The behavioral and neurobiological relationships between executive function and reading: A review and preliminary findings

Abstract

Despite decades of prior research, the mechanisms for how skilled reading develops remain elusive. Numerous studies have identified word recognition and oral language ability as key components to explain later reading comprehension performance. However, these components alone do not fully explain differences in reading achievement. There is ongoing work exploring other candidate processes important for reading, such as the domain-general cognitive ability of executive function (EF). Here, we summarize our work on the behavioral and neurobiological connections between EF and reading and present preliminary neuroimaging findings from ongoing work. Together, these studies suggest 1) that EF plays a supportive and perhaps indirect role in reading achievement and 2) that EF-related brain regions interface with the reading and language networks. While further work is needed to dissect the specifics of how EF interacts with reading, these studies begin to reveal the complex role that EF plays in reading development.

Lay Abstract

Many students struggle when learning to read. Our understanding of how and why reading development goes amiss is incomplete. In this paper, we recap past and present findings about how an important cognitive skill, executive function (EF), is related to reading achievement. We outline brain and behavioral findings and discuss how EF shapes typical and atypical reading development.

Introduction

The ability to read is a vital life skill linked to educational, vocational, and health outcomes (Cain & Oakhill, 2006; DeWalt et al., 2004). Despite its overall importance, many individuals struggle to attain adequate reading skills. According to the most current study of United States 4^th, 8^th, and 12^th graders by the National Assessment of Educational Progress (NAEP), approximately 60% of children and adolescents achieve at or below the level of proficiency for their grade (NCES, 2022). Even though a voluminous number of studies have tried to elucidate the causes and consequences of reading failure, understanding why some children become skilled readers, and some do not, is not fully understood.

Despite the lingering ambiguity, over 30 years of scientific research have pinpointed key components of reading development. These studies have shown that reading comprehension (RC) relies upon the appropriate maturation of word recognition (i.e., decoding) and oral language comprehension abilities. Thus, children must develop adequate skills in both components and learn how to weave those skills together to achieve successful reading. This formative framework, known as the Simple View of Reading (Gough & Tumner, 1986; Hoover & Gough, 1990), has been integral in advancing our understanding of reading development and disability. Numerous studies show that word recognition and oral language predict a substantial amount of variance in RC outcomes both concurrently and longitudinally (García & Cain, 2014; Kendeou et al., 2009; Quinn & Wagner, 2018). Furthermore, as predicted by the model, when children struggle with word recognition (dyslexia), RC also suffers (Hulme & Snowling, 2016; Ransby & Lee Swanson, 2003). While specific reading abilities are malleable early in development (Bus & IJzendoorn, 1999; Clarke et al., 2010; Frick et al., 2013; Melby-Lervåg et al., 2012), RC is often difficult to modify, especially in older children (Vaughn & Fletcher, 2012).

Although the success of the Simple View has furthered our understanding of RC development and subsequently guided intervention, gaps remain in our knowledge of how to address and treat reading difficulties. First, despite significant efforts to improve reading outcomes, NAEP scores have remained relatively stagnant (NCES, 2022), despite efforts to improve them (e.g., Reading First Initiative, 2000). Second, while studies have shown substantial success in remediating word recognition deficits (Gersten et al., 2020; Wanzek et al., 2010, 2018), remediating deficits in older readers or reading comprehension ability remains problematic (see Capin et al., 2023). Third, while the components of the Simple View account for a substantial amount of variance in RC, studies show that they do not fully explain reading outcomes. For example, some readers show adequate word recognition skills but poor RC abilities (i.e., specific reading comprehension deficit, or SRCD). Those with SRCD have poorer RC abilities than would be predicted by either word recognition or oral language alone (Spencer & Wagner, 2018), thus raising the consideration that other cognitive processes may impact RC (Catts et al., 2006; Landi & Ryherd, 2017). Given these points, there is a need to further our understanding of reading beyond the processes linked solely to word recognition and oral language.

