Reading comprehension: A computerised intervention with primary-age poor readers

The current study investigates the effectiveness of a computerized reading comprehension programme on the reading accuracy, reading comprehension and reading rate of primary-age poor readers. There is little published literature relating to computerized reading interventions in UK primary schools, and no previous studies have investigated the Comprehension Booster programme. Thirty-eight children (26 boys and 12 girls; aged 6:7 to 11:0) from two schools in East Yorkshire, UK, took part. Half of the participants (the intervention group) undertook the Comprehension Booster programme for a 6-week period, whilst the other half (the control group) continued with their usual teaching. Significant effects of the intervention were found, with increases in reading accuracy and reading comprehension for the intervention group. It is concluded that computerized reading programmes can be effective in improving reading skills, and these are particularly useful for pupils with reading difficulties in disadvantaged areas, where resources are limited and family support in reading is lower. However, such programmes are not a replacement for good teaching, and regular monitoring of children with reading difficulties is required. Further research is necessary to compare the programme used here to other conventional and computerized intervention programmes, using a larger sample. Copyright © 2017 John Wiley & Sons, Ltd.


Introduction
Comprehension is vital and is the ultimate goal of reading (Nation & Norbury, 2005) and, although reading fluency (automatic word recognition) is necessary for students to achieve comprehension (Moats, 1998), decoding skill does not necessarily mean that comprehension will follow (Nation & Norbury, 2005). Opportunity to read is significantly associated with reading achievement (Guthrie, Schafer & Huang, 2001). In order to become independent readers, children need to be provided with a wide range of reading experiences (Cunningham & Stanovich, 2001). Getting children to read more, especially outside of school, although vital for improving reading skill, is very difficult to achieve. The National Literacy Trust survey (Clark, 2014a) found that only 41% of 8-11 year olds read outside class daily, with 12% rarely or never reading outside of school (16% of boys and 8% of girls). The survey also found that 27% of children said that their parents don't care whether they spend any time reading and 47% state that they prefer watching television to reading (Clark, 2014a). Despite continual increases in reading standards over the last 17 years, 11% of English 11-year-olds do not meet the expected level (Level 4 or higher) in their SATs reading test (DfE, 2014). This figure is higher (18%) for pupils eligible for free school meals. Poor readers will miss out on a good deal of academic content due to their inability to acquire knowledge from text reading (Carnine & Carnine, 2004;Montali & Lewandowski, 1996;Visone, 2010) and Reading comprehension: A computerised intervention with primary-age poor readers 2 the requirement to learn from reading text increases as students progress through their education (Williams, 1998). Beyond education, research has shown poor literacy to be related to unemployment, low pay, poor health and offending (Morrisroe, 2014).
Reading fluency is essential for reading success (Hasbrouck & Tindal, 2006;Speece & Ritchey, 2005). A child who is fluent in reading is able to recognise the majority of words in a text quickly and without relying on phonological decoding.
Reading fluency requires a large amount of practice to become established and a lack of practice in reading and exposure to text negatively affects the automaticity and speed of word recognition (Cunningham & Stanovich, 2001). Poor decoding makes reading arduous and leads to errors and misreading of words (Torgesen, 2002) affecting the development of fluent word reading, which slows down the reading process and diverts effort away from understanding the meaning of the text (Hudson, Lane & Pullen, 2005).
Children who lack fluency in word decoding will have difficulty reading independently and their comprehension will be impaired (Ehri, 2005;Fuchs, Fuchs, Hosp & Jenkins, 2001;McKenna, 1998;Stanovich, 1986). Torgesen and Barker (1995) suggest that the majority of children with reading difficulties do not receive sufficient practice at decoding to reach a level of fluency that will allow them to read effectively. Due to their difficulty in phonological decoding, these children require far more practice and exposure to text to become fluent in word recognition (Ehri, 2005).
Reading comprehension problems have also been attributed to deficits in vocabulary knowledge (Beck, Perfetti & McKeown, 1982). If a child comes across a word that is not in his or her vocabulary then they are unlikely to identify it, even using phonological decoding. Furthermore, the higher the proportion of unfamiliar words within a text, the more likely it is that comprehension will be impaired. Pickering and Reading comprehension: A computerised intervention with primary-age poor readers 3 Gathercole (2004) suggest that the poor phonological storage skills of children with language difficulties may lead to early difficulties in acquiring new vocabulary. In addition, research has found that the relation between reading ability and vocabulary knowledge is reciprocal -children who read widely increase their vocabulary knowledge and are able to comprehend a wider variety of texts, thus increasing their reading experience and developing their vocabulary further (Cunningham & Allington, 2010). Lovett, Barron and Benson (2003) stress the importance of text reading practice using a controlled decodable vocabulary.
