Literacy Progress in Children with Dyslexia and the Role of Attention

Abstract

The main focus of this study was the role of attention in children with dyslexia (n = 137), who participated in a literacy-remediation program. Prior to and during four assessment moments (pretest and follow-ups 1–3), the attentional skill was assessed using the Attention Concentration Test (ACT). The ACT provides two measures for attentional skill, namely, working speed and distraction time. Working speed, unlike distraction time, was related to word and text decoding, and spelling (i.e., literacy level). Children who had high scores on working speed were more likely to perform better on all three literacy measures. Thus, working speed is related to literacy level. Literacy progress, the improvement due to the remediation was not affected by attentional skill. An unexpected finding was that a substantial number of children with dyslexia were unable to complete the ACT task without making errors. The same task administered to a younger group of readers without dyslexia proved that the task was not too difficult for this age sample. Attentional skill, albeit related to literacy skill, had no differential effect on remediation. All children profited equally.

Introduction

Developmental dyslexia refers to severe problems with the acquisition of reading and spelling (Vellutino et al., 2004). Depending on the source about 3–15% of children in primary education are facing problems with the acquisition of literacy skills (Pennington, 2009; Shaywitz et al., 1992; Snowling, 2000). Although scientific research combined with experience from the educational practice have led to best practices for early literacy intervention when reading and spelling skills are lagging behind (e.g., Griffiths & Stuart, 2013; Torgesen, 2005), for a subgroup of children with severe difficulties these interventions do not suffice. Because limited development of reading and spelling skills also causes unequal opportunities in education and society, it is paramount that science investigates the reason(s) for the non-responsiveness to literacy interventions.

The majority of remediation programs focus on strengthening phonological skills. After all, the most widely acknowledged explanatory factor for dyslexia is a phonological deficit (e.g., Torgesen, 2005). Phonological problems, however, only explain a limited amount of unique variance (16.1% according to a meta-analytic review by Melby-Lervåg et al., 2012), indicating that other factors also contribute to the problems that children with dyslexia face.

One factor is the capacity to pay attention. A lack of attentional control may be detrimental in two ways. One, reading problems are the result of another problem, such as ADHD. Precious tutoring time during reading instruction may get lost due to behavior management (Al Otaiba & Fuchs, 2006; Torgesen, 2005). The fact that reading problems and ADHD share a substantial amount of explained variance substantiates this assumption; 25–40% of children with ADHD meet criteria for reading disabilities and 15–40% of children with reading disabilities also meet criteria for ADHD (August & Garfinkel, 1990; Dykman & Ackerman, 1991; Maughan & Carroll, 2006; Pennington, 2009; Verhoef et al., 2019).

Second, reading problems may be affected by a lack of attentional control due to limitations of cognitive information processing. So-called capacity-limitation theories, which originate from Broadbent’s information processing theory, conceptualize attention as a limited and selective cognitive process (see Mole, 2017 for an extensive treatment of the history and development of the concept of attention in psychology; Scerif & Wu, 2014). Deficiencies in this capacity may hamper cognitive processing. Since reading is a quintessential cognitive process, it may be worthwhile researching whether children with reading problems are suffering from an attentional-processing deficit and to what extent this may hamper remediation.

Attention in reading and spelling

Learning to read requires learning a sequence of steps: Separation of spoken words into phonemes, identification of letters, learning rules that represent specific spoken words in print, gathering vocabulary, and assigning meaning to written language (Shaywitz & Shaywitz, 2008). Learning to read also requires attention (Laberge & Samuels, 1974). Initially, each subskill to be learned requires a substantial amount of attention, which decreases with increasing mastery. Each subskill sets the stage for entering the next phase in the reading process. The same holds for mastering spelling. Despite the fact that skilled reading occurs in a relatively automatic fashion, a certain amount of attention to the sequence of letters in the word and each letter or letter cluster is required (Reynolds & Besner, 2006). Decoding written words requires phonemes to be kept in sequential order and to be temporarily stored. Attention is related to the process of learning to read as well as to the process of decoding text in skilled readers. Given the notion that decoding skills of children with dyslexia lag behind those of their peers, it seems plausible that attention plays a role in mastering reading and spelling.

Reading requires processing a certain amount of orthographic information. Attention is usually limited to the current word and potentially the two neighboring words (Rayner, 2009). At the same time, other mental processes may compete with attention spent on the upcoming information, affecting the reading speed and reading accuracy, which appears particularly detrimental in less skilled readers (Leininger & Rayner, 2015).

