Title: Investigating the Development of Inhibitory Control in Early Childhood using the Tablet-based TapTap Task

The development of inhibitory control is an early construct of executive function (EF) and a key milestone of neurocognitive maturation. However, while its functional connectivity (fc) patterns have been well-defined in adults, its developmental trajectory in early childhood remains poorly understood. Using tablet-based EF tasks, we assessed inhibitory control in typically developing children between 2 to 6 years of age. Overall successful inhibition and improvement during competitive and non-competitive inhibition displayed converging patterns of fc. With increasing inhibitory control and task difficulty, additional frontal and parietal fc was recruited, and long-range connections established. Insights into early development and refinement of inhibitory control could provide essential information for the support in normal development as well as for the diagnosis and treatment of neurodevelopmental disorders with EF and inhibition difficulties, such as autism spectrum disorder. attention Results suggest that by 30 months, children may be in a translational period where they use both attentional systems to support self-regulation. Our findings contrasting children in this age range could reflect part of this early maturational process. Furthermore, our findings indicate a shift towards increased long-range fronto-parietal connections with age and experience when comparing the oldest with the youngest group. This could indicate a development towards adult-like connectivity patterns, as long-range connectivity and specialisation are direct contributors to adult’s superior inhibitory control skills further supporting the notion of our age-staged contrasts capturing an underlying maturational process.


Introduction
Early child development is characterized by behavioral responses to the environment around them. While this process starts in-utero, it progresses more rapidly postnatally when exposed to multiple stimuli (for example, Silbereis et al., 2016), mirrored by changes in functional development of the brain at rest (Bruchhage et al., 2020(Bruchhage et al., , 2021. Behavior describes an organism's external reaction or set of actions in response to such stimuli in its environment (for a review on behaviour: Levitis et al., 2009). As such, a behavioral circuit is understood to involve progression from perception of stimuli, and processing of relevance (useful to planned reaction) to action termination. The human brain's uniqueness lies in its ability to use information in decision-making processes ranging from the simplest, single-step decisions to more complex computations, all of which influence behavioral responses to stimuli (for a review, Ernst et al., 2005).
A child's demonstration of decision-making precedes existing measures of the skill-constructs collectively explained by Executive Function (EF) skills (for a review, Silbereis et al., 2016). These skills refer to a set of neurologically based processes considered essential for coordinating cognitive and metacognitive activities, and functional brain correlates at rest have been documented for EF (Reineberg et al., 2016). Existing research suggests that early EF skills such as inhibitory control develop rapidly in early childhood while higher-order abilities develop in older children and adolescents (Miyake et al., 2000;Davidson et al., 2006). Inhibitory control describes the ability to suppress automatic inappropriate responses (Bjorklund & Harnishfeger, 1995;Dempster, 1992). The development of inhibitory control is a key milestone and is considered a feature of frontal lobe maturation (Casey et al., 1997;Garavan et al., 1999). Inhibition influences the "Go/No-go" decision-making process that encompasses most of the decisions made during early childhood and serves as an early construct of EF that may be measured in a young child (Carlson & Wang, 2007). Inhibitory control involves two dissociable processes; a thresholded adjustment process involving a global response, called non-competitive inhibition; and a controlled selection process involving inhibition among coactive responses; called competitive inhibition (Erb et al., 2020). Inhibitory control is supported by a distributed neural network associated with EF processes, including regions in the frontal lobe encompassing frontal pole, inferior and middle frontal gyri, dorsal lateral prefrontal cortex, frontal eye fields, anterior cingulate cortex, insula, superior and inferior parietal lobules and the presupplementary motor area (Nomi & Uddin, 2015;Garavan et al., 1999;Dove et al., 2000;Chen et al., 2009;Schmitz et al., 2006). Inhibitory control abilities continue to mature from early childhood until into adolescence, getting more defined the older the child gets (Menon et al., 2001;Vara et al., 2014). While the development of EF skills and its neural correlates in adulthood is well established, brain networks supporting inhibitory control differ between children and adults (Fair et al., 2007;Church et al., 2017), and the neurological underpinnings of early EF constructs in typically developing children is less well defined.
Most studies of executive function in children are based on established proxies of EF such as parent-report measures; e.g., the Behavior Rating Inventory of Executive Functioning -Preschool version (BRIEF-P; Gioia et al., 2000) and show progressive gain of EF skills, often noted to occur in spurts of rapid development (Ferrier et al., 2014). BRIEF-P structural brain volume and functional connectivity correlations have been shown to develop from multi-regional to frontal lobe focused clusters from 3 to 6 years of age Lockridge et al., 2018), suggesting a key role of frontal regions in EF skill development and refinement. There is very limited task-based EF skill mapping in younger children as they grow towards later stages of early childhood. However, this information is necessary, both for assessing maturational trajectories in development of an essential metacognitive skill, and to study the factors that influence it.
This study aims to build upon existing research defining the developmental trajectory of neural correlates of EF in early childhood. Mapping the course of typical maturation from diffuse brain pathways to more defined specialized networks to enable successful inhibitory control and its subcomponents, will allow more objective study of early development of decision-making processes in the young child. For this, we used a tablet-based EF task assessing inhibitory control  in typically developing children between 19 and 71 months of age.

