Introduction
Cognitive and neurobiological characteristics of children with reading and attention difficulties
Reading refers to the ability to translate written language to spoken language using phonological, orthographical, and semantic processing (Glezer, Jiang et al., Reference Glezer, Jiang and Riesenhuber2009). The Simple View of Reading model (SVR) postulates that language processing (including semantics and the meaning of language presented orally) and word decoding (matching letters to sounds, i.e., phonological processing and visual processing or orthographical) are integrated to achieve reading comprehension, or understanding the meaning of text, and the ability to learn and integrate information from the text and retain it for future use (Gough, Reference Gough and Tunmer1986). These skills are supported by various brain regions, including the supramarginal/angular gyrus (AG), the left occipitotemporal (OT) area, particularly the visual word form area (VWFA), and the inferior frontal gyrus (IFG), respectively (Booth, Burman et al., Reference Booth, Burman, Meyer, Gitelman, Parrish and Mesulam2004; Sliwinska, Khadilkar et al., Reference Sliwinska, Khadilkar, Campbell-Ratcliffe, Quevenco and Devlin2012; Ishkhanyan, Michel Lange et al., Reference Ishkhanyan, Lange, Boye, Mogensen, Karabanov, Hartwigsen and Siebner2020). However, these brain regions do not uniquely support these aspects of reading. For instance, the left IFG has been linked with early phonological processing during visual word recognition (Cornelissen, Kringelbach et al., Reference Cornelissen, Kringelbach, Ellis, Whitney, Holliday and Hansen2009), pre-reading abilities such as phonology and speeded naming (Benischek, Long et al., Reference Benischek, Long, Rohr, Bray, Dewey and Lebel2020), and in syntactic processing, specifically in unification operations during sentence comprehension (Acheson and Hagoort, Reference Acheson and Hagoort2013). Similarly, the left AG has also been associated with semantic processing operations (Paz-Alonso, Oliver et al., Reference Paz-Alonso, Oliver, Lerma-Usabiaga, Caballero-Gaudes, Quiñones, Suárez-Coalla, Duñabeitia, Cuetos and Carreiras2018), and the VWFA integrates visual information with phonological and semantic processing contributing to the efficient recognition of whole words. Extended versions of the SVR model have recently proposed an integration of additional cognitive abilities that contribute to reading comprehension, such as executive function (EF) (Kim, Reference Kim2020).
EF refers to a set of cognitive abilities traditionally broken into three sub-components: inhibition, switching, and working memory updating (Miyake, Friedman et al., Reference Miyake, Friedman, Emerson, Witzki, Howerter and Wager2000; Miyake and Friedman, Reference Miyake and Friedman2012). EF plays a critical role in reading comprehension and supports reading comprehension together with word decoding and language processing (Nouwens, Groen et al., Reference Nouwens, Groen, Kleemans and Verhoeven2021). Inhibition involves active restrain or suppression of dominant responses, which improves focus on the relevant text (Borella, Carretti et al., Reference Borella, Carretti and Pelegrina2010). Switching aids in shifting attention between different mental states, activities, or tasks such as decoding words and comprehending sentences (Horowitz-Kraus, Reference Horowitz-Kraus2023). Working memory is crucial for holding and manipulating information during reading, such as remembering the beginning of a sentence while reading to the end (Cain, Oakhill et al., Reference Cain, Oakhill and Bryant2004; Butterfuss and Kendeou, Reference Butterfuss and Kendeou2018; Cartwright, Reference Cartwright2012). These EF components are essential for achieving fluent and effective reading (Horowitz-Kraus, Reference Horowitz-Kraus2023).
Neurobiologically, EF is thought to be governed by the cingulo-opercular (CO) and frontoparietal (FP) networks, together with dorsal and ventral attention networks (DAN, VAN) supporting attention abilities (Dosenbach, Fair et al., Reference Dosenbach, Fair, Cohen, Schlaggar and Petersen2008). These networks interact with traditional reading areas such as the IFG, AG, and VWFA in the left OT region (Vogel, Church et al., Reference Vogel, Church, Power, Miezin, Petersen and Schlaggar2013; Burgess and Cutting, Reference Burgess and Cutting2023). Therefore, targeting EF in interventions could potentially enhance reading abilities by leveraging these shared pathways and supporting the cognitive processes that underpin both EF and reading (Horowitz-Kraus, Vannest et al., Reference Horowitz-Kraus, Vannest, Gozdas and Holland2014; Nouwens, Groen et al., Reference Nouwens, Groen, Kleemans and Verhoeven2021; Horowitz-Kraus, Reference Horowitz-Kraus2023). This connection is particularly relevant given that improvements in EF have been shown to correspond with gains in reading abilities in typical readers (TR) as well (Sesma, Mahone et al., Reference Sesma, Mahone, Levine, Eason and Cutting2009; Horowitz-Kraus, Reference Horowitz-Kraus2016).