Attention and inhibition theory

Studying attention or attentional control has a long-standing tradition that started with Hylan’s (1898) so-called continuous-performance tasks (CPTs). A CPT consists of a series of relatively easy items (of similar type) which the subject has to act upon with a simple and fixed response as quickly as possible. Well-known examples are canceling out all letters A amongst a series of different letters (e.g., Uttl & Pilkenton-Taylor, 2001), marking figures with 4 dots only amongst a series of figures that contain 3, 4, or 5 dots, as in the Bourdon–Vos test (Vos, 1992), and click the red squares amongst a set of five different colored squares (e.g., Smit & van der Ven, 1995).

Van der Ven and colleagues (Hotulainen et al., 2014; Pieters & van der Ven, 1982; Smit & van der Ven, 1995; van der Ven, 2019; van der Ven et al., 1989; van der Ven et al., 2005, 2017) developed the inhibition theory, an attempt to describe attention from a mathematical perspective. To understand the analysis of time series based on inhibition theory (see Method section), we describe three fundamental assumptions of inhibition theory proposed by Smit and van der Ven (1995, p. 268):

1. During each mental activity, for which a minimal amount of attention is required, very short, alternating periods of attention and distraction emerge. During periods of attention (or work), the person is really involved in the task and during distraction periods he is distracted and not busy with the task at hand (periods of non-work). It is also assumed that periods of work are constant (i.e., consecutive working time periods are equal), and the amount of working time for each subtask is the same.

2. Inhibition resembles a kind of resistance to work. It increases during periods of work and decreases during periods of non-work (distraction).

3. Inhibition fluctuates during periods of work and periods of non-work. When inhibition increases during a period of work, the tendency to fall into a distraction increases, whereas a decrease of inhibition during a period of non-work increases the tendency to return to an attention period.

These assumptions lead to two criteria that should be met by tasks used to measure concentration. The first is that the task consists of a series of items with the same difficulty and the response times of each item must be registered. The length of the task is such that participants reach stationarity because reaction times tend to increase at the beginning of the task and then become stationary around a certain value. The second requirement is that the task is relatively easy and responses should be overlearned because attentional skill is considered in terms of the maximum potential of an individual.

The present study

Despite the fact that attention is a prerequisite for all academic skills, research on the role of attentional skill during reading, and more specifically in children with dyslexia, is scarce. An attention measure should not require skills such as knowledge or experience; a simple task is therefore needed (van der Ven, 2001). For that reason, only tests that contain easy and highly overlearned items, that also generate quantitative measures of attention, should be used. One such task is the attention-concentration task (ACT) developed by van der Ven and colleagues.

The main goal of this study was twofold. One, establish whether a relationship between attentional skill and reading and spelling level exists. That is, does lower attentional skill imply poor reading and spelling? Two, establish whether a relationship between attentional skill and reading and spelling progress exists. In other words, is attentional skill predictive of who will benefit from a remediation program?

Method

Participants

Participants were recruited from Braams & Partners, a Dutch clinic for the assessment and remediation of learning disorders. Only participants who met the criteria of severe dyslexia were included (criteria are stated below). A group of 137 students (45 girls and 92 boys) with a mean age of 9;1 years (ranged between 7;3 and 11;7 years). For most participants, scores on an intelligence test (WISC-III NL) were available. Intelligence quotients (available for 109 participants) varied between 79 and 131 (M = 103.6; SD = 12.5).

Participants were admitted according to a strict admission procedure prescribed by the so-called Protocol Dyslexie Diagnostiek en Behandeling [Protocol Dyslexia Diagnostics and Remediation] of Blomert (2006) and the Stichting Dyslexie Nederland (Dutch Dyslexia Foundation). This protocol has been accorded by the Dutch College of Health Insurance (College van Zorgverzekeringen) and was presented to the minister of Public health, Welfare, and Sports (Volksgezondheid, Welzijn en Sport). The criteria of the Dyslexia Protocol are: (a) the participant attended or attends a Dutch primary school; (b) the participant’s results on the word-decoding test (i.e., Drie Minuten Toets, [Three Minute Test]; Jongen & Krom, 2009) are at the 10th percentile or participant’s results on word-decoding tests are at the 16th percentile and results on spelling tests are at the 10th percentile (Schaal Vorderingen Spellingvaardigheid [Scale for progress in spelling skills], see de Wijs et al., 2010); (c) the participant has received at least 8 weeks of intensive and qualitative remediation, either individually or in a small group in the school, without showing significant improvement in reading and spelling (see criterion b); (d) the participant does not suffer from another developmental disorder other than dyslexia.