While various candidate processes may impact RC beyond word recognition and oral language ability (e.g., processing speed [Willcutt et al., 2005]; theory of mind [Kim, 2015], and inference making [Cain et al., 2001]), one set of processes that has garnered particular interest in the reading research field is executive function (EF). EF encompasses domain-general cognitive control skills, such as shifting, inhibition, and working memory (Diamond, 2013; Miyake et al., 2000), that have clear theoretical linkages to the reading process. To this end, discourse processing theories highlight reading as an integrative and dynamic process and, although not directly incorporating EF terminology, underscore the direct and indirect impact of EF abilities on both lower- and higher-level reading. For example, the foundational Construction-Integration model by Kintsch (1988) emphasizes that readers must form a mental model in order to comprehend texts. Readers must constantly update this mental model by integrating bottom-up text processing with top-down reader knowledge – thus shifting between different sources of information and utilizing working memory to integrate knowledge. While there are many discourse processing theories (e.g., Reading Systems Framework by Perfetti, 1999; Landscape Model by van den Broek, 1995), all are based on the tenant that successful reading requires individuals to interactively build mental models of text, thus implying the central importance of EF.

Consistent with the theoretical reasons for linking EF and reading, empirical studies have shown that reading is associated with EF skills, both concurrently and longitudinally. Meta-analyses report an approximate .30 effect size for the EF-reading link (Follmer, 2018; Spiegel et al., 2021) and a .57 effect size for the connection between reading disability and EF impairment (Booth et al., 2010). Therefore, there is little question about whether EF is associated with reading; instead, the questions are how EF is linked and if EF is important beyond the Simple View of Reading components. Below we review our cognitive and neuroimaging work that has focused on examining how EF may be linked to reading outcomes, followed by presenting some preliminary neuroimaging findings from an ongoing study. Together, insights across these studies reveal more about the complex role of EF in reading and support our emerging hypothesis that EF plays a faciliatory role. In particular, our findings suggest that areas of the brain that support EF may play a role in coordinating the left hemisphere’s reading and language networks. Critically, we hypothesize that EF becomes especially important to rely on for those with weaker reading skills.

Cognitive Findings

Most studies focusing on how EF components are linked to reading outcomes have utilized structural equation modeling approaches, allowing for the examination of both direct and indirect relationships among variables (e.g., Arrington et al., 2014; Christopher et al., 2012; Georgiou et al., 2018; Haft et al., 2019; Kieffer et al., 2013; Spencer, Richmond, & Cutting, 2020; Spencer & Cutting, 2021). For example, Georgiou and colleagues (2018) investigated how the EF components of updating, inhibition, and shifting related to RC ability in a cohort of young adults. Their results revealed that only the shifting component had a significant and direct relationship with RC. However, this study, as well as others, hadn’t yet included separate word recognition and oral language components or multiple measures of EF (Arrington et al., 2014; Christopher et al., 2012; Georgiou et al., 2018, Haft et al., 2019; Kieffer et al., 2013). Given the need for understanding earlier developmental trajectories and more granular aspects of EF linking to reading, we examined how two central components of EF, working memory and cognitive flexibility, related to the Simple View of Reading components in 271 children (Spencer, Richmond & Cutting, 2020). In this cohort of 9–14 year-olds, word recognition and oral language abilities mediated the relationship between EF and RC, with word recognition linking more to working memory and oral language linking more to cognitive flexibility. Additional work in our lab has probed what variables might mediate the relationship between EF abilities (i.e., working memory, shifting, and inhibition) and RC in younger readers (Spencer & Cutting, 2021). Consistent with prior work (e.g., Arrington et al., 2016; Haft et al., 2019; Spencer et al., 2020), Spencer and Cutting (2021) found that the association between EF and RC was mediated by word recognition. Interestingly, when examining gender as a potential moderator for these associations, there was a direct relationship between EF and RC for girls but not boys (with a trend-level significant difference directly comparing models for boys versus girls). These results indicate that girls may be able to draw upon EF skills more directly to facilitate their RC skills in order to perhaps compensate for weaknesses in other areas (e.g., poor word recognition).