Dyslexia is a specific reading disability with neurobiological origins (Lyon, Shaywitz et al., Reference Lyon, Shaywitz and Shaywitz2003) affecting about 7–10% of primary school students (Yang, Li et al., Reference Yang, Li, Li, Zhai, An, Zhang, Zhao and Weng2022), characterized by reading difficulties (RD), including below-average decoding accuracy and speed, and/or fluent word recognition skills, and often reading comprehension as a secondary consequence of poor word reading (Lyon, Shaywitz et al., Reference Lyon, Shaywitz and Shaywitz2003; IDA, 2011). However, some children with dyslexia also have developmental language disorder, which involves an impairment in language comprehension that can independently impair reading comprehension regardless of decoding ability (Catts, Hogan et al., Reference Catts, Hogan and Fey2003; Adlof and Hogan, Reference Adlof and Hogan2018).
Children with RD demonstrate difficulties in the classical EF components (inhibition, working memory, and shifting) as well as visual attention, speed of processing, and error monitoring in the linguistic and nonlinguistic domains (Horowitz-Kraus, Toro-Serey et al., Reference Horowitz-Kraus, Toro-Serey, DiFrancesco and Ward2015). These EF deficits are also observed in children with attention-deficit/hyperactivity disorder (ADHD) and may contribute to the high comorbidity between RD and ADHD (Germanò, Gagliano et al., Reference Germanò, Gagliano and Curatolo2010). Speed of processing, in particular, has been highlighted as a crucial factor in EF deficits associated with RD and contributes to the overlap between RD and ADHD (Peterson, Boada et al., Reference Peterson, Boada, McGrath, Willcutt, Olson and Pennington2017). This is in line with the multiple deficit model which suggests that various cognitive deficits, including rapid automatized naming and processing speed, collectively influence the manifestation of RD and ADHD (Rucklidge and Tannock, Reference Rucklidge and Tannock2002; Martinussen, et al., Reference Martinussen, Hayden, Hogg-Johnson and Tannock2005; McGrath, et al., Reference McGrath, Pennington, Shanahan, Santerre-Lemmon, Barnard, Willcutt, DeFries and Olson2011; McGrath, et al., Reference McGrath, Peterson and Pennington2020).
ADHD is characterized by behavioral symptoms of inattention, impulsivity, and hyperactivity, affecting about 8–10% of the population (Polanczyk, Willcutt et al., Reference Polanczyk, Willcutt, Salum, Kieling and Rohde2014). Even though ADHD and RD are defined as two different conditions in the DSM, they frequently co-occur; while 7% of the general population has a reading disorder, the probability of reading disorders among the ADHD population is 25–40% (August and Garfinkel, Reference August and Garfinkel1990; Boada, Willcutt et al., Reference Boada, Willcutt and Pennington2012). The relationship between RD and ADHD is a topic of ongoing debate among researchers, with some researchers suggesting that RD + ADHD comorbidity involves unique characteristics that do not exist in RD and ADHD alone (Rucklidge and Tannock, Reference Rucklidge and Tannock2002) and that there are mutual symptoms with different levels of severity (Katz, Brown et al., Reference Katz, Brown, Roth and Beers2011). On the other hand, some researchers consider RD + ADHD as a combination of both disorders (Shanahan, Pennington et al., Reference Shanahan, Pennington, Yerys, Scott, Boada, Willcutt, Olson and DeFries2006), as both children with RD-only and those with RD + ADHD exhibit varying degrees of EF difficulties that may be most affected in children with both conditions (Rucklidge and Tannock, Reference Rucklidge and Tannock2002; Al Dahhan, Halverson et al., Reference Al Dahhan, Halverson, Peek, Wilmot, D’Mello, Romeo, Meegoda, Imhof, Wade, Sridhar, Falke, Centanni, Gabrieli and Christodoulou2022; Al Dahhan, Halverson et al., Reference Al Dahhan, Halverson, Peek, Wilmot, D’Mello, Romeo, Meegoda, Imhof, Wade, Sridhar, Falke, Centanni, Gabrieli and Christodoulou2022). Both perspectives suggest that children with RD + ADHD may show greater gains from a reading intervention with EF principles embedded, although the specific EF domain shown to improve may vary across children with RD-only versus RD + ADHD.