The classroom teacher or remedial teacher at the school administered the reading and spelling tests and provided and evaluated the remediation program prescribed. Guidelines for the remediation program at school are accommodated for by the Protocol Leesproblemen en Dyslexie (Protocol Reading Difficulties and Dyslexia; Wentink & Verhoeven, 2003) and Protocol Leesproblemen en Dyslexie voor groep 5–8 (Protocol Reading Difficulties and Dyslexia for Grades 3–6; Wentink & Verhoeven, 2004). The main goal of these protocols is to achieve a functional level of technical reading (decoding) and spelling for the majority of students, based on learning in a meaningful context, learning by interaction with peers and teachers, learning of useful strategies, and extending effective reading time.

The tests administered at the clinic assessed whether criteria were met for severe dyslexia and thus for reading and spelling remediation at the clinic. The inclusion criteria were: (1) Intelligence quotient above 70 (based on recent WISC-III results); (2) participant’s results on word decoding at the 10th percentile or participant’s results on word decoding at the 16th percentile and results on spelling tests at the 10th percentile, (3) low scores (the 10th percentile) on phonological processing tasks (two out of six measures: rapid naming letters and digits; letter-sound identification accuracy and speed; and phoneme deletion accuracy and speed; see Blomert & Vaessen, 2009) or, when phonological processing scores were above the 10th percentile, low scores on memory tasks for phonological stimuli (e.g., sounds and syllables). A strict application of the criteria resulted in a relatively homogenous group. Pretests of intelligence, reading and spelling, phonological processing, memory, and attention were administered by certified school psychologists who worked at the clinic. Follow-up measures of reading and spelling were administered after approximately 3 (follow-up 1), 6 (follow-up 2), and 9 months (follow-up 3) of reading and spelling remediation.

Materials

Literacy tests

Word decoding was assessed with the standardized Three Minute Test [Drie Minuten Toets, DMT] of Jongen and Krom (2009). This test consists of three different cards containing single one-syllable words. Card 1 contains Vowel-Consonant (VC)- and CVC-words. Card 2 contains one-syllable CCV-, CCVC-, CVCC-, CCVCC-, CCCVC- and CVCCC-words. Card 3 contains multisyllabic words. Participants were asked to read as many words on each card as they could in 1 min. Test score was the sum of the number of words read accurately on all three cards. The publisher reported reliability of summed scores on Card 1, Card 2, and Card 3 with alpha’s varying between 0.94 and 0.97.

Text decoding was assessed with the standardized AVI toets pakket [AVI test package] of Visser et al. (1996); text 3A was used. Participants were asked to read aloud the text as fast as possible without making errors. Test score was the total amount of time in seconds required to finish the text. Reliability varied between 0.85 and 0.91 (Visser et al., 1996). Construct validity was sufficient to high; correlations with a word-decoding test was high (r = 0.85–0.90), but lower with tests measuring reading comprehension and vocabulary (r = 0.45–0.70; Visser, 1997).

Word spelling was assessed with the standardized PI-dictee [PI test] of Geelhoed and Reitsma (2004). Participants were asked to write down a list of orally presented words. The list consists of nine sets of 15 words each, increasing in difficulty. When 8 out of 15 words were spelled correctly, the next set was presented; the test was terminated when the participant spelled seven or less than seven words in a set correctly. The test score was the number of words spelled correctly. Reliability varied between 0.90 and 0.93.

Attention-Concentration Test (ACT)

The ACT is an experimental computer task (http://www.socsci.ru.nl/advdv/actengl.html; see van der Ven et al., 2017 for a detailed description). Participants were presented with 25 bars of colored buttons. Each bar consisted of 18 colored buttons (i.e., blue, orange, pink, yellow, green, or red) and each color was presented three times in each bar (see Figure 1 for an example). The bars were displayed one by one and the participants were asked to click all three red buttons in strict left to right order. After the participant had clicked the last red color of a bar, the next bar appeared. The participants were asked specifically to make no mistakes because one mistake resulted in the termination of the test. Instructions were provided verbally because of the participants’ reading problems. Participants also practiced ten bars before starting the actual test.

Figure 1. Example of the view of one of 25 rows displayed during the ACT.

