (χ2 (1) = 15.0; p < .001; power = 99.0%, CI [98.2, 99.5]), as well as a significant DISTRACTOR by SES interaction (χ2 (3) = 8.5; p < .05; power = 68.9%, CI [65.9, 71.8]; Fig. 6). According to HSD post-hoc comparisons, and irrespective of the distractor condition, participants from LSES backgrounds (9.6 ± 1.6%) made more MissRep than participants from HSES backgrounds (18.6 ± 1.2%). Additionally, only children from HSES backgrounds made more MissRep in the Dis3 condition (12.4 ± 1.7%) compared to the Dis1 condition (7.7 ± 1.5%).
Note. Mean percentage of missed response as a function of cue (informative and uninformative) and SES (HSES = higher SES, LSES = Lower SES). (p < .05 *, p < .01 **, p < .001 ***; Error bars represent
Discussion
We characterized different components of distractibility in young children to determine the extent to which SES impacts the development of these components in early childhood. Consistent with our previous work in older children and adults, we found in both 4- and 5-year-old children (i) a facilitation of RT after informative cues or early distractors (ii) a cost in RT after late distractors, suggesting that this paradigm is a valid measure of these constructs in younger children (see also Fig. S9 Supplemental Material 9 for a representation of the developmental trajectories of distractibility components from age 4 to 25). In addition, 4- and 5YO children showed high variability in RT, suggesting difficulties in sustaining attention. Finally, 5YO were faster to detect target sounds and missed fewer of them than their 4YO peers, whether or not relevant sounds were preceded by a distractor. Furthermore, 4YO children from LSES backgrounds had higher variability in response times, while this difference was not observed in 5YO. Both 4 and 5 YO LSES children made more erroneous responses to the cue and missed more targets than children from the HSES group, suggesting greater impulsivity and reduced sustained attention in the LSES group. Below, we discuss the implications of these findings, caveats to our conclusions, and future directions.
Difference in distractibility components between 4 and 5 years of age
Voluntary orienting and sustained attention
Both 4- and 5YO were as faster to respond after an informative rather than an uninformative cue, irrespective of the SES. This finding suggests similar attention orienting abilities at these ages, independent of SES. Voluntary attention orienting has previously been found either to be stable (Federico et al., 2017; Mullane et al., 2016; Ross-Sheehy et al., 2015; Rueda et al., 2004) or to increase (Commodari, 2016; Mezzacappa, 2004; Reis Lellis et al., 2013) in efficiency from 5 to 10 years of age. The present results provide additional support for early maturation of attentional orienting.
The number of missed responses in attentional tasks is a sensitive measure of sustained attention over long time periods (Kanaka et al., 2008), while the proportion of late responses is considered to be an index of short-term sustained attention (Petton et al., 2018). In the present study, and irrespective of the distractor condition, 4YO showed an enhanced missed response rate when compared with their 5YO peers suggesting reduced sustained attention abilities. However, no effect of age was observed for the late response type, suggesting that children’s ability to maintain attention over shorter periods of time is stable from 4 to 5 years of age. More missed responses in 4 than in 5 YO are then more likely to result from an overall lower vigilance level in the younger age group, though future research is necessary to confirm this hypothesis. Furthermore, 3YO children were excluded from analysis due to their random behavioral performance: an inability to adequately perform the task at 3 years of age could also be explained by reduced sustained attention ability, as previously observed in the literature (Mahone et al., 2001). Together, these results suggest that sustained attention may not be sufficiently developed at age 3 to perform this task and is developing between ages 4 and 5.
Distraction and facilitation effects triggered by distractors
In line with previous studies using the CAT in adults (Bidet-Caulet et al., 2015; ElShafei et al., 2019, 2020; Masson & Bidet-Caulet, 2019), we observed two distinct effects on RT triggered by the distracting sounds. First, distracting sounds played long before the target (Dis1 and Dis2) speeded RT compared to a condition without distractor (NoDis): this benefit in RT has been attributed to an increase in phasic arousal (Masson & Bidet-Caulet, 2019). Second, distracting sounds played just before the target (Dis3) slowed RT compared to conditions with a distractor played earlier (Dis1 and Dis2): this cost in RT is considered a good behavioral approximation of distraction (Bidet-Caulet et al., 2015; ElShafei et al., 2019; Hoyer et al., 2021; Masson & Bidet-Caulet, 2019). In addition, both phasic arousal and distraction effects were observed in 4- and 5-year-olds. Thus, this finding extends those previously obtained with school aged children (Hoyer et al., 2021) to preschool aged children.