Reading comprehension also requires complex thought processes, such as relating information in the text to wider knowledge, drawing inferences and making deductions. Cain and Oakhill (1999) found children with comprehension difficulties to be poor at inference-making and further research has shown that the relation between comprehension ability and inference-making skill remains strong when knowledge base is controlled for (Cain, Oakhill, Barnes & Bryant, 2001). However, an inference cannot be made without the ability to relate the information in the text to wider knowledge (Ackerman, Silver & Glickman, 1990). Much experience is required to develop these thinking skills and instruction that promotes critical thinking and considering alternatives in decision-making has been shown to augment reading comprehension (Fletcher, Lyon, Fuchs & Barnes, 2007).
Motivation is also essential for reading progress. Research has shown a bidirectional relation between reading ability and motivation to read (Morgan & Fuchs, 2007); good readers tend to enjoy reading and so read more, thus improving their ability further, whilst struggling readers tend not to enjoy reading (McKenna, 1998) and so read less, falling further behind their peers (Hasbrouck, Ihnot & Rogers, 1999) -this is Reading comprehension: A computerised intervention with primary-age poor readers 4 known as the 'Matthew Effect' (Stanovich, 1986). It has been suggested that the lack of motivation seen in children at risk of reading difficulty occurs within a year of starting school (Chapman, Tunmer & Prochnow, 2000). Children with reading difficulties are the most likely to benefit from reading practice but the least motivated to carry this out (Chapman, 1988).
Several researchers have shown independent silent reading to be beneficial in terms of reading achievement, comprehension, fluency and vocabulary (Kelley & Clausen-Grace, 2006;Reutzel, Jones, Fawson & Smith, 2008), particularly for students over the age of 8 years old (Kuhn, 2004;Reutzel, Petscher & Spichtig, 2012). However, there are some issues with using independent silent reading practice in the classroom, such as ensuring that students are actually able to read the books they choose (Donovan, Smolkin & Lomax, 2000) and keeping their eyes on the text (Hiebert, Wilson & Trainin, 2010). Nevertheless, research has shown computer-based guided silent reading interventions, using texts which increase in difficulty and length, to produce gains in reading comprehension and general literacy achievement (Rasinski, Samuels, Hiebert, Petscher & Feller, 2011;Reutzel et al., 2012). Reutzel et al. suggest that the success of the computer-based guided intervention is due to the program monitoring the children's comprehension and adjusting the text level to suit their ability (rather than children selecting texts that are too difficult for them), promoting motivation by giving students a choice of genre within the appropriate level of text, making students accountable for their time by monitoring their progress and focusing on the development of fluency, vocabulary and comprehension skills.
Reading aloud to children has been found to have a positive impact on language, phonological awareness, print concepts, comprehension and vocabulary (Swanson et al., Reading comprehension: A computerised intervention with primary-age poor readers 5 2011), and is especially effective when it involves repeated reading, vocabulary instruction and questioning (Karweit & Wasik, 1996). Repeated reading of passages or stories, either independently or reading along, has been found to develop reading rate, accuracy, expression, comprehension and retention of information (Armstrong & Hughes, 2012;Barnett & Seefeldt, 1989;Nelson, Alber & Gordy, 2004;Pappas & Brown, 1987;Therrien, Wickstrom & Jones, 2006). Computer interventions involving reading aloud with a computer have also been seen to be effective in increasing comprehension (Basil & Reyes, 2003;Swanson et al., 2011). It has been suggested that the provision of increasingly difficult text, along with corrective feedback is imperative for an improvement in reading (Chard, Vaughn & Tyler, 2002). Hall, Hughes and Filbert (2000) suggest that strategic feedback must include a review process where students return to questions answered incorrectly before moving on to new tasks. Hatcher, Hulme and Snowling (2004) argue that mainstream teaching does not generally enable children at-risk of reading difficulties to develop fluent and accurate reading and research suggests that this group of children need an individualised approach (Torgesen, 2005). However, Torgesen, Alexander, Wagner, Rashotte, Voeller and Conway (2001) purport that, although special education reading instruction is more effective than classroom instruction, it is not enough to bring reading disabled children's reading up to average levels. Due to their difficulty in phonological decoding, these children require far more practice and exposure to text to become fluent in word recognition (Ehri, 2005). However, teachers looking for strategies to improve the decoding and comprehension skills of poor readers are limited in the amount of instruction and practice they can provide due to teacher shortages, school finances and time factors (Fielding, Kerr & Rosier, 2007;Gibson, Cartledge & Keyes, 2011;Hall et Reading comprehension: A computerised intervention with primary-age poor readers 6 al., 2000). Furthermore, Moats (1999) suggests that many teachers are not adequately trained to provide the specific instruction necessary for students at risk of reading disabilities.
