Fostering word fluency of struggling third graders from Germany through motivational peer-tutorial reading racetracks

Abstract

Automation of frequently used words is a key component in the development of reading fluency. However, acquiring fast word recognition skills is a serious challenge for many children in their early years of formal education. Lagging word recognition leads to general reading problems, as fluency is a vital prerequisite for text comprehension. Recent research shows that the percentage of struggling elementary school readers in Germany is increasing, speaking to the need for widespread implementation of effective word recognition interventions. This pilot study aims to provide preliminary evidence of the effectiveness of peer-tutorial reading racetrack training with an integrated motivational system for the sight word fluency of German struggling elementary school students. The intervention comprised twelve 15-min teaching units over a period of three weeks. To encourage reading motivation, the intervention included graphing of performance scores and a group contingency procedure. A control-experimental group design (N = 44) with pre-, post-, and two follow-up measurements (each after five weeks) was employed to investigate the impact of the treatment on decoding sight words at an appropriate speed. Results demonstrated a significant performance increase in the treatment group, relative to the control group. The effect size can be considered very high (partial η² = .76), indicating that this brief training has the potential to enhance the word recognition of struggling elementary students.

The importance of remediating reading problems

The ability to decode printed or written symbols to arrive at meaning plays a vital role in academic success and the development of the human mind (Aaron et al., 2008). Tragically, an ever-increasing number of students do not even attain basic reading comprehension skills (Cibulka & Cooper, 2017). Worldwide, about 10 to 15% of the population experience severe difficulties in understanding written text, despite having attained at least a basic level of education (Dyslexia International, 2017; Sprenger-Charolles & Siegel, 2016). Moreover, the percentages of struggling readers seem to increase. In Germany (where this study took place), the Progress in International Reading Literacy Study Survey (PIRLS) indicated that there was an upward trend of weak German fourth-grade readers from 2011 to 2016. The percentage of struggling students rose from 16.9% in 2001 to 18.9% in 2016. Furthermore, the PIRLS survey revealed a decrease of reading motivation, especially for struggling readers (Harju-Luukkainen et al., 2020). The negative consequences of severe reading difficulties are far-reaching. They lead to generally poor performance in school and often a dramatic loss of motivation in all core subjects. In the long run, they increase the risk of school dropout, unemployment, and poverty (Macdonald et al., 2016).

Typically, reading problems are apparent early in the developmental sequence. Students who struggle experience difficulty understanding the relationships between letters and sounds and have trouble decoding phonologically regular words. It takes them much longer than normally achieving peers to acquire a sufficient amount of sight vocabulary, which in turn limits their reading fluency (Kendeou et al., 2009). Sight words are immediately recognized without paying attention to individual sounds that simplify the process of finding the pronunciation and meaning of familiar words automatically (Ehri, 2005). To achieve advanced reading skills, the automatic recognition of words as sight words is of crucial importance (Balass et al., 2010). To support students in their reading development, it is vital to examine how reading skills develop, where exactly problems in reading acquisition may arise, and how to strengthen important skills.

The signification of word recognition skills for text comprehension

Our understanding of the importance of word recognition is informed by contemporary reading models. The dual route theory (DRT) and dual route cascaded model (DRC) of reading (Coltheart, 2005; Coltheart et al., 2001) include two cooperative systems: (a) a lexical route (orthographical decoding) and (b) a non-lexical route (phonological recoding). The lexical route refers to a mental dictionary that helps to recognize words by sight; the non-lexical route enables skilled readers to identify parts of words and connect them to single sounds to decode unfamiliar words. Skilled readers use the non-lexical only for less frequent or unknown words, but struggling readers use it all the time (Coltheart, 2005). Ehri (2005) also considered sight word reading and a lexical route in her stage-based model of reading, which consists of four phases of reading acquisition: (1) the pre-alphabetic, (b) the partially alphabetic, (c) the fully alphabetic, and (d) the consolidated alphabetic phase. Children in these stages can be distinguished by the extent to which they incorporate phonological awareness and knowledge of grapheme-phoneme correspondences into the building of memory connections between written words and their pronunciation and meaning. Mature readers have succeeded in building up a visual vocabulary that enables them to retrieve words quickly and unconsciously via the lexical route from their mental lexicon (Ehri, 2005). Students who have already reached this state show a greatly facilitated reading process (Morris & Perney, 2018). One similarity across these theories is that sight-word reading (orthographic decoding) is far more efficient than relying on phonological recoding.