The ACT computer-program saves the reaction times of each consecutive bar, consisting of the time it takes the participant to click three red buttons. According to inhibition theory (as discussed in the introduction), each individual reaction time consists of a series of alternating concentration and distraction phases. The concentration phases are in fact real working times or working speed and the distraction phases reflect non-working times. Working speed is estimated by the minimum response time based on 25 consecutive reaction times in one test. The total distraction time was estimated by taking the square root of the average square of the reaction times of bars 16–25.

Unlike van der Ven et al. (2017) the natural logarithm of the measures was computed to avoid skewed distribution of scores. This concerned both the square root of the Mean Square Residual of reaction times (distraction time) and the minimum response time based on 25 consecutive reaction times in one test (working speed). Note that the natural logarithms of values below 1 result in negative values. van der Ven et al. (2017) provide all required information for the calculation of the ACT-measures as well as reliability measures. Test–retest reliability coefficient for working speed (the natural logarithm of the square root of the Mean Square Residual of bar 16–25) was r = 0.52, and for distraction time (the logarithm of the minimum) r = 0.78. The test–retest reliability coefficients were based on two sessions of the ACT.

Reading and spelling remediation program

The present intervention was provided according to the recommendations of Blomert’s (2006) Protocol Dyslexie Diagnostiek en Behandeling [Protocol Dyslexia Diagnostics and Remediation]. This “best practice” remediation program is based on the premise that dyslexia is strongly associated with a deficit in phonological skills. It targets both reading and spelling skills and links the processing of speech sounds to the processing of letters and words. The main goal for the participants is achieving a so-called functional level of reading decoding and spelling.

Participants received direct instruction and practice once a week from a certified school psychologist at the clinic for about 45 min with one of the parents attending. Instruction and practice consisted of word reading, story reading (aloud), instruction of spelling categories, practicing word and sentence spelling, and coaching the parent assisting with homework assignments. Participants were asked to comply with homework assignments (i.e., reading and spelling exercises) three to four times a week for about 30 min under the guidance of their parents. Parents received instruction manuals to make sure that the homework assignments were conducted properly. The length of the remediation program varied between 12 and 18 months (note that for this study, children were followed during the first 9 months). 116 participants stayed in the program for at least 9 months.

Word-decoding skills were practiced using a computer-programmed digital flash-card program. To enhance text-reading skills, children were offered texts and books that matched their interests. In most cases, participants’ word-decoding skills were below the required level of the texts they wanted to read. To accommodate for this, the school psychologist read the text with the participant during weekly sessions. The parents were asked to do the same when reading at home. Strategies used were guided reading strategies, reading aloud together, repeated reading, or participant reading aloud.

Spelling skills were practiced by means of systematic instruction of spelling rules and regularities. The instruction level was adapted to the current spelling level of the participant and slowly increased during the four-module remediation program. The first module targeted grapheme-phoneme correspondences; grapheme–phoneme knowledge was trained using visual scaffolds. Each sound category (i.e, short vowels, long vowels, digraphs, trigraphs, tetragraphs, consonants, and schwas) was assigned a symbolic scaffold; note that the schwa was not included in the practice materials until the third module when multisyllabic words were introduced. Also, a mnemonic card was introduced with sound categories and symbolic scaffolds (see Figure 2). These symbolic scaffolds were used to show that sounds (in alphabetic writing systems, such as Dutch) correspond to the letters of the alphabet or a letter cluster. Although most participants should have already learned this at school, most participants needed direct instruction with scaffolds to gain a proper understanding and command of the Dutch phoneme-grapheme couplings.

Figure 2. Mnemonic card with sound categories and symbolic scaffolds used during reading and spelling remediation. Sound categories from top to bottom: short vowels, long vowels, digraphs, trigraphs, tetragraphs, consonants, and schwas. Copyright 2020 by Braams. Reprinted with permission.

The second module concerned the spelling categories of monosyllabic words. Spelling categories refer to a specific rule and its exceptions, such as homophone sounds (e.g., <ch/g> or <au/ou> in Dutch) and words that end in an ambiguous phoneme–grapheme relationship. For example, a “d” at the end of a word is pronounced <t>. In case, the plural of a noun is pronounced and written with a “d,” such as in “paarden” (horses), the singular is also written with a d, “paard.” During this phase, previous spelling categories, including those of the first module, were repeated to strengthen previously learned skills and strategies.