Researchers have stressed the importance of technology in supporting struggling readers by providing the intensive, individualised instruction and extensive repeated practice that they require (Hasselbring & Goin, 2004;Torgesen, 2002), allowing children to work at their own pace with the provision of immediate feedback and instructive corrections (Hall et al., 2000) along with efficient monitoring of student progress (Kim, Vaughn, Klingner, Woodruff, Klein Reutebuch & Kouzekanani, 2006).
The adaptive nature of computerised learning tools makes them more effective as slower learners get more practice before progressing on to subsequent stages (McDonald Connor et al., 2009). Furthermore, this specialised instruction can be provided at a low cost but with high reliability (Torgesen, Wagner, Rashotte, Herron & Lindamood, 2010), whereas teachers are not always able to apply interventions with this high level of integrity (Moore & Fisher, 2007). A number of studies have found that using technology in reading activities, not only benefits reading ability, but also improves self-confidence and motivation (Adam & Wild, 1997;Hall et al., 2000;Mioduser, Tur-Kaspa & Leitner, 2000;Moore & Calvert, 2000) and reduces behavioural problems during instruction (Chen & Bernard-Opitz, 1993 (Clark, 2014b). By keeping struggling readers motivated, they are less likely to give up (Karemaker, Pitchford & O'Malley, 2010), which may help overcome the 'Matthew Effect'. Furthermore, the use of computerised reading instruction allows teachers more time to work individually with students who need particular help (Carnine, 1989, cited in Kim et al., 2006. the programs and wishing to continue using them (Gibson et al., 2011;Kim et al., 2006;Torgesen et al., 2010). However, there is little evidence of significant increases in reading rate in previous studies. Indeed, Torgesen et al. (2001) suggest that rate may not significantly improve due to the nature of reading comprehension tests. Post intervention, participants progress further through tests and therefore encounter harder passages with more unfamiliar words, which take longer to decode. They argue that increasing the reading rate of older poor readers is challenging as their previous lack of exposure to text limits their reading fluency.
Notwithstanding the recognised advantages of computerised reading programs, Gibson et al. (2011) argue that many commercial programs are being used in Reading comprehension: A computerised intervention with primary-age poor readers 8 classrooms without the empirical research to support their use. Karemaker et al. (2010) suggest that the majority of previous research studies on computer-assisted reading instruction are based on programs that present words in isolation, rather than storybook formats, and are not based in the natural classroom setting, making them less valid in terms of educational practice. Furthermore, Saine, Lerkkanen, Ahonen, Tolvanen and Lyytinen (2013) assert that most previous studies of computerised reading interventions have not used control groups who continue in mainstream instruction, rather than taking part in a specialist intervention program, and many do not report effect sizes.
Galuschka, Ise, Krick and Schulte-Korne (2014) support the assertion that research into reading programs should utilise randomised controlled trials.
Comprehension Booster is a computer program that provides pupils with interactive reading practice using a variety of texts, and opportunities to acquire new vocabulary and learn the thinking skills required for understanding text (Lucid Innovations, 2010). It is possible to set specific teaching plans for individual children or whole classes and speech can be enabled or disabled, according to pupil needs. Reports generated by the program allow each child's progress to be constantly monitored.
The aim of the present study is to investigate the effectiveness of Comprehension Booster in developing the reading accuracy, reading fluency and reading comprehension of primary school children with reading difficulties, using a randomised controlled trial. The current study is designed to fill a gap in the literature as there is currently no published evidence of the effectiveness of this program. Indeed, there is a paucity of studies looking at the effectiveness of computerised reading interventions in UK primary schools.
The hypotheses are: 1) there will be significantly more improvement in reading accuracy and reading comprehension, from baseline to post-test, in the intervention group than in the control group; 2) there will be no significant difference between the intervention and control group in improvement shown in reading rate from baseline to post-test.