Empirical research also has supported the superiority of orthographic decoding. Based on the fundamental tenets of the DRC model, Knoepke et al. (2014) showed that orthographic decoding is better at predicting sentence and text comprehension than phonological recoding in German elementary school children. Further, the results indicated that the ability to recognize words is a significant predictor of general reading skills in all grades. It has been documented repeatedly that high-frequency words are processed more quickly than low-frequency words (Fischer et al., 2014; Kennedy et al., 2013). A study by Masrai (2019) showed that mid-frequency words correlate with reading comprehension in L2 reading. Knowledge of only high-frequency words seems not to be sufficient for skilled reading comprehension (see also Schmitt & Schmitt, 2014; Schmitt et al., 2011). Furthermore, mid-frequency words have not been addressed adequately in pedagogy (Schmitt & Schmitt, 2014). According to other studies, students should receive training in lower frequency words and not only high-frequency words (Calabrèse et al., 2016; Stolowy et al., 2019).

Many struggling students face difficulties using the lexical path. They have restricted access to their mental dictionary and therefore are unable to comprehend the meaning of the text. This occurs because many words are not stored in the mental lexicon, and students cannot access them (Samuels, 2006). As a consequence of inefficient word-reading automaticity, they frequently rely on the non-lexical route (De Jong et al., 2012). This highlights how poor automatization is a key factor for reading fluency, and it may lead to a lack of naming speed and an overall insufficient reading speed (Balass et al., 2010).

As indicated above, reading fluency is a key prerequisite for comprehension. In fact, many studies have shown that the ability to decode correctly and swiftly has a bridging function between transferring the written code into the language and the ability to extract meaning from text (e.g., Kim et al., 2010; National Institute of Child Health and Human Development, 2010; Nese et al., 2013; Roehrig et al., 2008; Schwanenflugel & Kuhn, 2015). A central role of reading fluency is that it alleviates the demands on readers’ limited working memory capacity. In terms of reading acquisition, if too many cognitive resources are devoted to decoding, there is little capacity left for higher process capabilities (Guerin & Murphy, 2015; Paige, 2011; Rasinski, 2003).

According to Hayes (2016), the aforementioned considerations illustrated that the ability to retrieve words quickly is a fundamental reading skill and key qualification for further reading development. Volpe et al. (2011) also argued in favor of sight word reading and advocate for implementing sight word reading instruction in the classroom because it is important to establish those skills before students learn to read complete sentences or passages of text. Moreover, becoming better in sight word reading does not solely lead to increased reading fluency and comprehension; it also builds reading confidence and reduces reading frustration across weak readers (Musti-Rao et al., 2015). Other authors have also argued for training sight words in reading because a very important part of teaching reading is teaching sight words and irregular words as they contribute to text comprehension (Sullivan et al., 2013). Studies showed a positive relationship between word knowledge and word pattern knowledge on basic word reading (in poor German readers) (Rothe et al., 2015; Zarić et al., 2020; Zarić & Nagler, 2021) and on sentence-level reading (Zarić & Nagler, 2021). These results underline the necessity of training orthographic knowledge to enhance reading proficiency. Other studies showed a positive effect of sight word training on trained words, untrained words and word reading fluency (McArthur et al., 2015a, 2015b). Due to the significant impact of a limited number of sight words on reading competence, research on the effectiveness of interventions to promote the use of the lexical pathway is central. Furthermore, work in this area can provide teachers with evidence-based practices to develop these skills (Kuhn & Stahl, 2003). Intervening early after the first signs of reading difficulty has been shown to be particularly important to counteract long-term failure (Volpe et al., 2011). This is illustrated by the fact that, without intervention, around 74% of reading impaired children at the age of nine years maintain their deficits in secondary school (Lee & Yoon, 2017).

The necessity of repeated reading for fostering word recognition skills

Once children demonstrate sufficient phonological awareness and adequate decoding skills, word recognition can be promoted in various ways (Hjetland et al., 2017). The most common and most effective method to enhance the reading fluency in struggling learners is repeated reading (National Institute of Child Health and Human Development, 2000). It is defined as an approach “that consists of rereading a short and meaningful passage until a satisfactory level of fluency is reached” (Samuels, 1979, p. 404). The high potency of this method lies in the fact that children need repetition to automate the retrieval of words from the mental lexicon (Mraz et al., 2013).

In their meta-analysis, Chard et al. (2002) demonstrated that repeated reading leads to increased skills to decode accurately and effortlessly in elementary school children with learning difficulties (d = 0.68). Lee and Yoon’s (2017) meta-analysis, which included empirical studies of the last 25 years to estimate the effects of repeated reading, also confirmed the effectiveness and yielded a total Hedges' g of 1.41 (p < .001) for students with reading disabilities.

Ways to motivate students to engage in repeated reading

However, even very effective interventions like repeated reading miss their intended mark if students are unwilling to get involved in them. For so many children, decoding symbols to arrive at meaning is extremely arduous; they resent it in its entirety (Sabatini et al., 2018). For them, reading interventions must always be complemented with motivational techniques that encourage them to give it a try and to persist (Guthrie & Wigfield, 2000; Harju-Luukkainen et al., 2020; Klem & Connell, 2004).