Although prior and ongoing studies, including ours, suggest a role for EF in addition to word recognition and oral language in reading development, how these relationships emerge over time – and whether EF is a critical foundational skill or simply a correlate of word recognition and oral language – remains to be fully understood. For example, findings from a recent longitudinal study of early readers in an orthographically transparent language indicated that while EF was linked to reading comprehension, it did not provide additional explanatory power for predicting reading comprehension growth over time above oral language ability (Dolean et al., 2021). However, the indirect effects of EF on later reading comprehension outcomes were not considered. Other studies that have examined indirect effects of EF have shown that it does predict reading outcomes over time, albeit through the Simple View components; for example, a longitudinal study that followed Spanish-English bilingual adolescents from 6^th through 8^th grades revealed that EF predicted reading comprehension indirectly through either word recognition or oral language (Kieffer et al., 2021). Further, other literature suggests these relationships may differ depending on reading ability, be bidirectional, and begin to manifest at early ages (Peng et al., 2022). Finally, while EF is a significant predictor of academic readiness and achievement in early childhood (Diamond, 2013; St Clair-Thompson & Gathercole, 2006), the importance of EF and intervention response for struggling readers is less clear (Church et al., 2019). Establishing such associations will allow the field to continue advancing our understanding of how to intervene when reading development goes awry.

Overall, our studies, as well as others, suggest a complex picture of EF and reading and stress the importance of considering both the direct and indirect effects of EF on reading comprehension over time and across different reader profiles (Haft et al., 2019; Kieffer et al., 2013, 2021; Peng et al., 2022; Spencer, Richmond & Cutting, 2020; Spencer & Cutting, 2021). These studies collectively add to the growing literature indicating that 1) EF has subcomponents that differentially relate to distinct reading predictors and 2) EF appears to have an indirect role in reading comprehension. While intriguing, further interrogation is needed to make causal inferences. To this end, our lab is currently pursuing longitudinal and intervention studies to further elucidate these relationships (e.g., Spencer & Cutting, July 2021). Additionally, we are also probing EF’s role in reading using brain-based methodologies, which may serve to facilitate our understanding of how EF-related brain regions coordinate with the reading and language networks.

Neuroimaging Findings

In complement to the above cognitive studies, other studies have examined how the neurobiological networks that support EF intersect with and support brain areas readers utilize while processing written text at both the word and passage levels. Overall, we conjecture that neuroimaging methodologies may be particularly helpful in sorting out the role that EF may play in reading development. Cognitive studies have greatly expanded our understanding of the intersection of EF components and reading abilities; however, they are limited in enabling us to understand how different types of learners may have different neurobiological demands under various circumstances. Thus, using neuroimaging methodologies may add to our understanding of how EF influences academic outcomes by probing the underlying neural circuitry.

Various neuroimaging studies have found linkages between the brain networks that support EF and networks traditionally associated with processing written text. Resting-state imaging studies have helped to elucidate how the reading network overlaps with canonical EF networks. For example, over a decade ago, Vogel and colleagues (2013) used graph theory network analysis of resting state functional connectivity MRI data to define the reading network. They found that reading-related regions for both adults and children were sorted into pre-defined networks associated with EF and attention (i.e., the frontoparietal and cingulo-opercular control networks) rather than clustering into a separate grouping. In a similar vein, we found, using a large-scale synthesis of publicly available functional MRI (fMRI) data on Neurosynth.org, that the reading-related activity had the most overlap with the dorsal attention network (22%), the frontoparietal network (FPN; 19%), and the default mode network (17.8%, Bailey et al., 2018). These overlaps suggest that the processing of written text involves, to a certain extent, neural activity that has also been implicated in attention and cognitive control tasks. These findings lay the basic framework for understanding which of the brain’s general networks are also utilized for the more specific cognitive task of reading.