To date, few studies have investigated distraction in preschool-aged children using active behavioral tasks. Recent results from oddball paradigms, however, suggest that the distraction effect decreases from age 4 to 5, 5 to 6 and 6 to 9-10 before reaching the adult level (Wetzel et al., 2018). On the contrary, in the present study, the RT distraction effect is found similar between 4 and 5-year-olds. This discrepancy may be explained by differences in protocols. Discrimination (Oddball task; Wetzel et al., 2018) is more demanding than detection (CAT): it could then be easier for 4YO children to deal with distracting information during detection tasks compared to discrimination tasks. Taken together, these results suggest improvement from 4- to 5-years-old in the ability to ignore distracting information. However, future studies using both the CAT and an oddball task in the same sample of children are necessary to improve our understanding of the development of distraction during the preschool age.
Motor control and impulsivity
In the CAT paradigm, cue, distractor and random responses are conceptualized as complementary measures of impulsivity and motor control. The present findings indicate that none of these measures were modulated by age, suggesting that impulsivity and motor control efficiency are not different between 4 and 5 years of age. Irrespective of age, participants made more anticipated responses to the target in presence of a distractor, showing that processes triggered by distractors contribute to increase impulsivity to subsequent target. However, the mean rate of anticipated responses was quite variable across participants, and analyses suggested that this effect may not be robust.
Differences in distractibility as a function of SES
Distraction
Distraction can result in enhanced RT, late response or even missed responses in trials with distracting sounds. No effect of SES was found either on the RT distraction effect, or on late responses. The HSES group showed increased missed response rates in trials with a short distractor-target interval (Dis3) compared to condition without distractors. A different pattern was found in the LSES group: the rate of missed responses was similar irrespective of the distractor condition. Power analyses highlighted the robustness of the main effect of SES on the missed response rate but revealed a moderate power for the distractor by SES interaction. Together, these results suggest that LSES children may be less attentive in environments with or without auditory distracting information because of reduced global sustained attention abilities, while HSES children may be more sensitive to distraction and would thus tend to miss more targets in distractor conditions. Therefore, the present study does not provide clear support for increased distractibility in children from LSES backgrounds.
This finding contradicts previous research seeking to characterize behavioral differences in attention during the preschool years and as a function of SES. Results from event-related potential (ERP) studies of auditory selective attention suggested a reduced ability to suppress distraction in LSES children (Stevens et al., 2009; Giuliano et al., 2018; Hampton Wray et al., 2017). This discrepancy may be explained, at least in part, by differences in the tasks used to assess attentional capacities. Studies suggesting a reduced ability to suppress distracting information in LSES children have used a dichotic listening task with embedded probes. Behaviorally, this paradigm solicits the child’s ability to inhibit the continuous stream of distraction from the unattended channel, but also unexpected and isolated probe sounds. Since these tasks are more demanding, a child may need to recruit the executive system, the development of which is delayed in children from LSES backgrounds (Lawson et al., 2013; Noble et al., 2012). Recent findings also suggest that for children from LSES backgrounds there may be a potential biological cost of achieving performance similar to their peers from HSES backgrounds (Giuliano et al., 2018). Future studies that combine different paradigms in the same sample of LSES and HSES children are necessary to further elucidate this pattern of results.