Training reading by stimulating both EF and reading
Interventions stimulating both EF and reading may be particularly important for children with RD + ADHD and can be achieved using an EF-based reading training (“READ”) (Horowitz-Kraus et al., Reference Horowitz-Kraus, Cicchino, Amiel, Holland and Breznitz2014, Horowitz-Kraus, Cicchino et al., Reference Horowitz-Kraus, Cicchino, Amiel, Holland and Breznitz2014, Horowitz-Kraus Reference Horowitz-Kraus, Toro-Serey, DiFrancesco and Ward2015; Horowitz-Kraus and Holland, Reference Horowitz-Kraus and Holland2015; Horowitz-Kraus Reference Horowitz-Kraus, Toro-Serey, DiFrancesco and Ward2015b; Horowitz-Kraus Reference Horowitz-Kraus, DiFrancesco, Kay, Wang and Holland2015c; Cecil et al., Reference Cecil, Brunst and Horowitz-Kraus2021). This is a reading fluency training program with embedded EF principles targeting inhibition, shifting, working memory, visual attention, and speed of processing (Horowitz-Kraus, Hershey et al., Reference Horowitz-Kraus, Hershey, Kay and DiFrancesco2019; Cecil, Brunst et al., Reference Cecil, Brunst and Horowitz-Kraus2021). During this training, sentences are presented on the computer screen and the letters are deleted at an individualized rate compared to their initial reading speed. This manipulation is triggering speed of processing (changes in deletion speed), working memory (more letters are being processed in a given time, compared to the individual’s regular comfortable speed), and the reader’s needs to inhibit themselves from regressing with their gaze to the beginning of the sentence (for more information, see Horowitz-Kraus et al., Reference Horowitz-Kraus, Vannest, Kadis, Cicchino, Wang and Holland2014; Horowitz-Kraus, Cicchino et al., Reference Horowitz-Kraus, Vannest, Kadis, Cicchino, Wang and Holland2014; Horowitz-Kraus Reference Horowitz-Kraus, Hershey, Kay and DiFrancesco2015; Horowitz-Kraus and Holland, Reference Horowitz-Kraus and Holland2015; Horowitz-Kraus Reference Horowitz-Kraus, Toro-Serey, DiFrancesco and Ward2015b; Horowitz-Kraus Reference Horowitz-Kraus, DiFrancesco, Kay, Wang and Holland2015c; Cecil et al., Reference Cecil, Brunst and Horowitz-Kraus2021). Previous studies using this training among children with RD-only demonstrated improved EF (inhibition, working memory, switching), better visual attention abilities, and faster speed of processing following this training, alongside improved reading fluency, accuracy, and comprehension (Horowitz-Kraus, Reference Horowitz-Kraus2013, Horowitz-Kraus et al., Reference Horowitz-Kraus, Vannest, Kadis, Cicchino, Wang and Holland2014, Horowitz-Kraus, Cicchino et al., Reference Horowitz-Kraus, Cicchino, Amiel, Holland and Breznitz2014; Horowitz-Kraus Reference Horowitz-Kraus, Hershey, Kay and DiFrancesco2015; Horowitz-Kraus and Holland, Reference Horowitz-Kraus and Holland2015; Horowitz-Kraus, Toro-Serey et al., Reference Horowitz-Kraus, Toro-Serey, DiFrancesco and Ward2015; Horowitz-Kraus Reference Horowitz-Kraus, DiFrancesco, Kay, Wang and Holland2015c, Cecil et al., Reference Cecil, Brunst and Horowitz-Kraus2021). One possible mechanism for this improvement involves the release of the bottleneck of working memory during the decoding process, encouraging the readers to shift the reading strategy to an orthographic, holistic reading (Breznitz, Reference Breznitz and Misra2003). It may also be that this deletion manipulation, combined with triggering EF, helps synchronize the auditory and visual modalities, reducing the proposed neural noise in these children (Hancock Reference Hancock, Pugh and Hoeft2017), which results in better reading fluency (Breznitz, Reference Breznitz2006). When comparing improvement among TR, RD-only, and RD + ADHD groups trained on the READ program, all groups showed enhancement in reading skills: single word/nonword reading, contextual sentence reading speed, contextual reading fluency and rate, which were transferred also to a novel text (i.e., not the sentences they were trained on) (Horowitz-Kraus, Hershey et al., Reference Horowitz-Kraus, Hershey, Kay and DiFrancesco2019). However, a comparison of neuroimaging results revealed differential changes in the functional connectivity (FC) between EF networks (CO and FP) during a lexical decision fMRI task (Horowitz-Kraus, Hershey et al., Reference Horowitz-Kraus, Hershey, Kay and DiFrancesco2019). Specifically, children with RD + ADHD showed greater increases in FC in neural networks related to ventral attention, dorsal attention, visual processing, and EF (Horowitz-Kraus, Hershey et al., Reference Horowitz-Kraus, Hershey, Kay and DiFrancesco2019), while children with RD-only showed increased FC in both higher and lower-level visual processing, as well as in networks related to language (semantic) and attention abilities (Horowitz-Kraus, Hershey et al., Reference Horowitz-Kraus, Hershey, Kay and DiFrancesco2019). These findings demonstrate the different effects of the READ training on the two clinical populations versus TR using fMRI during reading-related tasks. Whether this gain in EF engagement in these populations is also generalized to non-reading tasks (i.e., during resting-state fMRI) is another question examined in the current study.
Despite the reported behavioral gains in cognitive and reading abilities in children with RD + ADHD when using the EF-based reading intervention, it is still unknown if (1) the changes in neural circuits associated with EF also present in a task-free condition (i.e., resting-state fMRI condition) and (2) whether those who suffer from more severe EF challenges (such as in those with RD + ADHD) gain more from training (behaviorally and neurobiologically) than RD-only and TR. We hypothesize that due to the primary deficit in EF in children with RD + ADHD, this group will show the lowest scores on reading and EF measures and greater improvement in EF and reading with training. We also suggest that lower FC and a greater effect of training will be observed in this group on FC of networks supporting EF (CO and FP) and attention abilities (VAN, DAN) relative to children with RD-only and TR. Finally, we also hypothesize that reading improvement can be predicted by pre-training EF and resting-state fMRI data and gains in EF and in resting-state data.