In addition to the resting state neuroimaging studies showing neurobiological linkages between the areas that support EF and reading, studies using task-based functional connectivity analyses offer a more concrete basis for some of our central hypotheses about the faciliatory role of EF in reading processes. Two such studies in adolescents found that across in-scanner reading, math, and EF tasks, three regions within the frontal cortex were commonly activated (see Banich et al., 2023 for a review; Wang et al., 2020a; Kim et al., 2022), suggesting a more domain-general function. These findings are consistent with earlier studies from our lab with solely fMRI reading tasks, revealing that activity and connectivity with frontal regions are important for RC (Aboud et al., 2016; Aboud et al., 2019), predict RC outcomes, and response to reading intervention (Aboud et al., 2018). For example, Aboud et al. (2016) examined the neural correlates of word recognition and reading comprehension and explored which areas were unique versus overlapping between the two tasks. The regions of common activation across the two tasks were then used as seeds for functional connectivity analyses, revealing 1) that prefrontal, working memory (WM) regions were connected to the angular gyrus and 2) that better comprehenders showed more connectivity between the two regions compared to poorer comprehenders. Further, as the prefrontal seed became more active, there was an increased coupling between reading-related regions (i.e., left occipitotemporal and left angular gyrus), which was modulated by out-of-scanner working memory ability. These findings suggest that the role of EF during text processing may be to facilitate connections within and between central reading hubs.

Another neuroimaging study in our lab probed how functional connectivity relationships might predict responsiveness to reading intervention (Aboud, Barquero & Cutting, 2018). We hypothesized that if EF did indeed play a faciliatory role in connecting reading-related regions, connectivity between prefrontal regions and reading and language areas, but not prefrontal brain activity alone, would predict intervention responsiveness. Findings showed that those who responded to the intervention had more involvement and connectivity of prefrontal regions while processing single words, with connectivity between prefrontal regions and reading-related regions predicting better responsiveness to reading intervention. Importantly, the responder and non-responder groups did not differ in their baseline reading skills, so behavioral scores at pre-intervention alone could not be used to predict which poor readers would or would not respond to intervention. Similar to Aboud et al. (2016), as the prefrontal region associated with EF became more active, more coupling was observed between reading-related regions (i.e., left middle temporal gyrus to inferior parietal lobe, and left inferior frontal gyrus to right inferior frontal gyrus). While other studies have examined how baseline task-based fMRI activity predicts response to intervention or reading growth (e.g., Nugiel et al., 2019; Hoeft et al., 2007; Krafnick et al., 2022), and a few have also used task-based connectivity measures to predict intervention response (Horowitz-Kraus et al., 2015; Horowitz-Kraus & Holland, 2015) or reading growth (Jasińska et al., 2021 (fNIRS)), the Aboud et al. (2018) study design enabled the examination of whether the way in which EF-related brain areas interact with reading-related regions is predictive of intervention response. These findings offer additional support for our hypothesis that brain regions supporting EF play a role in the processing of written text, and these regions link to variation in reading skills, with greater involvement of EF regions yielding more robust reading outcomes.

Together, these neuroimaging studies suggest that regions of the brain that support EF play a role in the processing of written text, and these regions link to variation in reading skills. Greater involvement of EF regions, either directly or indirectly, yields more robust reading outcomes.

Early Academic Achievement and Intervention Response: The Role of Executive Function

The aforementioned cognitive and neuroimaging findings from our lab laid the foundation for our ongoing NIH-funded project, “Early Academic Achievement and Intervention Response: Role of Executive Function.” In this project, we are examining the role of EF in predicting reading and math growth in a large cadre of children from kindergarten through 1^st grade, capturing neurobiological measures starting in the fall of kindergarten. In 1^st grade, the poor readers in the group receive reading intervention, along with pre- and post-intervention MRI scans. Our overarching hypothesis is that neural networks supporting EF (i.e., the FPN) facilitate connections between skill-specific (i.e., reading and math) brain regions, which will then predict academic growth and outcomes. Further, we hypothesize that for poor readers, EF connectivity with reading-specific regions will predict responsiveness to reading intervention during 1^st grade. In other words, we hypothesize that the role of EF in academic growth is not through activity in the FPN itself, but instead through the FPN coupling with skill-specific areas of the brain. The findings outlined below are an initial pass at investigating whether our approach can begin to test some of these hypotheses. While data collection is ongoing and a significant caveat must be placed on them in that these analyses are preliminary, findings are promising.