Sustained attention
Although behavioral distraction was similar for children from LSES and HSES backgrounds, we found evidence that both 4- and 5YO from LSES backgrounds had more difficulty in sustaining attention, as they missed more targets irrespective of the distractor condition. We also documented increased variability in RT across experimental blocks in 4YO children in the LSES group, such that RT SD almost doubled between the first and third blocks in this group. Power analysis for this last interaction effect indicated a 70% chance of replication. Previous studies using a child-friendly ERP paradigm assessing sustained selective auditory attention (i.e., the ability to focus on a relevant channel among others over time) have also found differences in brain activity as a function of SES in this age group (Stevens et al., 2009; Hampton-Wray et al., 2017, Giuliano et al., 2018). Importantly, consistent with the present results, one study shows that SES differences evident at age four fade by age five (Hampton-Wray et al., 2017). There is also evidence that preschool-aged children from LSES backgrounds show differential functional maturation of prefrontal cortices (Lawson et al., 2013; Noble et al., 2012). Taken together, these results suggest that preschool-aged children from LSES backgrounds may show an attentional imbalance characterized by reduced voluntary sustained attention that might be particularly pronounced at age four. This might represent an adaptation to certain environmental contexts in which regularly interrupting ongoing voluntary attention processes is necessary to maintain constant reactivity to possible changes in the environment.
Motor control and impulsivity
Children from LSES backgrounds made significantly more cue responses than their peers from HSES backgrounds, suggesting higher impulsivity in LSES children. HSES children showed less cue responses than LSES children, but more cue responses in informative than uninformative trials: this suggests that children from HSES backgrounds, were more likely to respond impulsively to relevant - but non-target - stimuli when they were highly predictive of the forthcoming target. However, the cue and SES type interaction effect should be considered with caution: first, cue responses were relatively scarce; and second, power analyses revealed only a moderate chance for this effect to be replicated in other studies. Despite these limitations, our findings suggest that impulsivity to relevant stimuli (responses to cues) would be increased in LSES children, but there is no difference in SES groups in impulsivity to irrelevant stimuli (responses to distractor).
Limitations of the present study
There are several limitations to the current study. First, the protocol used does not provide measures of stress, motivation, or tonic arousal: these factors are believed to influence attentional performance. Second, this cross-sectional study did not provide longitudinal data. Further studies are then needed to characterize individual variability in attention performance and to better identify predictors of attentional efficiency in children and examine developmental trajectories of efficiency in the same participants. Third, SES was operationalized by proxy using the school affiliation of the child, limiting the ability to characterize and control for variability within SES. The present results thus need to be replicated using a more rigorous, individual-level operationalization of SES that characterizes multiple dimensions of the construct (Hackman et al., 2010; Ursache & Noble, 2016), and with age groups more closely balanced as a function of SES categories (i.e., here Bayesian analyses indicated an imbalance in SES in the 5-year-old group). It is also important to note that SES is not the sole factor than can explain differences in attention during the preschool age: several other related variables, such as the time spent by parents playing with their child (e.g., Cheung et al., 2017) or the time spent by children participating to activities outside the class (e.g., Diamond et al., 2007) have been shown to have a positive impact on a child cognitive abilities. It will be fruitful for future studies attempting to replicate the present results to include such factors in their design and analyses. Finally, the power calculated for several effects described here can be considered as insufficient to draw robust conclusions regarding potential differences in attention as a function of SES. It will thus be important to assess distractibility in larger samples of children, and to employ different attention tasks to better characterize inter-individual variability in attention performance within SES groups (e.g., 16 LSES and 5 HSES children were excluded from analyses because the task was too difficult).
Conclusion and future directions
To our knowledge, this is the first study to examine differences in specific facets of distractibility during the preschool age. Results show that 4YO have reduced voluntary sustained attention ability compared to their 5YO peers as indexed by a greater number of missed responses. Sustaining attention over time was found to be more difficult for children coming from LSES backgrounds, but no evidence for enhanced distraction in this population was found. While this difference in sustained attention may be maladaptive in certain contexts like school, it is important to consider that regularly interrupting ongoing attentional processes may be adaptive by allowing a child to maintain more constant reactivity to possible changes in the environment. Elucidating differences in distractibility during the preschool age as a function of SES is relevant not only to our basic understanding of disparities related to SES, but also for the timing or nature of interventions targeting attention skills (Neville et al., 2013; Posner et al., 2015; Tang & Posner, 2009). Thus, this study showed that the CAT is a potentially powerful tool for assessing attention in preschool children whose sustained attention is sufficient to perform the entire task, and that the CAT can provide important information for further evidence-based approaches to develop and refine programs that seek to train attention processes in children.
References