Method
Participants
Fifty-five right-handed, native English-speaking children between the ages 8 and 12 were classified into three groups: 18 TR (10 males, 8 females), 18 children with RD-only (9 males, 9 females), and 19 children with both RD and ADHD (RD + ADHD; 10 males, 9 females). None of the participants had visual and hearing impairments or a history of neurological or psychiatric disorders other than RD and/or ADHD or intellectual disability (screened with the Test of Nonverbal Intelligence, TONI-III). Participants included in the RD + ADHD were previously diagnosed with ADHD by a professional before study participation and demonstrated currently elevated ADHD symptoms on the short version of the Conners-3 parent-report scale administered in our study (i.e., T > 60) (Conners, Reference Conners2008). Children with RD had an official diagnosis of a specific learning disorder with impairment in reading and the reading impairment was verified using a list of reading tests outlined below, with scores less than a standard score of -1.5 or below in at least two reading measures of this list as done in previous studies (Kovelman, Norton et al., Reference Kovelman, Norton, Christodoulou, Gaab, Lieberman, Triantafyllou, Wolf, Whitfield-Gabrieli and Gabrieli2012; Freedman, Zivan et al., Reference Freedman, Zivan, Farah and Horowitz-Kraus2020; Taran Reference Taran, Farah, DiFrancesco, Altaye, Vannest, Holland, Rosch, B. and Horowitz-KrausAccepted). Children in the RD + ADHD group adhered to the criteria of both ADHD and RD outlined above, whereas children with RD-only adhered only to the RD criteria and did not have elevated ADHD symptoms (i.e., Conners T < 60). Participants in the TR group showed averaged reading scores in the list of reading tests below (i.e., scores > 1 standard score below the mean) and did not have elevated ADHD symptoms (< 60 in the Conners).
Study procedure
Participants performed the reading and EF behavioral testing and a resting-state fMRI scan prior to completing a computerized EF-based reading intervention for 4 weeks, five times per week remotely. Behavioral testing and another fMRI resting-state scan were conducted after completing the intervention. The study was conducted at Cincinnati Children’s Hospital Medical Center (CCHMC), Ohio, USA, in accordance with the Helsinki Declaration and was approved by the CCHMC Institutional Review Board. Parents signed a written informed consent and children 10 years and older also signed a written assent. Participants and families were compensated for their time and travel. See Figure 1 for the study design.
Study procedure.
The EF-based reading training (READ)
The reading intervention lasted for 4 weeks, with 5 days of practice each week (approximately 20 min per session) totaling 20 sessions. The pretest reading evaluation phase included 20 sentences that remained “still” on the screen. The reader had to push the space bar when they finished reading the sentence and then answer a comprehension question. The initial reading rate was calculated based on the sentences in which the comprehension questions were answered correctly during the evaluation phase: the overall time to read the sentence was divided by the number of characters for this sentence and averaged across the msec/letter calculated across all correctly answered sentences.
During each session of the intervention, participants were presented with 50 different sentences one at a time, with varying levels of complexity. In each sentence, letters were progressively deleted from left to right at a specific pace. After reading each sentence, participants had to answer a multiple-choice comprehension question to ensure they read and understood the sentence. The pace of letter deletion was made faster by 5% after 7 out of 10 correctly comprehended sentences. All children performed at least 17 sessions (i.e., 85% of the overall training sessions).
Behavioral measures
Standardized (scale and standard) scores were calculated for each reading and EF measure and were used to evaluate the effect of the intervention for each of the reading groups. Different test versions were used for the pre/post-testing when available (TOWRE, GORT). The test administration order was randomized between the pre- and post-intervention testing sessions. An effort was made to keep a maximum of 2 weeks difference between the pre-testing session and the beginning of the intervention and from the end of the intervention to the post-testing session.
Reading measures included: (1) word reading accuracy/orthography, or the ability to pronounce printed words, was assessed with the Test of Word Reading Efficiency (TOWRE), sight word efficiency (SWE) subtest (Torgesen, Rashotte et al., Reference Torgesen, Rashotte and Wagner1999), standard score; (2) phonological processing was assessed with the pseudoword decoding efficiency subtest from the TOWRE (Torgesen, Rashotte et al., Reference Torgesen, Rashotte and Wagner1999), standard score; and (3) contextual oral reading fluency, rate, and comprehension were assessed with theGray Oral Reading Tests, 3rd Edition (GORT-III), scaled score (Wiederholt and Bryant, Reference Wiederholt and Bryant1992).
EF measures included: (1) inhibition: the Color-Word Interference subtest (condition 3) from the Delis–Kaplan Executive Functions System (D-KEFS) (Delis, Kaplan et al., Reference Delis, Kaplan and Kramer2001), scaled score; (2) working memory: the digit span subtest from the Wechsler Intelligence Scale for Children, 5th Edition (WISC-V), scaled score (Wechsler, Reference Wechsler2014); (3) switching: Sky Search DT from the Test of Everyday Attention (TEA-CH) (Manly, Robertson et al., Reference Manly, Robertson, Anderson and Nimmo-Smith1999), scaled score; (4) speed of processing (SOP), nonverbal: coding subtest from the WISC-V, scaled score, and verbal: naming letters subtest from the Comprehensive Test of Phonological Processing (CTOPP), scaled score (Torgesen, Rashotte et al., Reference Torgesen, Rashotte and Wagner1999).