Preliminary Findings: Overview

The current investigation examines neural and behavioral relationships for a subset of kindergarten children (N = 23) from the ongoing longitudinal project described above. During their visits, children completed out-of-scanner testing to characterize their behavioral EF and reading abilities. Additionally, they participated in a reading-related fMRI task, where they viewed letters, two-string letter combinations, and words. Here, we investigate 1) the relationship between behavioral EF ability and neural activity during early reading, 2) functional connectivity between EF-related brain regions and the rest of the brain, and 3) the relationship between significant functional neural connections and reading abilities.

Preliminary Findings: Methods

Participants

Forty-eight participants completed behavioral and neuroimaging data collection during the fall (Visit 1) and spring (Visit 2) of kindergarten. For the current preliminary study, we will discuss neuroimaging and behavioral EF data collected during Visit 1 and behavioral reading performance collected during both visits. After excluding participants for insufficient head coverage (n = 2), excessive in-scanner movement (n = 9), and irregular task performance (n = 17, see Supplementary Materials: fMRI Task, Acquisition, and Preprocessing for more information), the final analysis included twenty-three children (n = 23; 11 girls; M_{age Visit 1} = 5.7 ± 0.3 years; M_{age Visit 2} = 6.2 ± 0.3 years). This percentage of usable data is consistent with other fMRI studies of kindergarten children (e.g., Centanni et al., 2019; Wang et al., 2020b). A majority of the sample was identified as White and non-Hispanic/Latinx. All participants had normal hearing and vision, no history of developmental disorders, psychiatric diagnoses, traumatic brain injury, or epilepsy, and had age-standardized IQs > 85. See Supplementary Table 1 for more demographic information and Supplementary Figure 1 for an overview of children’s reading and EF abilities. Parental consent and child assent were obtained prior to data collection, and all study procedures were carried out in accordance with the university’s internal Institutional Review Board.

Behavioral Testing

To assess various components of children’s behavioral EF skills at Visit 1, we administered four different behavioral assessments to derive three EF components: 1) inhibition/attention scores were derived from the Flanker Test from the NIH Toolbox (Weintraub et al., 2014), 2) cognitive flexibility scores were derived from the Dimensional Change Card Sort (DCCS) Test from the NIH Toolbox (Weintraub et al., 2014), and 3) working memory scores were generated from the Verbal Attention and Numbers Reversed subtests on the Woodcock-Johnson (Tests of Cognitive Abilities-IV (WJ-IV; Schrank & Wendling, 2018). To assess children’s behavioral reading abilities at Visit 1 and Visit 2, we administered three behavioral assessments to derive two reading components: 1) basic reading scores were generated from the Letter-Word Identification and Word Attack subtests of the WJ-IV (Schrank & Wendling, 2018) and 2) passage comprehension scores were pulled from the Passage Comprehension subtest of the WJ-IV (Schrank & Wendling, 2018). See the Supplementary Materials for further explanation of behavioral measures. All behavioral analyses included age as a covariate. Standard scores were used for the inhibition and cognitive flexibility measures, while W scores were used for WJ-IV measures.

fMRI Task and Analyses

During the Visit 1 fMRI session, children participated in two runs of a reading-related task that were presented electronically using E-Prime 2.0 software (Psychology Software Tools, Pittsburgh, PA; RRID:SCR_009567). All participants were trained on the task in a mock scanner prior to scanning. Participants saw a series of 140 randomized tokens in the center of the screen that were either uppercase letters (60 stimuli), legal two-letter combinations (40 stimuli), or simple words (40 stimuli). Further task details can be found in the Supplementary Materials.