Participants
The 38 participants attended two primary schools in the north of England. These schools were selected due to them having above average numbers of disadvantaged children (as measured by Free School Meal data), high numbers of pupils with Special Educational Needs (SEN; as measured by School Action Plus and Statement data) and low levels of attainment in reading (as measured by Key Stage 2 test data). Both Reading comprehension: A computerised intervention with primary-age poor readers 10 schools' most recent Ofsted reports rated them as requiring improvement. Summary data (percentages and quintile scores) for both schools are presented in Table 1.
School 1 is a Church of England Voluntary Controlled Junior (7-11) school in East Yorkshire, UK. It is a large mixed school, with 378 pupils on the roll. The majority of pupils at the school live on two large council estates nearby and most are of White British origin, with a low proportion of children speaking English as an Additional Language. The latest Ofsted report states that pupil achievement in reading requires improvement and that reading needs to be regularly assessed in order for learning to meet the pupils' needs.
School 2 is a community primary (4-11) school in East Yorkshire, UK. It is a large mixed school, with 350 pupils on the roll. The majority of pupils at the school are of White British origin, with a small number from a wide range of minority ethnic heritages. The school has a low proportion of children speaking English as an Additional Language. The latest Ofsted report states that the school needs to raise attainment in reading.
[ Table 1 about here] Participants for the study were randomly selected from the school's SEN register of pupils on School Action Plus (meaning their progress has been limited and they require the support of external services) for literacy difficulties. Some of the pupils had other co-morbid needs, including numeracy, hearing and behavioural difficulties.
Summary data for participants are presented in Table 2.
Participants from School 1 were 18 pupils (12 boys and 6 girls) aged from 7:4 to 11:0, with a mean age of 9:5 (SD=1.14). Participants were matched in pairs on the basis Reading comprehension: A computerised intervention with primary-age poor readers 11 of age and gender, with one child from each pair selected at random and placed into the intervention group and the other into the control group. The intervention group consisted of 9 pupils (6 boys and 3 girls) aged from 7:4 to 11:0, with a mean age of 9:6 (SD=1.23). The control group consisted of 9 pupils (6 boys and 3 girls), aged from 7:6 to 10:10, with a mean age of 9:4 (SD=1.11).
Participants from School 2 were 20 pupils (14 boys and 6 girls) aged from 6:7 to 11:0, with a mean age of 8:4 (SD=1.40). Participants were matched in pairs on the basis of age and gender, with one child from each pair selected at random and placed into the intervention group and the other into the control group. The intervention group consisted of 10 pupils (7 boys and 3 girls) aged from 6:7 to 10:9, with a mean age of 8:6 (SD=1.55). The control group consisted of 10 pupils (7 boys and 3 girls), aged from 6:7 to 11:0, with a mean age of 8:2 (SD=1.30).