Neuroimaging measures
The scan took place in Cincinnati Children’s Hospital using a 3T Philips Achieva MRI scanner (Philips Medical Systems, Best, the Netherlands). Participant’s head movements were managed by stretchable bands fastened to both ends of the head-coil device employed during the scanning process. An MRI-compatible audio/visual system was used to display a cross-hair during the resting-state scan (Avotec, SS3150/SS7100).
During the functional MRI scans, a T2*-weighted blood oxygen level-dependent (BOLD) technique was used, with a gradient echo planar imaging sequence (EPI) with the following specifications: TR/TE = 2000/35 ms; FOV = 24 × 14.8 × 24 cm; matrix size = 64 × 64; slice thickness = 4 mm resulting in a voxel size = 3.75 × 3.75 × 4 mm3. During the resting-state scan, the entire cerebrum was covered by 37 acquired axial slices. The fMRI scanning process yielded 183 image volumes, with a total acquisition time of 5.5 min.
Additionally, a 3D T1-weighted inversion recovery gradient echo whole-brain scan was obtained for each participant. This scan was used for anatomical co-registration and spatial normalization of the fMRI data with the following parameters: TR/TE = 8.05/3.68 ms; FOV = 22.4 × 25.6 × 19.2 cm; matrix = 256 × 256; slice thickness = 1 mm and flip angle = 8°.
Resting-state condition
In order to evaluate the effect of the EF-based reading intervention on the FC of EF brain networks, the resting-state condition was used before and after training. Children were instructed to look at a gray ”+” positioned in the center of a projector screen and were asked to avoid closing their eyes for the duration of the scan.
fMRI data preprocessing
After reconstructing the fMRI data, data were preprocessed using the default pipeline in CONN Version 22a (Whitfield-Gabrieli and Nieto-Castanon, Reference Whitfield-Gabrieli and Nieto-Castanon2012). The preprocessing included segmentation into gray matter, white matter, and cerebral fluid tissue classes, followed by slice-timing correction, realignment, co-registration, and normalization to the MNI (Montreal Neurological Institute) standard space. After the normalization transform was obtained, it was applied to the functional data, which underwent resampling to 3mm3 voxels and smoothing using the default 8 mm full width at half maximum Gaussian kernel (Horowitz-Kraus, Hershey et al., Reference Horowitz-Kraus, Hershey, Kay and DiFrancesco2019; Freedman, Zivan et al., Reference Freedman, Zivan, Farah and Horowitz-Kraus2020). Frame displacement cleaning and censoring were performed with a threshold of FD = 0.03 mm. No participants were removed due to excessive movement, as the average and standard deviation for the FD in each group did not differ significantly (see Table 1).
Baseline differences in demographic, clinical, reading, and EF behavioral measures in RD-only, RD + ADHD, and TR groups
Note. RD = reading difficulties; ADHD = attention-deficit/hyperactivity disorder; TR = typical readers; M = mean; SD = standard deviation; ScS = scaled score; SS = standard score; SWE = sight word efficiency; PDE = phonemic decoding efficiency; GORT = Gray Oral Reading Test; TOWRE = Test of Word Reading Efficiency; CTOPP = Comprehensive Test of Phonological Processing; WISC = Wechsler Intelligence Scale for Children.
Identifying person-specific network activity was achieved using a predetermined Power Atlas template (Power, Cohen et al., Reference Power, Cohen, Nelson, Wig, Barnes, Church, Vogel, Laumann, Miezin, Schlaggar and Petersen2011). This atlas divides the brain into 264 regions of interest (ROIs), organized into various functional networks, and it has been used to examine FC between different brain regions. The ROIs associated with cognitive networks and EF (cingulo-opercular [CO], frontoparietal [FP], visual attention [VAN], ad dorsal attention [DAN]) were defined using this atlas (Power, Cohen et al., Reference Power, Cohen, Nelson, Wig, Barnes, Church, Vogel, Laumann, Miezin, Schlaggar and Petersen2011). FC was defined as the average time series for each of the ROIs in the network. Average FC within and between these networks was calculated for each group, before and after the intervention. See Figure 2 for the selected networks.
Spatial maps for the cognitive control networks. The upper images display top-down network maps: cingulo-opercular (CO) in blue and frontoparietal (FP) in pink. The lower part displays bottom-up network maps: ventral attention (VAN) in red and dorsal attention (DAN) in green. The images are shown in terms of neurological orientation, with “L” indicating left and “R” indicating right.
Data analyses
To compare the effect of the intervention on reading and EF abilities, several 3 Group (TR, RD-only, RD + ADHD) × 2 Test (Test 1, Test 2) repeated measures (RM) ANOVAs were conducted for the five reading measures and five EF measures. Analyses were corrected for multiple comparisons (five tests within the reading/EF domain) using a Bonferroni correction.
To compare the effect of the intervention on FC, several 3 Group (TR, RD-only, RD + ADHD) × 2 Test (Test 1, Test 2) RM ANOVAs were conducted for the four within-network FC values and the six between-networks pairs, resulting in 10 FC scores overall (see Table 3). Analyses were corrected for multiple comparisons (four or six tests for within- and between-network connectivity, respectively) using a Bonferroni correction.