fMRI and connectivity analyses were conducted in SPM12 (http://www.fil.ion.ucl.ac.uk/spm/; RRID:SCR_007037) and the CONN toolbox (CONN 21b; https://www.conn-toolbox.org; RRID:SCR_009550), respectively. Subject-level contrast maps of Reading > Fixation were brought to a one-sample t-test to analyze activity related to reading processing while controlling for lower-level visual input. To examine the neural correlates of EF ability, three regressions were performed between each participant’s contrast beta map and their respective behavioral EF scores (i.e., Flanker, DCCS, and WM). These covariate analyses were performed within a mask of significant FPN activations, derived from the Yeo et al., 2011 resting state network parcellations, to test whether neural activity was significantly related to our EF covariates of interest. GLM results were reported at the whole-brain level using AFNI’s 3dClustSim (compilation date 2022; Cox et al., 2017; RRID:SCR_005927) to derive an appropriate cluster-correction threshold and correct for multiple comparisons. Group-level findings were subjected to thresholding of p-corrected < 0.05 (p-uncorrected < 0.001; k > 225). For connectivity analyses, seed regions were derived from significant clusters within the FPN mask for the covariate analyses (see Figure 2 below). Connectivity results were run at the whole-brain level, cluster p-FDR corrected < .05, and voxel p-uncorrected < .001. For details about GLM and connectivity processing steps, see Supplementary Materials).

Figure 2. Increased left MFG and right dlPFC activity is associated with better behavioral EF performance.

a) The flanker covariate GLM analysis of Reading > Fixation revealed increased activity in the left MFG and right dlPFC. Results are displayed at p-corrected < .05 (p-uncorrected < .001, k > 225).

b) & c) Correlations between flanker task performance and peak activations within the left MFG (Figure 2b; r₂₂ = 0.56) and right dlPFC (Figure 2c; r₂₂ = 0.46).

Preliminary Findings: Results

Neural Activity During Reading

The Reading > Fixation contrast revealed a robust set of activations in expected reading and language areas (Martin et al., 2015), including increased activity in the left inferior frontal gyrus (IFG), supplementary motor area (SMA), bilateral inferior parietal lobules (IPLs), and bilateral occipitotemporal cortices (OTCs) (Figure 1, Supplementary Table 2).

Figure 1. Reading and language network activity for Reading > Fixation.

The GLM of letters, letter combinations, and words compared to the fixation baseline revealed activity in reading and language brain regions including left IFG, SMA, bilateral IPLs, and bilateral OTCs. Results are displayed at p-corrected < .05 (p-uncorrected < .001, k > 485).

Neural Activity and Behavioral EF

Next, we conducted covariate analyses of neural activity for Reading > Fixation with our continuous measures of out-of-scanner behavioral EF (i.e., Flanker, DCCS, and WM) at Visit 1 within the hypothesis-driven FPN mask. Activity for the reading task within the FPN mask did not significantly correlate with DCCS or WM performance. However, flanker task performance showed significant positive correlations with activity in the left middle frontal gyrus (MFG; Brodmann Area 6) and right dorsolateral PFC (dlPFC; Brodmann Area 8) (Figure 2a, Supplementary Table 3). We extracted peak voxel activity (i.e., maximum t-values) within these two clusters for each participant. Flanker performance significantly positively correlated with peak voxel activity within the left MFG (Figure 2b; r₂₂ = 0.56; p < .05) and right dlPFC (Figure 2c; r₂₂ = 0.46, p < .05). Overall, children who had increased activity in these regions also tended to have better performance on the flanker task (i.e., better inhibitory control).

Functional Connectivity

The significant clusters from the flanker covariate GLM were used as seeds for a whole-brain functional connectivity analysis. While the time course for the left MFG didn’t show significant connectivity to other brain regions, the right dlPFC seed was positively correlated with activity in the left ventral OTC, more specifically, the left fusiform gyrus (Brodmann Area 37; x = −40, y = −69, z = −16; k = 399; Figure 3). Functional connectivity scores between these regions for the reading task were less than the fixation baseline connectivity scores, but still positive overall.

Figure 3. Functional connectivity between right dlPFC and left ventral OTC.

Functional connectivity analysis from the flanker covariate analysis right dlPFC seed (left, in red) revealed positive connectivity between the left ventral OTC (right, in blue) for Reading > Fixation. Results are significant at the whole-brain level, cluster p-FDR corrected < .05 and voxel p-uncorrected < .005.