The third module pertained to the spelling of multisyllabic words. The most important and the most difficult rules to master are vowel reduction or degemination in words with lax vowels (e.g., singular “raam” [window] is spelled “ramen” in its plural version) and consonant doubling in polysyllabic words with tense vowels (e.g., the singular “ster” [star] is spelled “sterren”). After participants mastered all disyllabic spelling categories, they were confronted with three-and-more syllabic words, which require the application of more than one spelling category per word. Spelling categories from the first and second modules were also rehearsed.

The fourth module targeted the set of most common loan words in Dutch. Loan words or non-native Dutch words mainly originate from English, French, and German. In most cases, the spelling of the original language is preserved, which causes severe deviations from the prototypical Dutch phoneme-grapheme correspondences. Although training of these categories relies partly on drill-and-practice strategies, direct instruction and training were based on the orthographic structure to strengthen participant’s awareness of regularities in orthography (e.g., in Dutch the morpheme <-ge> in “garage” and “etalage” stems from the French language). The final spelling module was only presented to participants who were eligible or who mastered all categories from earlier modules. Spelling categories taught in previous modules were also rehearsed along with the spelling of the loan words.

Procedure

Participants were referred for diagnosis and remediation by their schools according to the standardized Dutch Protocol Dyslexia Diagnosis and Treatment [Protocol Dyslexia Diagnostics and Remediation] (Blomert, 2006). Upon referral by the school and after the school declared that they had provided intensive remedial help for at least 8 weeks prior to the intake, parents requested the clinic for diagnosis and if necessary a remediation program. On intake, all literacy and the attention-concentration tests required for this study were conducted.

Participants were referred and diagnosed in 2011, 2012, or 2013. If participants met the inclusion criteria, parents were informed about the scientific research conducted at the clinic before the child started the reading and spelling remediation program. All parents of the children who participated in this study gave their permission for the use of their child’s test scores; privacy was guaranteed. The data of those who did not wish to join were removed. Less than 5% of selected participants refused to participate.¹

Results

Before attempting to answer the research questions, we need to address a rather unexpected finding. The pretest session revealed that a substantial number of our participants appeared unable to perform the attention-concentration test correctly within three attempts (88 of the 137 participants; 64%). As said, according to inhibition theory, every person who understands the task should be able to finish 25 consecutive bars of the ACT flawlessly; which is required for a valid interpretation of the ACT-scores.

This surprising and unanticipated result forced us to conduct a parallel study to investigate whether students with dyslexia need to practice the ACT more thoroughly. Half of the participants were randomly assigned to either the ACT-training condition or the control condition (in which no ACT-training was offered). The ACT-training group consisted of 62 participants (19 girls and 43 boys), aged 7;3 to 11;0 years. The control group consisted of 75 participants (26 girls and 49 boys), aged 7;5 to 11;7 years. On the pretest, the ACT-training and control group did not differ on outcome measures used for this study (t-tests resulted in p’s > 0.05 for Word decoding, Text decoding, and Word spelling). The experimental group received ACT training as part of the reading and spelling remediation program. In this training, the participant practiced the first ten bars of the ACT (see Materials section). The children practiced every other week during 9 months with a total of 18 practice sessions; six sessions between pretest and follow-up 1, six sessions between follow-up 1 and follow-up 2, and six sessions between follow-up 2 and follow-up 3. The control group did not receive any training on top of the reading and spelling remediation program. Participants and their parents were not aware of the differences between groups. Table 1 presents the success rates of participants of those who had practiced the ACT and those who had not.

Table 1. Number and percentages of participants who succeeded in completing 25 bars of the ACT within three attempts at pretest and follow-ups 1–3 for experimental and control group.

Condition	Pretest		Follow-up 1		Follow-up 2		Follow-up 3
	n	Succeeded	n	Succeeded	n	Succeeded	n	Succeeded
ACT-training	62	22 (35.4%)	60	26 (43.3%)	60	28 (46.7%)	52	38 (73.1%)
Control	75	27 (36.0%)	72	28 (38.9%)	62	43 (69.4%)	57	35 (61.4%)

After six training sessions at Follow-up 1, it seems that more students who were actually practicing succeeded to complete 25 bars of the ACT than those who had not. After 12 sessions, at Follow-up 2, the control group had more students finishing the ACT without errors, but the final measurement at Follow-up 3, revealed again that more ACT-training participants (11.7%) completed the ACT flawlessly than the control-group participants. Interestingly, the number of students in the control condition who completed the ACT without errors increased from 36% at pretest to 61.4% at Follow-up 3; the numbers in the ACT-training group increased from 35.4% to 73.1%. Despite extensive training, a substantial number of children did not achieve a flawless score at the end of the ACT training. Note, however, each child managed to complete at least one ACT session flawlessly during the research period.² These data could thus and were used in the analyses to test the research questions.