Results
Behavioral results
Baseline differences between the groups in demographic, clinical, reading, and EF measures
Chi-squared tests and ANOVA results (Table 1) comparing demographic, clinical, reading, and EF measures before the intervention (i.e., baseline) revealed that the groups did not differ in age, sex, IQ, or framewise displacement (index of head motion during the scan). The RD + ADHD group had higher symptom scores on the Conners scale and was the only group in which children were taking stimulant medication at the time of the study and continued taking it throughout the study period. On reading measures, the TR group consistently scored significantly higher than the RD-only and RD + ADHD groups, which did not significantly differ from each other. On measures of EF, the TR group also scored significantly higher than the RD-only and RD + ADHD group on tests of inhibition (D-KEFS Color-Word Interference) and verbal SOP (CTOPP, naming letter speed), whereas there were no significant group differences on tests of switching (TEACH, Sky Search DT), verbal working memory (WISC digit span), and nonverbal SOP (WISC Coding). See Table 1 and Figure 3 (T1).
Results of the RM ANOVA for reading (top) and EF (bottom) behavioral measures from pre- to post-intervention in RD + ADHD, RD-only, and TR groups. The y-axis represents the mean scaled/standard score results. Standard deviations are noted. Asterisks indicate that the Group*Test interaction is significant between-group differences at T1 or significant within-group change from T1 to T2: (**p < .01, *p < .05).
The effect of the EF-based reading intervention
Reading measures: RM ANOVA results (see Table 2 and Figure 3) revealed a main effect of group for reading fluency (p < .001), rate (p < .001), comprehension (p < .001), orthographical processing (p < .001), and phonological processing (p < .001) abilities, with significantly higher reading performance among the TR group relative to both RD groups and no difference between RD-only and RD + ADHD. A significant main effect of test was found for reading fluency (p = .002), rate (p = .016), comprehension (p < .001), and orthographical processing (p = .013), indicating overall improvements from Test 1 to Test 2. A Group × Test interaction was found for reading fluency (p = .015), although this did not survive correction for multiple comparisons, such that the TR group showed significant improvement following the intervention (p < .001), but the RD-only (p = .865) and RD + ADHD (p = .104) groups did not. Although there was no Group × Test interaction for reading comprehension (p = .271), there was a significant improvement in reading comprehension in the TR group (p = .041) and the RD + ADHD group (p = .001), but not in the RD-only group (p = .329).
Results of the RM ANOVA for reading and EF behavioral measures from pre- to post-intervention in RD + ADHD, RD-only, and TR groups
Note. RD = reading difficulties; ADHD = attention-deficit/hyperactivity disorder; TR = typical readers; M = mean; SD = standard deviation; ScS=scaled score; SS = standard score; TOWRE = Test of Word Reading Efficiency; SWE = sight word efficiency; PDE = phonemic decoding efficiency; GORT = Gray Oral Reading Test; CTOPP = Comprehensive Test of Phonological Processing; WISC = Wechsler Intelligence Scale for Children; DKEF = Delis–Kaplan Executive Functions; TEACH = Test of Everyday Attention for Children.
***p < .001, **p<0.01, *p < .05. Results in bold font survived the correction for multiple comparisons within a statistical effect and domain (reading/EF; 5 tests; p < .01).
EF measures: RM ANOVA results (see Table 2 and Figure 3) revealed a main effect of group for inhibition (p = .034), verbal working memory (p = .001), and verbal SOP (p < .001). Overall, EF scores were higher in the TR group compared to RD + ADHD and RD-only groups, with no difference between the RD groups. A main effect of test was found for inhibition (p = .001), although a marginal Group × Test interaction (p = .081) suggests this was driven by the RD-only (p = .002) and RD + ADHD groups (p = .017), with no effect in the TR group (p = .860). A main effect of test was also found for verbal working memory (p = .028), although a Group × Test interaction (p = .028, did not survive correction for multiple comparisons) suggests this was driven by the TR group (p < .001), with no effect of test among the RD-only (p = .905) or RD + ADHD groups (p = .673). Finally, a main effect of test was also found for verbal SOP (p = .009), but this was qualified by a Group × Test interaction (p = .008, did not survive correction for multiple comparisons), such that the RD-only group showed significant improvement (p < .001), but the TR (p = .645) and RD + ADHD (p = .999) groups did not.
Neuroimaging results
Baseline differences in functional connectivity between the groups
No significant differences in FC within or between networks were found between the groups before intervention (i.e., at baseline; see Table 3 and Figure 4).
Functional MRI matrices for Test 1 and Test 2 functional connectivity of cognitive control and attention networks. Functional MRI matrices for Test 1 (left column) and Test 2 (right column) functional connectivity of cognitive control (FP, frontoparietal; CO, cingulo-opercular) and attention networks (DAN, dorsal attention network; VAN, ventral attention network) for RD + ADHD (top row), RD (middle row) and TR (lower row). Hot color represents a higher correlation coefficient value, and cold color represents a lower correlation coefficient value (scale is noted to the right).