Participant Demographics
Data used in this study were drawn from the ongoing BAMBAM (Brown university Assessment of Myelination and Behavioral development Across Maturation) study of neurotypical brain and cognitive development. From the BAMBAM cohort, 38 (23 boys, 15 girls) typically developing children between 19 and 71 months of age were selected for inclusion in this study. General participant demographics are provided in Table 1. 6 With a 66% Caucasian and 26% non-white ratio (8% N/A), spanning families with income ranges from $10,000 -$29,999 up to $200,000 and above annual income, with a mean average of $70,000 to $89,999 real household income, our cohort is comparable to the Rhode Island population (71.4% white and 28.6% non-white, average income of $85,527 as of 2018; United States Census, https://www.census.gov/quickfacts/fact/table/RI,US/INC110218).
As a broad background, children in the BAMBAM cohort were born full-term (>37 weeks gestation) with height and weight average for gestational age, and from uncomplicated singleton pregnancies. Children with known major risk factors for developmental abnormalities at enrolment were excluded. These included: in utero alcohol, cigarette or illicit substance exposure; Preterm (<37wks gestation) birth; small for gestational age or less than 1500g; fetal ultrasound abnormalities; preeclampsia, high blood pressure during pregnancy, and gestational diabetes; 5 minute APGAR score <8; NICU admission; neurological disorder (e.g., seizure disorder); and psychiatric or learning disorder in the parents or siblings (including maternal depression requiring medication in the year prior to pregnancy). In addition to screening at the time of enrollment, ongoing screening for clinically concerning behaviors using validated tools were performed to identify at-risk children and exclude them from subsequent analysis.
Additional inclusion criteria for these analyses included: 1. Complete and artifact-free anatomical and rsFMRI dataset; 2. Completed neurocognitive assessments; 3. Valid medical history information including gestation duration, birth weight, and 5-minute APGAR scores; and 4. Family history information including family income and parental education levels. Each child provided a single neuroimaging and neurocognitive dataset to this analysis.

MRI Acquisition and Functional Connectivity Processing
All neuroimaging data were acquired on a 3T Siemens Trio scanner with a 12-channel head RF array. rsFMRI data were acquired during natural sleep with the following parameters: TE=34ms, TR=2.5s, flip angle=80 degrees, field of view=24x24cm 2 , imaging matrix=80x80, and 32 interleaved 3.6mm slices (for a voxel resolution: 3x3x3.6mm 3 ). BW=751Hz/pixel, and GRAPPA acceleration factor of 2. We acquired 164 volumes for a total acquisition time of approximately 7 minutes. To achieve successful scanning without sedation, scans were scheduled around the child's usual nap time. Once asleep, the child was transferred to the scanner bed. To maintain the sleep state, we used noise-absorbing foam inside the bore, headphones, and custom MRI pulse sequences with reduced gradient slew rates to reduce noise levels. Children were positioned in an immobilizer to keep them still and monitored visually by research staff at the scanner bed as well as outside the scanner using an IR camera. Physiological monitoring via a pulse-oximeter was also used to identify child distress as well as to time scans to periods of deep sleep.
T1-weighted magnetization-prepared rapid acquisition gradient echo anatomical data were acquired with an isotropic voxel volume of 1.2x1.2x1.2mm 3 , resampled to 0.9 x 0,9 x 0.9mm 3 Sequence specific parameters were: TE=6.9ms; TR=16ms; inversion preparation time=950ms; flip angle=15 degrees; BW=450Hz/Pixel. The acquisition matrix and field of view were varied according to child head size in order to maintain a constant voxel volume and spatial resolution across all ages (Dean et al., 2014).
To extract fc values, the rsFMRI data were first preprocessed (including realignment, centering, motion correction and scrubbing) with the CONN-fMRI toolbox for SPM 8 (Whitfield-Gabrieli et al., 2012) on MATLAB and registered to our child study template using FSL FLIRT (Smith et al., 2004) and ANTS (Avants et al., 2014). ROI-to-ROI connectivity analyses were performed, computing the correlation of spontaneous BOLD activity between network regions. Thirty-two anatomical ROIs were used in the network analysis (Whitfield-Gabrieli et al., 2012). Using the implemented CompCor strategy (Behzadi et al., 2007), the effect of nuisance covariates including BOLD signal fluctuations from CSF, white matter and their derivatives, as well as the realignment parameter noises were reduced. Data were simultaneously band-pass filtered (0.008<f<0.09HZ) and global signal regression was applied, including derivatives. Finally, we scrubbed volumes (Power et al., 2014) if there was significant motion during data acquisition (i.e., DVARS > 5 or FD were > .5).