After completion of the experiment, the assumption that children this age should be able to complete the ACT task quite easily was tested. A group of 34 children without dyslexia, with a mean age of 7;10 years (ranged between 6;8 and 9;6 years) was asked to perform the ACT task. All children managed to complete the task flawlessly at the first attempt, except for one child who needed a second one. The average age of this group of children was considerably younger (1;3 years) than that of the group with dyslexia.

ACT performance and literacy level

True attention is estimated by the best possible performance on the ACT, represented by the lowest scores on working speed and distraction time during the 9 months of the study. Pearson correlation coefficients were calculated to explore the extent to which ACT performance was related to literacy skills measured at pretest and Follow-ups 1 through 3; see Table 2.

Table 2. Pearson correlations and corrected Pearson correlations between ACT measures and literacy skills at pretest and follow-ups 1–3.

		n	Working speed			Distraction time
			r	r′	p	r	r′	p
Pretest	Word decoding	111	−0.25	−0.35	<0.01	−0.12	−0.13	0.17
	Text decoding	93	0.27	0.38	<0.01	0.06	0.06	0.57
	Word spelling	133	−0.29	−0.40	<0.01	−0.13	−0.14	0.11
Follow-up 1	Word decoding	124	−0.18	−0.25	<0.01	−0.07	−0.08	0.38
	Text decoding	126	0.15	0.21	0.02	0.05	0.05	0.58
	Word spelling	133	−0.19	−0.26	<0.01	−0.04	−0.05	0.57
Follow-up 2	Word decoding	118	−0.24	−0.33	<0.01	−0.17	−0.19	0.04
	Text decoding	110	0.10	0.14	0.15	−0.001	−0.001	0.99
	Word spelling	128	−0.16	−0.22	0.01	0.06	0.07	0.43
Follow-up 3	Word decoding	103	−0.26	−0.36	<0.01	−0.22	−0.25	0.01
	Text Reading	92	0.14	0.20	0.06	0.05	0.05	0.64
	Word spelling	111	−0.19	−0.26	<0.01	−0.06	−0.06	0.53

Note. p-Values presented pertain to r′-values and are computed using https://www.socscistatistics.com/pvalues/pearsondistribution.aspx

Due to low test–retest reliability of the ACT measures (see materials section), Pearson correlation coefficients may have been underestimating the extent to which ACT measures relate to literacy measures. Therefore, correlation coefficients were corrected for attenuation by the formula, as suggested by Muchinsky (1996, p. 67): $${\rho }_{xy\mathrm{ }}=\frac{r_{xy}}{\sqrt{r_{xy}}}$$

Corrected Pearson correlation coefficients between best working speed times on the one hand and literacy levels on the other were generally significant, albeit small to moderate (Cohen, 1988). Absolute values varied between 0.14 and 0.40, with a median of 0.26. Corrected Pearson correlation coefficients between best distraction time and literacy level were generally not significant and tended to be small. Absolute values varied between 0.001 and 0.25, with a median of 0.065. Thus, superior working speed was associated with good word decoding, text decoding, and spelling, but distraction time was not.

ACT performance and literacy progress

To test for the effects of ACT performance, working speed and distraction times were used as covariates in the analyses on literacy progress during and after the remediation program with GLM repeated measures ANCOVA’s. Time was used as a within-subjects factor with four levels (pretest, follow-up 1, follow-up 2, and follow-up 3). Results for each literacy measure will be reported after age in months as well as best ACT working speed and distraction time of each participant were added as covariates. Note that, degrees of freedom were adapted due to the violation of the assumption of sphericity, using Greenhouse–Geisser (when ε < 0.75) or Huyn–Feldt (when ε > 0.75). Descriptive statistics are presented in Table 3.

Table 3. Descriptive statistics of literacy measures at pretest and follow-ups 1–3.