A number of motivational concepts usually have a marked positive effect on the academic achievement of students, including positive reinforcement, self-monitoring, and praising (Alberto & Troutman, 2008; Copper et al., 2008). In particular, graphing of individual performance scores that allows students to monitor their own progress has demonstrated an especially strong impact on learning positive behaviors (Amato-Zech et al., 2006; Legge et al., 2010) and school performance (Gunter et al., 2003). Often, instruction combines individual motivational techniques. For example, self-monitoring is often used in combination with positive reinforcement (Briesch & Chafouleas, 2009; Joseph & Eveleigh, 2011).

Motivational methods can be applied on an individual level (e.g., positive reinforcement, self-monitoring, praising) and/or on a group level (e.g., group contingencies; Gunter et al., 2003; Stephen & Singh, 2017). Despite the importance of tracking one’s own learning progress, when teaching a whole class of approximately 30 students, using motivational techniques on a group level has advantages over using them on an individual level. Group contingencies, which occur when all group members work together to achieve a certain reward (Kerr & Nelson, 2006), emphasize peer influence to reduce problematic and disruptive behavior as well as to trigger positive behavioral attitudes and enhance social behavior (Donaldson et al., 2011; Ginsburg-Block et al., 2006; Hulac & Benson, 2010; Ling & Barnett, 2013; Ling et al., 2011; Wills et al., 2016). In addition, they may foster peer support and community in a classroom (Groves & Austin, 2019). This approach to motivation is especially suitable for use with a whole class of diverse learners and generally has demonstrated a positive impact on students’ performance (Bowman-Perrot et al., 2013; Pappas et al., 2010; Rohrbeck et al., 2003).

A cooperative game-based approach to integrate repeated reading into regular inclusive classroom instruction

Unfortunately, no matter how potent a particular approach may be, it seldom finds its way into daily classroom instruction (Johnson & Semmelroth, 2014). A primary reason for the often relatively wide research-practice gap in education lies in the prevalent attitude among teachers that many evidence-based techniques are not compatible with everyday school life and the fact that mostly they are not instructed in using evidence-based practice (Hirschkorn & Geelan, 2008; Scheeler et al., 2009). It is very challenging to find ways to adequately attend to the individual needs of a particular child without neglecting the rest of the class. However, researchers must meet this requirement and guide teachers if we want schools to successfully practice inclusion.

Therefore, we implemented the incorporation of repeated reading in conjunction with motivational components in the classroom by adopting a cooperative learning game approach. Class-wide peer tutoring, a system in which learners help each other in working pairs, can unburden teachers while providing everyone with intensive practice time (Bond & Castagnera, 2010). This technique is mindful of individual differences in an inclusive setting, so that all students can be involved whether or not they are performing well. It has been demonstrated to be especially useful when trying to foster reading skills in children (Dufrene et al., 2006; Mercer et al., 2011; Spencer, 2006). Class-wide peer tutoring has been found to enhance the reading competency of children with and without behavioral problems (Bowman-Perrott et al., 2013; d = 0.77), as well as students with differing levels of social competencies (Ginsburg-Block et al., 2006; d = 0.28) and learning-related behaviors (Ginsburg-Block et al., 2006; d = 0.45).

To apply repeated reading by letting students teach each other, we used an educational board game, racetracks. Its underlying idea is simple: Flashcards with particular practice words are placed face down on blank cells on a game board, often designed to look like a Formula 1 circuit. The tutee rolls a die and moves the playing piece the respective number of spaces forward. As it lands on a cell, the tutor picks up the corresponding flashcard, presents the word and asks the tutee to read it out loud. In case of a pause of more than three seconds or a mistake, the tutor models reading the word and asks the tutee to repeat it. The game has no winners or losers; its entire purpose is to provide intense sight word practice (Sperling et al., 2019).

Racetracks has shown positive effects on reading fluency in several studies (Barwasser et al., 2021; Erbey et al., 2011; Green et al., 2010; Grünke, 2019; Grünke & Barwasser, 2019; Hopewell et al., 2011; Hyde et al., 2009; Sperling et al., 2019 and it can be used with a wide range of learners (Falk et al., 2003), which makes it useful for students in inclusion settings. However, all respective studies used a one-on-one teaching approach, unsuitable for inclusive classroom instruction.

Research questions

Even though racetracks and the graphing of performance scores, as well as group contingencies techniques, have been evaluated a number of times with samples from different populations, to our knowledge, these interventions have never been implemented in combination and within a framework of class-wide peer-tutoring. Thus, we cannot draw on the findings from previous studies as we specify our research questions, and we must be cautious as we phrase them. Our work is a pilot study that aims to explore whether crucial components of full-scale projects are feasible.

In particular, we wanted to pursue the following questions:

• (1) Does the implementation of a peer-tutorial reading racetracks intervention with motivational components in an inclusive school setting improve the reading fluency of struggling third-grade readers?

• (2) If there are effects of the intervention, do those effects on reading fluency persist five and ten weeks later?