TapTap iPad Task
This is an iPad-based experimental task presenting the participant with randomly ordered static images, appearing around the screen. It is based on principles of motor and cognitive inhibition, as a reflection of a child's ability to decide about the 'appropriateness' of a response based on predetermined, age-based, instructional criteria. Following a simpler reaction time module, where the participant is expected to tap on the image of a dog that appears in sequence, the participant then moves on to two Inhibition tasks. In the non-competitive inhibition module, the child is expected to tap on the image of a dog, while avoiding the image of a ball. Stimuli are presented in four, 10-image blocks for a total of 40 images. The competitive inhibition module presents the participant with increasing decision-making complexity, with the added step of classifying the stimulus before determining if it is appropriate to click/tap. The child is expected to tap on the image of an animal (dog, horse, cow etc.) while avoiding the image of a ball. As with noncompetitive inhibition, competitive stimuli are also presented in four, 10-image blocks for a total of 40 images.
A participant successfully inhibits when he/she can avoid tapping on the image of a ball. A score for successful inhibition is computed from the number of balls avoided, corrected for the participant's attentiveness and interaction with the assessment, and reflects the ability of the participant to 'inhibit'. The Successful Inhibition Score ( ! ) is computed by the following linear representation: with Ic = number of correct inhibitions, To = number of targets timed out, Tt =total number of targets in assessment, and To /Tt = interactivity of the participant.
The interactivity of the participant is a ratio of the amount of times the participant ignored the positive target stimulus over the total number of Target stimuli presented. The interactivity is the multiplied by the correct inhibitions to correct for ignored rather than inhibited instances. The ignored values are then subtracted from the number of correct inhibitions, giving the score.
In order to assess improvement during successful non-competitive and competitive inhibition, we calculated the delta of the error rate from the last minus the first of the four 10-image blocks.
Error rates were calculated per block by using the total number of balls clicked minus the number of total images by number of blocks.

Statistical Analysis
All age groups were of similar size and closely matched for biological sex to remove potential confounding influences (Table 1). Statistical analyses using the CONN toolbox were performed to 1) study the effect of the TapTap variables of the older over the youngest age group (5 year-olds > 2-3 year olds; 3-4 year olds > 2-3 year olds); and 2) study the influence of the TapTap variables on functional connectivity independent of age using a regression model. All analyses were corrected for biological sex and age (adjusted to a 40-week gestation) for the regression analysis.
For all analyses, statistical significance was defined as p≤.05 with FDR seed-level correction applied for all possible pairwise correlation between the 32 seeds.

Successful Inhibition Score
With increasing successful inhibition score, connectivity increased in frontal, temporal and subcortical relay stations (cerebellum, thalamus, pallidum), while posterior occipital and parietal functional connectivity decreased (for more details, see Figure 1, Table S1).

Improvement during competitive inhibition
We also assessed the improvement shown as a child proceeded through the blocks of stimuli, with error reductions expected from block 1 through block 4 (delta error reduction). With increasing delta error rate on non-competitive inhibition, overall fc decreased in frontal, parietal, temporal and putamen (for more details, see Figure 2, Table S1).

Improvement during competitive Inhibition
Increasing delta error rate on competitive inhibition led to an overall increase in the somatosensory, insular, occipital and pallidum fc. Analysis by age revealed increases in occipital, insular and anterior cerebellar fc in 3-4 year-olds when compared with 2-3 year-olds. A similar pattern of increased connectivity was noted when comparing 5 year-olds with 2-3 year-olds, specifically in frontal, temporal, occipital, parietal, insular, posterior cerebellar and caudate (for more details, see Figure 3, Table S1).

Discussion
In this study, we found that overall successful inhibition as well as improvement during competitive and non-competitive inhibition was associated with convergent patterns of functional connectivity (Figures 1-3). These patterns showed refinement with task difficulty and with children improving with age.