	Pretest			Follow-up 1			Follow-up 2			Follow-up 3
	n	M	SD	n	M	SD	n	M	SD	n	M	SD
Word decoding	113	110.50	45.72	125	147.30	47.80	118	163.16	50.34	104	175.14	51.27
Text decoding	94	118.54	52.19	127	88.70	41.42	112	74.95	27.17	93	68.72	22.34
Word spelling	135	42.41	21.56	135	65.90	19.85	130	77.99	20.16	113	87.18	19.79

Word decoding

The main effect of Time was significant after age was added as covariate, F(2.11, 170.64) = 12.83, p < 0.001, η²_p = 0.14. The main effect of Time remained significant after working speed was added as covariate, F(2.12, 169.76) = 10.99, p < 0.001, η²_p = 0.12, and also when subsequently distraction time was added as covariate, F(2.10, 167.81) = 13.36, p < 0.001, η²_p = 0.14.

Text decoding

The main effect of Time was significant after age was added as covariate, F(1.93, 108.18) = 13.96, p < 0.001, η²_p = 0.20. The main effect of Time remained significant after working speed was added as covariate, F(1.92, 105.84) = 6.73, p = 0.002, η²_p = 0.11, and also when subsequently distraction time was added as covariate, F(1.93, 106.13) = 14.25, p < 0.001, η²_p = 0.21.

Word spelling

The main effect of Time was significant after age was added as covariate, F(2.81, 302.96) = 13.52, p < 0.001, η² = 0.11. The main effect of Time remained significant after working speed was added as covariate, F(2.87, 300.82) = 2.79, p = 0.04, η² = 0.03, and also when subsequently distraction time was added as covariate, F(2.85, 299.16) = 10.21, p < 0.001, η² = 0.09.

To summarize, on average the participant group showed progress in all literacy skills during the course of the remediation program and this effect did not change after controlling for best ACT performance, indicating that ACT performance does not contribute to the effect of the reading and spelling remediation program.

Discussion

The main goal of this study was to investigate the potential role of attentional skill in the remediation process of children with dyslexia. Attention was operationalized by the ACT, which is based on inhibition theory. High levels of working speed and low levels of distraction time (resulting in low values on working speed and distraction time, respectively) assessed during a repetitive task served as an indicator of attentional skill. This study revealed three interesting findings pertaining to dyslexia and ACT completion, the relationship between ACT-performance measures and literacy levels, and the predictive value of the ACT for literacy progress.

ACT completion

The first one is important and unexpected. Unlike their younger peers, the majority of children with dyslexia had problems performing the ACT task without any errors. Although each of them had at least one attempt out of four in which they performed the task flawlessly, this achievement does not seem to come as easily as in children without dyslexia, who were also more than a year younger. All children were familiar and fully aware of the goal of the task. The fact that ACT practice did not improve performance much, suggests that a certain amount or type of concentration required to avoid mistakes is lacking in some of the children with dyslexia. Whether or not the reduced performance of the children with dyslexia is a signature of their problem is a question that requires more study. A study by Hotulainen et al. (2014) suggests that this issue may not be limited to children with specific learning difficulties. They tested a group of 358 Finnish children with an average age of 16 years and who were attending secondary education. In their sample, 15.6% of the participants (see Table 1 on p. 245) failed the ACT task. The version of the ACT they used was more complicated. Not only did they use dice rather than colors, but the number of targets per bar also varied randomly. We also need to emphasize that the participants in Hotulainen et al.’s study were only offered one attempt.

ACT performance and literacy level

ACT performance is based on two complementary measures derived from the ACT time series, that is, working speed and distraction time. The analyses revealed that all literacy level tests (i.e., word and text decoding, and spelling) were related to working speed, whereas distraction time was not. Children with higher levels of working speed are indicated to have a slightly increased chance of being the better readers and spellers within the group of children with severe reading and spelling problems. An important caveat is that the amount of explained variance never exceeded 16%, with an average level of 7%. Remember, this finding is based on correlational analyses and thus does not reveal a causal direction.

Working speed according to inhibition theory is the actual time a participant needs to complete the task. It is considered to be a pure measure of information-processing speed. According to inhibition theory, working speed is to a certain extent determined by the participant, whereas distractions are autonomous interruptions beyond the control of the individual. Word and text decoding tests are also measures of working speed. As such, at least the word and text decoding measures would be expected to be correlated with a measure of working speed. Note that, the relationship between working speed and spelling is not explained by the speed of information processing, because there is no time pressure with respect to spelling.