Methods

Design

An experimental control group design with a pre- and post-, as well as a follow-up measurement was applied. The treatment group received a reading racetracks (RR) intervention, and the control group worked with math racetracks (MR). To ensure a high level of internal validity, both groups were taught at the same time, for the same duration, by the same number of special education college students, and using similar materials. The 18 interventionists worked in teams of two, in a control class or treatment class.

The control group received an MR training with multiplication exercises and the treatment groups received an RR intervention with 30 words for reading training. All participants from both groups were compared regarding sight word fluency through a pretest (t1), posttest (t2), and two follow-up measurements (t3 and t4). Both groups met four times a week for 15 min, lasting a total of three weeks. All participants attended 12 intervention sessions which took place with the whole class in a classroom with the class teachers present. The first follow-up happened five weeks after post testing, including two weeks of autumn school holidays, and the second follow-up was conducted ten weeks after post testing.

Participants and setting

The final sample (N = 44) consisted of third graders from nine classes of five different elementary schools in a high-socioeconomic metropolitan area of Germany. On average, 192 students were enrolled in each school (SD = 8.4). In every school, except one, two whole classes participated either as experimental group or control group.

Although the RR and MR trainings were implemented in a classroom context with all students present, we focused only on those with the lowest and highest skill levels. To identify our sample, we conducted the Salzburg Reading Screening Tests (SLS; Mayringer & Wimmer, 2014) with 225 students. The SLS measures reading fluency by asking children to decide whether different sentences that are presented to them make sense within an assigned time period. We chose the SLS due to school time constraints, as it is a group screening. Thus, we conducted it with the whole class at once to capture the reading performance of all students of the participating classrooms.

Using the results, we ranked students based on their reading performance from lowest to best performance. Subsequently, we paired the weakest students (tutees) were with the strongest students (tutors). The classification process involved dividing participants into tutors and tutees by taking (a) the ones with the best scores (tutors, reading level at least > 100) and (b) the ones with the weakest scores (tutees, reading level maximum of < 89). Those students with an average reading score were paired together and participated in the intervention, but no data were collected from them and they were therefore not considered part of the sample. Also, no data was collected from the tutors. However, we only collected data from those children selected as tutees (N = 44), because the weaker readers should not have been able to read the training words, but the tutors, the stronger readers, should have been able to read the words and to give adequate feedback and help. In all, the final tutees had a reading quotient (LQ) of M = 74.6 (SD = 8.4) and the tutors an LQ of M = 110.4 (SD = 8.5). According to the SLS manual, an LQ of 80–89 is considered below average, 70–79 is considered weak, and less than or equal to 69 is considered very weak. The LQ expresses the extent to which the measured reading ability deviates from the average of the norming sample. The same scaling is used as for intelligence test measurements: where 100 stands for the mean value with an SD of 15 in each case.

Additionally, to gain more information about the tutees, an intelligence test (CFT-20-R, Weiß, 2006; German adapted Version of the Cultural Fair Intelligence Test; Cattell & Cattell, 1963) and the Strength and Difficulties Questionnaire (SDQ; Goodman, 1997) were implemented. The CFT-20-R consists of the following four subtests: series continuation, classifications, matrices, and topological conclusions. According to its manual, the correlations with other German intelligence tests range between r = 0.57 and r = 0.73. Additionally, the CFT-20-R is moderately correlated with school grades in mathematics (r = 0.50). The SDQ consists of five scales (emotional symptoms, behavioral problems, hyperactivity/dislike, peer relationship problems, and prosocial behavior), the first four scales of which can be combined into an overall problem value. Each scale consists of five items with three gradients (0 = not applicable, 1 = reasonably applicable, and 2 = definitely applicable) that are typically completed by teachers. Studies on the psychometric characteristics of the German version of the SDQ indicate good internal consistency. In addition, the German version shows good validity (Becker et al., 2004). Due to time constraints, for the SDQ measurement, the short (16 instead of 25 questions) form was completed by the teachers. For the total problem score, values between 12 and 15 are considered borderline and values between 16 and 40 are considered abnormal. SDQ results for the reading racetrack group were M = 9.8 (SD = 5.7) and for the control group M = 9.4 (SD = 6.8; Table 2). More specific information about the children (age, gender, special needs, and German L2) was collected through a teacher questionnaire developed by the authors.

To split up the participants into experimental group and control groups, matched pairs of tandems, consisting of a tutor and a tutee, were identified based on the pretest results per school. Each pair was randomly assigned to either the treatment group (working with reading racetracks) or to the control group (working with math racetracks). No significant differences between the experimental and control group, in terms of gender, age, special needs, German as L2, cognitive abilities, reading proficiency and externalizing problem behavior were found (Tables 1 and 2).