Functional connectivity measures within and between EF and attention networks before and after the intervention in children with RD-only, ADHD + RD, and typical readers
Note. RD = reading difficulties; ADHD = attention-deficit/hyperactivity disorder; TR = typical readers; M = mean; SD = standard deviation; CO = cingulo-opercular; FP = frontoparietal; VAN = ventral attention network; DAN = dorsal attention network.
***p < .001, **p < .01, *p < .05. Results in bold font survived the correction for multiple comparisons within a statistical effect and measurement type (4 tests, p < .0125; or 6 tests, p < .008).
The effect of training on within and between functional connectivity of the EF and attention networks
The RM ANOVA revealed no main effects of group. However, the main effects of test were observed for FC within VAN (p = .015) and between CO-FP (p = .006), VAN-DAN (p = .028), and FP-DAN (p = .010). Overall, between-networkFC values decreased from pre- to post-intervention, whereas within-VAN connectivity increased. Although there were no significant Group × Test interactions, exploratory effects of test within subgroups revealed significant decreases in the RD + ADHD group for CO-FP (p = .001) and FP-DAN (p = .004) connectivity and in the TR group for VAN-DAN connectivity (p = .025). See Table 3 and Figures 4 and 5.
Changes in within and between networks funcitonal connectivity following intervention.
Discussion
The goal of this study was to determine whether children with RD and comorbid ADHD, characterized by greater behavioral attention and EF challenges, show a differential response to an EF-based reading intervention (READ) on behavioral and neurobiological correlates (resting-state fMRI) of EF relative to children with RD with ADHD (RD-only) and without reading or EF challenges (TRs). In line with our hypotheses, children with RD + ADHD exhibited greater gains in reading comprehension and larger changes in FC between EF networks following the EF-based reading intervention than did children with RD-only. While these results align with theoretical frameworks such as the extended SVR model (Kim, Reference Kim2020) and the neural noise theory (Hancock Reference Hancock, Pugh and Hoeft2017), we were underpowered to detect significant interaction effects. Future studies with larger samples are needed to validate these findings and further explore the complex interplay between EF, brain connectivity, and reading development in children with both RD and ADHD. This research is critical to inform our understanding of individual differences in response to intervention and the role of EF in reading toward the goal of precision education.
Stimulating EF during reading training
In the past decade, several studies have advocated for the role of EF in the reading process (Kim, Reference Kim2020) and especially in reading fluency (Kieffer, Vukovic et al., Reference Kieffer, Vukovic and Berry2013; Church, Cirino et al., Reference Church, Cirino, Miciak, Juranek, Vaughn and Fletcher2019; Kieffer and Christodoulou, Reference Kieffer and Christodoulou2020; Taran, Farah et al., Reference Taran, Farah, DiFrancesco, Altaye, Vannest, Holland, Rosch, Schlaggar and Horowitz-Kraus2022; Horowitz-Kraus, Rosch et al., Reference Horowitz-Kraus, Rosch, Fotang, Mostofsky, Schlaggar, Pekar, Taran and Farah2023; Horowitz-Kraus, Reference Horowitz-Kraus2023; Taran, Farah et al., Reference Taran, Farah, Gashri, Gitman, Rosch, Schlaggar and Horowitz-Kraus2023). These studies focused on the involvement of EF in reading among those with RD, pointing at the EF challenges as another contributor to their lack of reading fluency. Theoretically, it was suggested that one of the causes of RD is increased neuronal noise, specifically in systems related to visual and auditory processing (Hancock Reference Hancock, Pugh and Hoeft2017). We have recently demonstrated that during the READ training, which includes a manipulation that encourages fluent reading (i.e., when the letters are deleted from the screen similarly to the manipulation suggested in the current intervention), greater synchronization between the visual and auditory networks is observed in those with RD (Horowitz-Kraus, Rosch et al., Reference Horowitz-Kraus, Rosch, Fotang, Mostofsky, Schlaggar, Pekar, Taran and Farah2023).
One of our previous assumptions was that this manipulation also triggers EF during training and hence engages EF to help synchronize these auditory and visual networks to reduce the neural noise (suggested in [Cecil, Brunst et al., Reference Cecil, Brunst and Horowitz-Kraus2021]). This was also supported by the greater functional connections between regions in the visual cortex and the CO network and DAN among children with RD-only during a reading fluency task following training (Taran, Farah et al., Reference Taran, Farah, Gashri, Gitman, Rosch, Schlaggar and Horowitz-Kraus2023). This group of readers also showed greater network efficiency (i.e., graph theory measures) within the CO networks during the resting state in relation to word reading gains (Horowitz-Kraus Reference Horowitz-Kraus, Toro-Serey, DiFrancesco and Ward2015b). Although the assumption regarding the connection between improved reading following training and greater visual and auditory networks in relation to the EF and attention networks has not validated yet, the current study results support the involvement of EF and attention networks in reading improvement among groups of children with varying levels of EF abilities, which warrants consideration when discussing linguistic and cognitive components supporting reading, such as the SVR.