To examine whether there was an association between children’s reading abilities and the degree to which their right dlPFC and left vOTC were connected during reading, we extracted participant-level connectivity z-scores between the two regions for each condition (i.e., one connectivity score between right dlPFC and left vOTC for Reading and another for Fixation). We then conducted multiple linear regression analyses to predict Visit 1, Visit 2, and growth in reading abilities (Visit 2 - Visit 1) from connectivity scores during reading. Connectivity between right dlPFC and left vOTC during in-scanner reading did not significantly predict Visit 1 or Visit 2 scores for BR or PC, or growth in BR ability. However, connectivity between right dlPFC and left vOTC during in-scanner reading significantly predicted growth in PC ability (R² = 0.32, F_2,20 = 4.62, p < .05; Connectivity: β = 0.44, t = 2.40, p < .05, age: β = 0.34, t = 1.83, p = 0.08; Figure 4).

Figure 4. Functional connectivity predicts growth in reading ability.

Functional connectivity between right dlPFC and left vOTC during in-scanner reading was significantly associated with children’s growth in PC ability from Visit 1 to Visit 2 (r₂₂ = .43).

While the multiple regression analyses were promising, we wanted to see if this relationship between growth in reading ability and connectivity during reading would remain significant after additionally controlling for general functional connectivity patterns (i.e., during fixation baseline). To account for this potential relationship between reading trajectories and general connectivity of the right dlPFC and left vOTC, we performed a hierarchical linear regression analysis (summarized in Table 1) to predict growth in PC scores (Visit 2 - Visit 1) from connectivity during reading while also controlling for age and connectivity during fixation.

Table 1.

Hierarchical linear regression analysis results predicting PC growth (Visit 2 - Visit 1).

Independent Variables	R 2	Δ R2	β	t	F
Model 1	.12	-			1.36
1. Age			.35	1.65
2. Fixation Connectivity			.03	0.14
Model 2	.33	.21			3.18*
1. Age			.33	1.77
2. Fixation Connectivity			−.14	−0.72
3. Reading Connectivity			.49	2.47*

^*= p < .05

The first model predicted growth in PC ability from the independent variables age and connectivity during fixation. The overall model was not significant (F_3,19 = 1.36, p = .28). The second model predicted growth in PC ability (Visit 2 - Visit 1) from age, connectivity during fixation, and connectivity during reading. The overall model was significant (F_4,18 = 3.18, p < .05) with connectivity between right dlPFC and left vOTC during reading (β = 0.49, t = 2.47, p < .05) as a significant predictor. The first model explained 12% of the variance in PC growth, while the second model explained 33% of the variance. This 21% variance increase was statistically significant (F_1,19 = 6.10, p < .05).

Preliminary Findings: Discussion

The current study presented preliminary findings for cognitive and neuroimaging relationships in a cohort of kindergarten children. We investigated 1) neural activity during a letter and word reading fMRI task, 2) how activity during that task, within the frontoparietal network, was associated with children’s behavioral EF performance, 3) how those EF-related regions were connected to the rest of the brain during reading, and 4) if that functional connectivity was predictive of reading trajectories. Compared to the fixation baseline, letter and word reading recruited the engagement of reading- and language-associated regions (Martin et al., 2015), indicating that our task recruited neural activity as designed. Of particular interest, we observed bilateral diffuse activation of ventral OTCs, including the putative visual word form area (VWFA), which is particularly tuned to processing word forms across development (Dehaene & Cohen, 2011; McCandliss et al., 2003). The engagement of the putative VWFA and its right hemisphere homolog is typical in early readers, with activations becoming more left-lateralized as children gain reading expertise and fluency (Turkeltaub et al., 2003). In addition to the ventral OTC activations, we also saw robust activity in the left IFG and temporoparietal regions. These regions are thought to form the phonological decoding circuit that early readers rely on to sound out written words (Pugh et al., 2001; Turkeltaub et al., 2003).