Table 1 Socio-Demographic Characteristics of Participants Comparing MR and RR

Table 2 Results of SDQ and Reading Test Comparing MR and RR

Further, to determine the words for the reading racetracks intervention, word-reading pre-tests were used. This pretest consisted of two PowerPoint presentations. On the slides were individual words, each with three consecutive hashtags, which the children were told to read within 1s of the presentation. The one 1-s cycle was automatically preset, so the slides changed by themselves after 1s. We utilized a list of the 1000 most frequently used German words (https://wortschatz.uni-leipzig.de/de) and the database ChildLex (Schroeder et al., 2015) in addition since the prior mentioned list refers more to older students’ vocabulary and words mentioned there are of lower frequency for children but still important with regard to the future. The ChildLex database was used to determine the words’ overall frequency.

Each PowerPoint presentation contained 70 words to increase the probability of finding 30 words for the intervention that were not familiar sight words for all subjects. The two presentations were divided into two days to avoid overtaxing the children. The final 30 words of training, which were the same for each participant and were not yet stored as sight words across all children, were of different syllables and a rather mid-frequency of M = 38.4 (SD = 37.0) ("Appendix"). This means that the words appeared on average 38 times per million words in a corpus. In comparison, low-frequency words appear five times and high- frequency words more than 100 times per million words (Brysbaert et al., 2018).

Dependent variables and measurement

The dependent variable in this study was automatic recognition of the training words, operationalized by the number of words that the participants read correctly within 1 s after having been presented with it (words read correctly, WRC). As suggested by Ehri (2005), “reading words within one second of seeing them is taken to indicate sight word reading” (p.136). A PowerPoint presentation that contained all 30 training words was used, each on a 1-s timer, to determine whether each word was read by sight. The words were presented in random order for each measurement. All tutees were measured independently in another room to reduce variables of potential influence. The 30 training words were the same for all participants.

Interventions

The intervention was conducted in the regular class setting. Both children of a pair (tandem) sat opposite each other and the racetrack game in the middle of them. At the start signal, the tutee rolled the dice and moved the piece to the appropriate square according to the number (see Fig. 1). Flashcards on this space were turned over and the tutee was asked to read them out loud. The job of the tutor was to attend to the correct pronunciation of the word and provide feedback to the tutee if needed. Tutees had three seconds to correct themselves. If someone did not make it, the tutor read the word aloud correctly, and the tutee was asked to read it again. Flashcards that had been read were left with the front side up to ensure that the other words would be turned around and read so that the tutees could not land on the space of an already read word. After that the tutee rolled the dice again and the procedure was conducted for 10 min. If all flashcards were read before the ten minutes ended, they were shuffled and placed back on the racetrack cells.

Fig. 1

Self-made racetrack board

After finishing the game, there was a brief word-reading assessment that the tutees graphed. The interventionists used a 1-min timer, and the tutee was asked to read as many words correctly as possible within this minute. Then, using a blank graph, the amount of correctly read words was drawn in the provided cells. The first row was filled in at the first intervention day, and another row was completed on each intervention day. The pairs also received rewards based on their progress over time. A pair was given one point for reading the same number of words correctly as they had the previous intervention. They received two points if there was improvement from the last session. These points were collected in front of the class, where each pair added the amount of points in a bottle to work toward achieving an overall class-wide improvement goal.

The same procedure was used for the control group intervention except the words on the flashcards were replaced with math problems. For the motivational system, the tutee was asked to answer as many problems correctly as possible within two minutes. The interventionists were intensively trained for two days in the implementation of the treatment and control group. In addition, there was a standard treatment protocol with the same instructions on implementation, which the interventionists had to follow.

Material

Materials for the experimental and control groups were exactly equivalent except for the content of the flashcards. Both groups were provided with a racetrack board consisting of 30 empty spaces for the flashcards. The reading racetrack group (experimental) had 30 flashcards with the words on one side, and math racetrack group (control) were provided 30 flashcards with multiplication problems on one side. The tandems which were not included in data collection got different words than the tandems from which data was collected since they were too good readers. Both groups were also given a self-graphing sheet entitled either “Reading Racetrack” or “Math Racetrack.” These sheets consisted of 12 rows with 30 empty cells, each of which represented an intervention day.

Treatment Integrity

To draw valid conclusions regarding the effectiveness of an intervention, treatment integrity is required (Hagermoser et al., 2009). Therefore, a detailed script was provided for the interventionists. The guidelines contained the exact procedure per session regarding the reading intervention and the reward system. All interventionists and the class teachers were instructed in the intervention procedure in addition to all students of the class (including all tutees and tutors—also those from whom no data were collected). To assess adherence, exposure, quality, and dosage of the intervention, the interventionists were asked to complete a checklist after each session. The questionnaire contained questions like: “Did you, as interventionist, follow the script?”, “Did you adhere to the time frame?”, and “Were all the materials available?”. After each session, the interventionists completed the 18-question questionnaire and submitted a list of students who had participated in the intervention. Additionally, one-third of the sessions (six) were observed by an independent observer who used the treatment integrity sheet to assess the implementation of the session. Interrater agreement equaled 100% between the interventionists and between the external observers. Finally, the intervention was carefully monitored by the first author, who conducted weekly meetings with all interventionists and maintained almost daily contact with all interventionists.