Although RD and ADHD are highly comorbid disorders, both associated with EF difficulties to varying degrees (Rucklidge and Tannock, Reference Rucklidge and Tannock2002; Katz, Brown et al., Reference Katz, Brown, Roth and Beers2011), children with RD + ADHD showed greater gains in reading comprehension following the READ training. It is important to note that while children with both RD and ADHD are thought to have greater challenges in contextual reading fluency, rate, and comprehension, in addition to lower cognitive abilities associated with reading (such as switching and inhibition), our RD + ADHD group did not differ from the RD-only group in their baseline reading and EF performance. The absence of baseline differences in EF and attention FC patterns in the RD + ADHD group was unexpected given the established literature suggesting more pronounced EF deficits in children with comorbid RD and ADHD. This discrepancy might suggest that the sample size was insufficient to capture the expected differences due to the heterogeneous nature of EF deficits in ADHD or that the measures used may not have been sensitive enough to detect these variations. Additionally, this may be due to participants taking their prescribed stimulant medication during the neuropsychological testing and when doing the READ training, given the known positive effects of stimulants on cognitive and academic functioning in children with ADHD (Hawk, Fosco et al., Reference Hawk, Fosco, Colder, Waxmonsky, Pelham and Rosch2018). In future studies with larger samples, comparisons between children with ADHD not taking stimulant medication versus those that are will be important to understand whether and how treatment with stimulant medication may impact outcomes from an EF-based reading training.
The results of the current study generally align with previous findings in children with RD + ADHD and RD-only showing distinct changes following the EF-based reading intervention in their functional networks while performing a word-reading task, although we were underpowered to detect significant Group × Test interaction effects. Previous studies have reported increased functional connections between EFs and visual processing-independent components among RD + ADHD children, whereas RD-only children showed enhanced increased functional connections between attention-semantic, attention-visual processing, DAN-visual processing networks (Horowitz-Kraus, Hershey et al., Reference Horowitz-Kraus, Hershey, Kay and DiFrancesco2019). However, that study focused on a reading task where children had to engage neural circuits associated with reading, which might be the reason for the enhanced change found among those with RD-only relative to our current findings in a non-task resting state. It seems that for children with RD + ADHD, even training of 4 weeks (instead of 8 weeks, as observed in (Reference Cecil, Brunst and Horowitz-Kraus16, Taran, Farah et al., Reference Taran, Farah, Gashri, Gitman, Rosch, Schlaggar and Horowitz-Kraus2023) is a sufficient time to observe changes in EF and attention-related networks, as found in the current study.
In line with the neural noise theory, we suggest that the EF bottleneck among children with varying levels of EF is released following training with the EF-based reading intervention. By releasing this bottleneck, we postulate that more automatic visual-auditory integration occurs, allowing more fluent reading (suggested in (Horowitz-Kraus, Rosch et al., Reference Horowitz-Kraus, Rosch, Fotang, Mostofsky, Schlaggar, Pekar, Taran and Farah2023)). Here, we found that the effect is more pronounced after 4 weeks in those suffering from more significant EF and reading challenges (RD + ADHD). Further studies on audio-visual integration processes during reading tasks should be conducted to verify this point.
Study’s limitations
The findings of this study should take into consideration the following limitations. First and foremost, the current study does not include a control group of non-trained individuals (or individuals who train on another control program). It might be that the results received in the current study resulted from a practice effect, and hence the specificity of the effect of this EF-based reading training should be further determined using a control group (or reading without manipulation, for example). Second, this study included 55 participants, which allowed sufficient power to demonstrate the main effects, but insufficient power to detect differential responses to the intervention among the reading groups and may limit the generalization of these results. Moreover, the current study employed an ROI-to-ROI analysis. Performing the analysis in a higher resolution, such as voxel-to-voxel, can lead to more precise results and a better understanding of the intervention’s effect on FC. The length of resting-state data was 5.5 min, which may affect the strength of the analysis. A longer data acquisition should be considered in the future. We also believe that 4-week intervention is not sufficient for children with RD-only as compared to 8-week intervention, as we did find a more dramatic effect in those children after 8 weeks (Cecil et al., Reference Cecil, Brunst and Horowitz-Kraus2001; Taran, Farah et al., Reference Taran, Farah, Gashri, Gitman, Rosch, Schlaggar and Horowitz-Kraus2023), and hence the effect on neural circuits supporting EF and attention in children with gradual levels of EF and RD should be also examined after 8 weeks of training. Finally, the current study focused on EF and attention networks only and to ensure an effect of sensory networks as well as suggested in the current model, a full-scale study including both EF, attention, and visual-auditory networks should be included.
Conclusions
Our study demonstrated that children with varying levels of EF difficulties (children with RD + ADHD, RD-only, and TRs) showed improvements in reading and EF performance in response to an EF-based reading intervention, with some preliminary evidence of differential response in brain and behavioral performance across groups. These findings add to a growing literature demonstrating the important role of EF in the reading process, which should be taken into account for future interventions and clinical diagnoses. Lastly, the results also suggest a mechanism for its involvement when considering the continuum of EF and reading abilities across groups in the framework of the neural noise hypothesis in the context of reading.
Acknowledgements
Funding statement
This study was supported by the National Institute of Child Health and Human Development (R01 HD086011; PI: Horowitz-Kraus).
Competing interests
The authors have no conflict of interest to declare.
Open access