Hotulainen et al. (2014) studied the relationship between attentional skill and scientific reasoning/school achievement. They report that attentional skill explained 3% of the variance of scientific reasoning. After they grouped the participants in low, below average, above average, and high attentional skill, the analyses revealed that the high attention group outperformed the other groups on both scientific reasoning and school achievement. These findings indicate that the ACT may reveal learning potential in general. Scientific reasoning, school achievement, including reading and spelling appear to be related, albeit limited, to attentional skill. We would like to emphasize that no causal relationship can be ascribed to these findings.

The fact that researchers sometimes find significant, usually low, correlations between reading level and cognitive or linguistic skills/processes and sometimes they do not, is in itself an interesting phenomenon. It proves that there is not a (uni)causal relationship between some cognitive skills and reading and spelling performance. A better way of understanding these phenomena is in terms of a complex system. Reading and spelling problems are, like any other learning or psychological problem, the result of the interaction between numerous interacting components as described in complex systems theory (e.g., Wallot & Van Orden, 2011; Wijnants et al., 2012). The fact that so many different processes, functions, or components are to some extent related to reading and spelling level suggests that a large number of processes contribute to the development of a skill, such as literacy. After all, becoming literate is greatly subsidiary to speech and language and as such grounded in the dynamics of the body. For example, pronouncing a single utterance requires the coordination of at least 70 muscles (Turvey, 2007). Moreover, all readers constitute a unique constellation of their own history, development, and the task they are faced with. It is the mutual interdependence of the connections between all components rather than the components themselves that determine reading skills, and for that matter all skills.

ACT performance and literacy progress

None of the ACT-performance measures were related to literacy progress. Although word and text decoding skills, as well as their spelling performance, improved substantially during the 9-month remediation program, this was not affected by their attentional skill. In other words, irrespective of their performance on the ACT, all children appeared to have similar chances of improving their literacy skills. Our findings mimic those of Walda et al. (2014) on executive functions and Walda et al. (2019) on the null effects of working memory as predictive powers for reading and spelling remediation. Both studies showed small to absent relationships between cognitive functioning and reading and spelling level, but no relationship between those cognitive processes and reading and spelling progress.

Implications for education and remedial practice

The results of our study are clear about the relationship between attentional skill and reading level. Children with low reading and spelling levels are more likely to perform relatively poorly on an attention task. Since this relationship is not very strong, attention measures cannot be used to distinguish between poor and good readers. More important for practice is the fact that irrespective of attentional skill, all children benefited from the remediation program, which is well-structured and applies a phonologically based didactic. Earlier research reveals that explicit direct instruction with lots of practice opportunities prevents the development of reading and spelling problems in almost all children (see Bosman & Schraven, 2017; Compton et al., 2014; Griffiths & Stuart, 2013). Thus, instruction time and the hours at school should be devoted to reading and spelling activities rather than to the training of cognitive skills that have no bearing on literacy development.

Acknowledgments

This research was supported by Braams & Partners, Clinic for Learning Disorders, in Deventer, the Netherlands. Special appreciation is expressed to the children who participated, their parents, teachers, and the reading specialists of Braams & Partners. Lydia Weenk, Yvonn van de Grootheveen (Braams) and Maaike ten Buuren (Braams) contributed to supplemental data gathering of the experiment.

Additional information

Funding

This work was supported by the Dutch Research Council (NWO) under Grant [023.010.019].

Notes

1 Due to the clinical nature of the study, data were not always complete. First, participants sometimes happened to miss one or more assessment moments, which led to the situation that the number of participants varied across analyses. Second, visual analysis of scatter diagrams showed that four cases had substantial outliers on the ACT measures, which could only be explained by assuming that in those cases there was a temporary interruption of the subject's normal test setting; the data of these ACT-sessions were removed from the data file. Third, some log files were accidentally removed from the computers without there being a back-up. As a result, the number of participants that failed three consecutive ACT tests is unknown. To accommodate for this uncertainty, a supplemental study with a new group of participants meeting the same criteria was performed. The study revealed that 8 out of 23 participants (34.8%) were able to finish the ACT within three attempts (3 succeeded the first attempt, and 5 succeeded within two attempts). The percentage of participants unable to finish the ACT within three attempts in the pilot is similar to the percentage of participants for whom data of the ACT were available in the original experiment (49 out of 137, or 35.8%). It thus seems rather unlikely that the results are affected by loss of data files.

2 Only 4 children completed the ACT at all four attempts, 24 children completed three out of four attempts, 49 children completed 2 out of four, and 60 children completed only one attempt.