Social validity

Social validity is necessary to determine the acceptance and usefulness of interventions (Briesch et al., 2013; Wolf, 1978). Using the Usage Rating Profile—Intervention by Briesch et al. (2013), we distributed a questionnaire to assess the acceptance, understanding, and feasibility of the intervention among students and teachers. The social validity was assessed by separate seven-item questionnaires for teachers and students. Both questionnaires used a 6-point Likert scale ranging from 0 (no agreement) to 5 (absolute agreement). Items on the teacher questionnaire were created to assess their understanding (e.g., “The intervention is a good way to improve the reading fluency of students”), acceptance (e.g., “I would use the intervention in my lessons as well”), and perceptions of the feasibility of the intervention (e.g., “The total time required for the intervention procedure was manageable”). Similarly, items on the student questionnaire were designed to assess their acceptance (e.g., “I gladly came to the sessions”) and understanding of the intervention (e.g., “I understood the purpose of the intervention well”). For the students’ survey, the interventionists left the room and the teachers read out loud each scale with all the items in order (a) to avoid bias in the answers if the interventionists had done the questioning, and (b) to ensure that all students understood the questions.

Data analysis

To answer the research question, a 2 (Conditions: Treatment, Control) × 4 (Time: Pre, Post, Follow-up 1, Follow-up 2) repeated measures analysis of variance (rmANOVA) was conducted to examine the intervention effect on correctly read words per second. Based on the rmANOVA, separate analyses were carried out examining the differences between treatment and control group, as well as between pre-, post- and follow-up data.

If the assumptions of sphericity were not fulfilled (Mauchly test of sphericity), the F estimate was based on the Green House Geisser correction with adjustment of the degrees of freedom (Field, 2013). The significance of between-subjects and within-subjects effects was tested by using independent and pairwise Welch test comparisons, applying the Bonferroni adjustment (using the mean difference of time_i and time_j). The significance level was set at p < 0.05.

Violations of equality of error variances and covariance were tested by means of the Levene test and the Box's test. A robust rmANOVA using the R-package WRS2 (Mair & Wilcox, 2019) was performed to validate the effects.

Results

Preliminary analysis

Prior to the main analysis, the distribution of the dependent variable was tested. The Mauchly test for sphericity was significant (Mauchly-W = 0.33, p < . 001). To correct this violation, the Greenhouse–Geisser adjustment was used. When testing for homogeneity of the error variances, as assessed by Levene’s test, the results for the post-test (p < .01) and the Follow-up 1 condition (p < .05) were found to deviate significantly from the null hypothesis, which was evident for the homogeneity of covariances, as assessed by Box’s test (p < .001). These violations were addressed by the application of a robust rmANOVA using the R package WRS2 (Mair & Wilcox, 2019).

Main analysis

Descriptive results for treatment and comparison groups at pre-, post-, follow-up 1, and follow-up 2 test are summarized in Table 3.

Table 3 Descriptive Statistics for Words Read Correctly

There was no statistically significant pretest difference between students in the reading racetracks and the math racetracks group for WRC t(41.86) = 0.007, p = .94, d = 0.002.

The rmANOVA to test the interaction effect (Time × Group) showed a significant result, Greenhouse–Geisser F(1.71, 71.73) = 133.60, p < .001, partial η² = 0.76. The inner-subject main effect for time indicates that this factor has a statistically significant influence on WRC for the reading racetracks group, F(1.62, 34.02) = 203.96, p < .001, partial η² = 0.91. Bonferroni tests for pairwise comparisons revealed a statistically significant increase of WRC after intervention (22.64, p < .001, d = 5.77) that remained stable through the follow-up 1 assessment (21.77, p < .001, d = 4.09) and follow-up 2 (18.55, p < .001, d = 3.59) in comparison to the pretest. Although the performance between the posttest and follow-up 1 remained constant (0.86, p = 1.00, d = 0.14), there was a decrease in the WRC from posttest to follow-up 2 (4.09, p < .001, d = 0.66), as well as from follow-up 1 to follow-up 2 (3.23, p < .001, d = 0.45), albeit with medium effect. Regarding the math racetracks group, there was also a significant influence of the main factor time on WRC, F (1.64, 34.42) = 4.13, p < .05, partial η ² = 0.16. The pairwise comparisons assessing the influence of time revealed no significant differences. This refers to the comparison of the pretest with the posttest (0.86, p = .31, d = 0.39) and with follow-up 1 (1.50, p = 0.26, d = 0.46) and follow-up 2 (2.10, p = .15, d = 0.57), as well as between the posttest and follow-up 1 (0.64, p = 1.00, d = 0.19) and follow-up 2 (1.23, p = .52, d = 0.32), and between follow-up 1 and follow-up 2 (0.59, p = 1.00, d = 0.13).