Next, we investigated how activity within the FPN for the reading task was associated with out-of-scanner behavioral performance on three EF measures: the flanker task (inhibitory control), the dimensional change card sort task (shifting), and a verbal and numerical composite working memory score. While neural activity was not significantly associated with shifting or working memory scores, activations in the left MFG and right dlPFC were associated with children’s performance on the flanker inhibitory control task. Children with relatively more activity in these regions tended to have better inhibitory control, which is consistent with both behavioral findings showing a link between EF and word reading (Haft et al., 2019; Spencer & Cutting, 2021), as well as neuroimaging findings that indicate FPN recruitment is associated with various components of EF (Bunge & Wright, 2007; Engelhardt et al., 2019; Roe et al., 2018). We hypothesize that the reason why inhibition, but not the other EF measures, was significantly linked to neural activity was related to the age of our participants. Inhibition is a foundational EF component that appears earliest, improves significantly during the preschool years, and remains relatively stable throughout later years (Best & Miller, 2010; Miyake et al., 2000). Thus, it is possible that neural associations with other EF components would appear at later developmental windows.

Most central to our current hypotheses about how EF may be linked to reading development, we conducted a novel analysis by examining how EF-related regions identified from the covariate analysis coupled to other brain regions. Using a seed-to-voxel connectivity analysis, we found that the time course of the right dlPFC seed was significantly associated with the time course of the left ventral OTC (Bouhali et al., 2014). Activity within this EF brain region was positively correlated with activity in this central reading-related region, suggesting a relationship between EF and reading neural attunement during letter and word processing. While we saw this interesting connection between the EF and reading networks, we didn’t see other reading or language-related regions implicated. We conjecture that this connectivity relationship was observed, in part, due to the age of the children in our sample and the demands of the task. As was seen in Aboud et al., 2016, we would expect that connectivity between the EF network (e.g., the prefrontal cortex) and higher-level comprehension regions (e.g., the angular gyrus) would be observed in older children performing a reading comprehension task, with EF regions helping to orchestrate the complex cognitive process of comprehending prose. Further, we investigated if right dlPFC and left vOTC connectivity predicted concurrent or longitudinal behavioral reading outcomes. We found that children who showed more connectivity between these regions during our reading task had more growth in their reading comprehension ability. Connectivity between right dlPFC and left vOTC did not predict children’s word recognition ability, either concurrently or longitudinally. Considering that the functional connectivity relationship we investigated was a higher-level EF region interfacing with the reading network, it follows that the behavioral association would be present with a more complex reading task (comprehension). Importantly, we found that this neural relationship was implicated in reading growth, rather than concurrent ability. We conjecture that the connectivity between these networks will increase as children develop and that higher-level reading comprehension regions (e.g., the angular gyrus) may also begin to interface with the executive systems.

Overall, while our results must be interpreted with caution and replicated with larger sample sizes in future cohorts, they provide insight into how to examine reading growth and performance using neural indices, which may be particularly important as we explore predictors of response to reading intervention in poor readers.

Conclusion

The current article describes previous and ongoing work examining how EF is linked to reading achievement. There is little doubt at this juncture that EF is related to reading outcomes. However, how it is related to reading achievement, especially in poor readers, is of ongoing interest. Behavioral work continues to elucidate how EF components interact with reading abilities and how EF may play a supportive role in various aspects of reading development, though seemingly not directly on reading comprehension. Additionally, neuroimaging findings suggest that EF-related brain regions and their functional connectivity with the rest of the brain are crucial for typical reading development and response to reading intervention in those who struggle. Lastly, the preliminary findings from our ongoing project reveal 1) which FPN regions (linked to EF) support early letter and word reading processing, 2) how those EF regions interface with the reading network, and 3) how functional connectivity between EF and reading regions is important for reading growth. With data collection still underway, we will soon be able to probe longitudinal neural relationships and their importance for various reading trajectories. Our hope is that our ongoing work will expand our understanding of how EF is related to reading and other academic outcomes (i.e., math) and whether that relationship is critical to improving outcomes for those with dyslexia and other learning difficulties. Together, our work reveals more about the complex interaction between EF and reading development and serves as a platform for future investigations that will further our understanding of how to improve reading instruction and performance.