Patterns of successful inhibition
In order to investigate inhibitory control accounting for participants' attentiveness, we studied fc correlations with increasing successful inhibition score (Figures 1-3). Higher successful inhibition was associated with increased fc in frontal, temporal and subcortical relay stations (cerebellum, thalamus, pallidum), while posterior occipital and parietal functional connectivity decreased (for more details, see Figure 1, Table S1). As children got better at inhibition with age, these widespread patterns refined towards fc increases in frontal, parietal and subcortical relay stations (posterior cerebellar and caudate), paired with parahippocampal and parietal fc decreases. Our results parallel previous reports of inhibitory control being supported by a distributed neural network, with strong focus on frontal and parietal fc recruitment (Nomi & Uddin, 2015;Garavan et al., 1999;Dove et al., 2000;Chen et al., 2009;Schmitz et al., 2006). Both regions have been shown to support behavioral performance of attentional disengagement (Csibra, Johnson, & Tucker, 1997;Csibra, Tucker, & Johnson, 1998), a key element of asserting inhibitory control (for example, O'Connor et al., 2012).
While taking participants' attentiveness into account, our successful inhibition score displayed a very different fc pattern when contrasting the youngest groups (2-3 year-olds and 3-4 year-olds 13 respectively; Figure 1) than the youngest with the oldest group (2-3 year-olds and 5 year-olds respectively). Early in development, inhibitory control is primarily driven by the orienting attention network (Posner et al., 2014). By the age of four, an observed shift in self-regulation takes place, enabling later forms of control to be driven by executive attention networks. Results suggest that by 30 months, children may be in a translational period where they use both attentional systems to support self-regulation. Our findings contrasting children in this age range could reflect part of this early maturational process. Furthermore, our findings indicate a shift towards increased longrange fronto-parietal connections with age and experience when comparing the oldest with the youngest group. This could indicate a development towards adult-like connectivity patterns, as long-range connectivity and specialisation are direct contributors to adult's superior inhibitory control skills (Mehnert et al., 2013), further supporting the notion of our age-staged contrasts capturing an underlying maturational process.

Differences between non-competitive and competitive inhibition
When contrasting the more complex task of competitive inhibition with non-competitive inhibition, two very different patterns emerged (Figures 2-3). As children got better at noncompetitive inhibition, fc was reduced in frontal, parietal, and temporal regions (Figure 2). When comparing the older and in turn more experienced children with the youngest group, the initially wide-spread fc patterns reduced in number of connections and clusters with age. This significant reduction in overall fc recruitment with inhibition ability could indicate an early streamlining of neural networks in performance of executive functions.
In contrast to non-competitive inhibition, where only one response has to be inhibited, competitive inhibition is a multistep selection process involving inhibition among several coactive responses (Erb et al., 2020). To master this more complex task, we demonstrated fc patterns of competitive inhibition increases in occipital, insular and cerebellar recruitment with age and skill, with highest competitive inhibition ability being associated with the establishment of frontal, temporal and parietal long-range connections and hubs (Figure 3). This additional fc recruitment could reflect a reaction to the more complex nature of the task, as more concentration and focus are needed to fulfil each step, with an additional layer of inhibition added. Findings from previous studies investigating work load and working memory maintenance showed that children displayed increased frontal and parietal cortical activation (Buss et al., 2014). The increase in activation in the parietal cortex was further higher in 4 year-olds than in 3 year-olds, possibly reflecting sensitivity to work load increases. Interestingly, our results also indicate additional recruitment of parietal fc in the older and in turn more experienced group when compared to the youngest group, which could be in response to the increased work load of the competitive inhibition task.

Conclusion
Inhibitory control abilities are core EF and continue to mature from early childhood until late into adolescence, getting more defined the older the child gets (Menon et al., 2001;Vara et al., 2014).
Gaining insights into maturational dynamics involved in the development of early EF can in turn allow to study influences on these processes, including genetics, nutrition, and sleep. Deficits of inhibitory control are common in neurodevelopmental disorders, such as autism spectrum disorder (Schmitt et al., 2019) and emerge as early as 24 months (St John et al., 2016). Insights into the early development and refinement of inhibitory control and other EF processes could provide essential information for the diagnosis and possibly treatment of such disorders.

Limitations
While our findings indicate additional fc recruitment with increasing task complexity and work load, our current imaging modality is unable to assess whether additional recruitment is done in a sequential manner. Furthermore, we are aiming to replicate our results in a larger cohort with different cultural settings, in order to replicate our findings and disentangle possible interferences of prior tablet-based skills.

Data availability statement
Data can be made available upon request.

Compliance with ethical standards
This study was approved by the local Internal Review Board and active written informed consent was collected for all participants given by the parent and/ or caretaker on behalf of their child.