The analysis of the between-subject main effect indicates that there was a significant main effect for group, F(1, 42) = 137.98, p < .001, partial η² = 0.77. Although there was no group difference for the pretest t(41.86) = 0.007, p = .94, d = 0.03, significant group differences were shown for all further measurement times (posttest: t(29.95) = 308.45, p < .001, d = 5.29; follow-up 1: t(33.63) = 126.17, p < .001, d = 3.39; follow up 2: t(36.81) = 81.06, p < .001, d = 2.55).

Due to the significance of the Levene test and the Box`s test for homogeneity of the error and for homogeneity of covariances, a robust rmANOVA was conducted, which supports the results of the rmANOVA in terms of the interaction effect Time × Group (F(3, 15.48) = 185.10, p < .001, η² = 0.97), as well as the main effect for time (F(3, 15.48) = 197.50, p < .001, η² = 0.97) and group (F(1, 19.90) = 160.48, p < .001, η² = 0.96).

Social acceptability of the reading racetrack intervention

To ascertain the social validity of the intervention, results from questionnaires for the teachers (Table 4) and the reading racetracks students (Table 5) are provided. Across all seven items, the six teachers rated the intervention in all three areas (acceptance, understanding, feasibility) as positive, with a general rating between 3 (agree) and 5 (absolutely agree). The items SocV5 (“I would use the intervention in my teaching”) and SocV7 (“The material resources required for this intervention were appropriate”) had the highest ratings. In total, questionnaires were returned by six of the nine teachers.

Table 4 Social Validity Questionnaire Teachers

Table 5 Social Validity Questionnaire Students

For the reading racetracks students (n = 22), the average of all seven items in the questionnaire ranged from 3 (agree) to 4 (strongly agree). The items SocV1 ("The racetrack helped me to read words correctly"), SocV4 (“I learned a lot during the intervention”) and SocV5 (“I came to the sessions with pleasure”) were rated highest.

Discussion

Main findings

Struggling readers often demonstrate problems in automatic word recognition. As a result, they also have lasting difficulties with higher-order reading processes, such as text comprehension (Ravitch, 2010). The automated decoding of words is a basic prerequisite for the development of advanced reading skills (Hayes, 2016; Knoepke et al., 2014; Tunmer & Chapmann, 2012). It therefore seems important to investigate which interventions can effectively and sustainably help students overcome their challenges with fluent word reading. The current pilot study examined the extent to which a peer-tutorial reading racetrack training supplemented by motivational components is an effective method to promote reading fluency of students with word-reading difficulties.

Our results indicated that our approach was very effective in increasing the reading fluency of the trained words by students in the experimental group. Even though the participants in the control condition also played a racetracks game (MR), implemented peer-tutoring procedures, and were motivated in exactly the same way as those who practiced reading, they did not show a comparable performance gain. Fortunately, this effect was still clearly evident ten weeks after the end of the intervention. Although data from the second follow-up showed a slight decrease in WRC, the students in the experimental group still read significantly more words than those in the control group. Thus, our results align with the findings of previous studies (e.g. Erbey et al., 2011; Green et al., 2010; Grünke, 2019; Hopewell et al., 2011; Hyde et al., 2009), and give an indication of the long-term effectiveness of the treatment.

In addition, the study provides an indication that words with a mid-frequency can also be trained effectively. The learning of mid-frequency words seems to be of importance for the development of reading competence (Masrai, 2019), as the processing of these words can have a significant slowing effect on lexical processing (Fischer-Baum et al., 2014; Kennedy et al., 2013). Thus, in addition to automated recognition of high-frequency words, acquisition of a comprehensive sight vocabulary of lower frequency words can also strengthen lexical processing (Calabrèse et al., 2016; Stolowy et al., 2019).

Furthermore, the intervention appears to be an equally effective and economical method for inclusive education, as it has been successfully applied using peer-tutoring in an inclusive classroom available to all students. In addition, responses from the teachers' and tutees’ social validity questionnaires indicated support for the intervention in the three areas of understanding, acceptance, and feasibility. In future studies, it would also be interesting to look at the evaluation of the social validity of the tutors.

Limitations and further research

The results of this research must be interpreted with caution. A preliminary small-scale study like ours does not allow for far-reaching conclusions about the effectiveness of our approach. This is especially true given that it was conducted in a specific geographical area in Germany and included a constrained number of trained sight words. Thus, the results can only be seen as a first indication of the usefulness of applying reading racetracks with certain motivational techniques in a class-wide peer-tutorial setting. Further studies with a larger sample should be carried out to substantiate the finding. This would allow for the consideration of additional control variables, such as spoken language(s), socioeconomic status, working memory, and home reading environment for matching experimental and control groups, in addition to the control variables used: age, gender, LD, German L2, reading proficiency, and reading behavior. Further, it would be of interest to consider whether differential effects on the effectiveness of the method can be mapped as a function of student characteristics. With regard to the use of diagnostic instruments, it should be noted that, due to time constraints of the school, the SLS was used in the present study, as it is possible to use this as a group test. However, the test measures reading fluency at sentence level and also assesses sentence comprehension. For future studies, it seems more appropriate to use a diagnostic instrument that measures reading fluency at the word level, such as the Salzburg Reading and Writing Test (SLRT-II; Moll & Landerl, 2010).

Moreover, it remains questionable whether the method is superior to another drill and practice approach. We compared the outcomes of our training with those of a math racetrack intervention. An interesting issue would be to examine the effectiveness of reading racetracks, relative to other methods for strengthening sight-word recognition.

Last, it is still unclear how far-reaching our results are for the reading progress of students since we did not measure transfer effects on unknown words and general reading fluency. However, Knoepke et al. (2014) showed that, according to the dual route cascaded model (Coltheart et al., 2001), the aspect of orthographical decoding skills, addressed in our study is an important prerequisite for reading comprehension and that sight word training can have positive effects on trained and untrained words as well as sentence reading fluency (e.g. McArthur et al., 2015a, 2015b). Thus, we would anticipate a transfer effect on other words, which, of course, we cannot prove on the basis of our data. Future studies should not only examine the effectiveness of reading racetrack interventions on reading fluency at the word level, but also evaluate the impact of the intervention on reading fluency at the sentence and text levels, as well as reading comprehension.

Practical implications

The research presented provides valuable suggestions to support readers who have difficulties in basic reading skills in an inclusive context. This is of particular importance, because a considerable number of students with learning problems are taught in classrooms with very diverse learners. For instance, in Germany, 48.65% of children and youth with special educational needs (i.e., including those with behavioral and language problems) in inclusive schools experience severe academic difficulties (KMK, 2018). The desire to support them in an inclusive context according to their individual needs is often hampered by a lack of resources and the absence of teaching methods that address a heterogeneous population at class level (McLeskey & Waldron, 2011; Schmidt et al., 2002). It is therefore vital to identify methods that are evidence-based, socially valid, easy to apply, and suitable for meeting the needs of different individuals (Mitchell & Sutherland, 2020). Even though many evidenced-based practices are known for struggling students, many general and special education teachers do not use them in their teaching (Maheady et al., 2013). One reason for this may be that these practices have not been adequately taught to teachers and thus they do not know how to use them (Scheeler et al., 2009). Therefore, it is of immense importance to teach the methods to the teachers in a comprehensible way and furthermore to pay special attention to very easy to use and effective interventions, like here the reading racetracks in a peer-tutorial setting. Moreover, the teachers were present during the whole procedure and also received instruction beforehand. Researchers should make practices comprehensible for teachers to facilitate their implementation in classrooms. As mentioned before, given the scarce resources available in schools, it is important that individual support for students can also take place in class. The present study shows that this is possible, even if the students work together in pairs and speak quietly to each other while reading. As reading comprehension is the key to accessing the curriculum and to academic success, it is irremissible to make sure that students acquire the necessary prerequisite skills, especially sight word recognition. As Hayer (2016) documented, sight-word training increases the ability to decode and comprehend text. Especially for struggling readers, integrating this into the lesson plan seems to be particularly important. In addition to practicing high-frequency words, automating lower frequency words can lead to an increase in reading fluency related to words with irregular sound patterns (Calabrèse et al., 2016; Stolowy et al., 2019). These findings underline the practical importance of training high- and also mid-frequency words.

Due to the low implementation effort, reading racetracks can be integrated easily into everyday school life and thus represents an evidence-based method to support students’ ability to recognize sight-words. Through the use of flashcards, it is easy to adapt the words to the students' individual needs beyond sight-word reading. Thus, the intervention can also be adapted to practice basic vocabulary among students with German as a second language (Grünke & Barwasser, 2019; Sperling et al., 2019). In follow-up studies, it would be interesting to systematically assess the optimal intervention dosage. Here, for example, the implementation of controlled single-case studies may be a useful approach, as they allow for a detailed view of the students’ learning process (Horner et al., 2005).

The game-based implementation of the training and the use of motivational components at individual and class levels encourages students to engage in the typically monotonous learning of word reading over a longer period of time (Amato-Zech et al., 2006; Lämsä et al., 2018; Legge et al., 2010). In addition to fostering students’ skill with word reading or vocabulary, the implementation of peer tutoring is also a way of promoting social integration (Mitchell & Sutherland, 2020). This is of particular importance, as these appear to be significantly lower for students with special educational needs than for those without (Krull et al., 2018).

Conclusion

Despite the limitations, the intervention is highly effective in improving the word-reading fluency of struggling readers. In particular, the low costs and minimal effort required for the intervention make it practical for everyday teaching in inclusive education. Further, the results illustrate the long-term effect of the intervention.

Appendix

See Table

Table 6 Words of Training

6.