Materials
Reading ability was measured using the Neale Analysis of Reading Ability -Revised (NARA-II; Neale, 1997 questions. Fig. 1 shows screen shots from the program including a story screen and a question screen.
[ Fig. 1  have it spoken and a definition provided. In order to work through the levels, students are required to correctly answer comprehension questions which test their ability to extract key information from texts, apply inferential thinking and assimilate complex ideas. At each level of the program, students are required to build on these key reading comprehension skills.

Procedure
Ethical approval for the study was granted by the Department of Psychology, University of Hull Ethics Committee, in accordance with British Psychological Society guidelines. Researchers involved in data collection underwent a Disclosure and Barring Service check. Consent from the schools' head teachers were gained prior to letters being sent by the school to parents or guardians of the selected children giving the full details of the study and allowing them the option to withdraw their child from it. Before any data collection took place, the children were also verbally informed of the details of the study and told that they did not have to take part if they didn't want to and that they could stop at any time they chose to. Each participant was assigned an ID code by the Reading comprehension: A computerised intervention with primary-age poor readers 14 school, which was used on the test material and computer program, in order to ensure anonymity.
The NARA-II test was administered pre and post-intervention to measure any changes in reading performance. In order to reduce practice effects and control for order effects, School 1 pupils were administered Form 2 for the pre-intervention test and they were encouraged to change these if they wished at any point. All participants in the intervention started on the Starter Level, due to their below average reading ability.
They were able to refer to the text when answering the comprehension questions and were required to pass a level in order to move onto the next. The intervention program incorporated the optional non-fiction passages, as well as the standard fiction passages, so that participants could choose the type of text that they preferred. The control group did not take part in the intervention and, instead, continued with their usual SEN provision, which varied according to individual need but mostly involved individual or small group teaching outside of the classroom, with a focus on phonics.
At the end of the intervention period, all participants were re-administered the NARA-II using the same procedure as previously stated. On completion of the study, the participants were thanked for taking part and debriefed with regard to the expectations of the study. In order that the participants in the control group were not disadvantaged by not having taken part in the intervention, participants in this group (and other pupils in the school) were given access to the program for the following half term (six-week period).

Statistical methods
Three mixed ANOVAs (analysis of variance) are used to test group differences in terms of improvements made from baseline to post-test in reading accuracy, reading comprehension and reading rate. The between-subjects IVs are group (intervention versus control), school (School 1 versus School 2) and age group (6-8 year olds versus Reading comprehension: A computerised intervention with primary-age poor readers 16 9-11 year olds), whilst the within-subjects IV is time (baseline versus post-test). The DVs are reading accuracy, reading comprehension and reading rate. Cohen's d (Standard Mean Difference) is used as a measure of effect size to compare differences from baseline to post-test for the intervention and control groups. Partial eta squared is used as a measure of effect size for the interactions.

Results
Descriptive statistics based on the Standard Scores for each NARA-II reading subscale (accuracy, comprehension and rate) for each group are shown in Table 3.
The mean standard scores for the intervention and control groups on each reading measure at baseline and post-test are shown in Fig. 2.
[ Fig. 2 about here] Due to the differences between schools in the amount of time spent each week on the intervention programme, school was included as an IV in the analysis. The mean standard scores for both intervention groups on each reading measure at baseline and post-test are shown in Fig. 3.
[ Fig. 3 about here] Due to the wide age range of the children in this study, age group was included as an IV in the analysis, with comparisons being made between the mid primary children (6-8 year olds) and the late primary children (9-11 year olds). The mean Reading comprehension: A computerised intervention with primary-age poor readers 17 standard scores for both age groups from the intervention sample on each reading measure at baseline and post-test are shown in Fig. 4.
[ Fig. 4 about here] Prior to any further analysis taking place, the data were checked for normality (using the Kolmogorov-Smirnov test), homogeneity of variance (using Levene's test) and outliers (using boxplots). The Kolmogorov-Smirnov test shows that the distribution of data for each of the variables are not significantly different from a normal distribution. Levene's test shows that the assumption of homogeneity of variance has not been broken for any pairwise comparison on any variable. Two outliers were identified (a high score on pre-intervention reading comprehension for an intervention group participant and a high score on pre-intervention reading rate for a control group participant). The effect of these outliers will be shown below.

Reading accuracy
The ANOVA for reading accuracy shows a significant main effect of time  Table 3 There is no main effect of age group on reading accuracy (F[1,30]=.03, NS).
There is also no significant interaction between age group, group and time

Reading comprehension
The ANOVA for reading comprehension shows a significant main effect of time There is a significant interaction between school, group and time (F[1,30]=10.59, p<.01, ƞ 2 p=.26). As seen in Table 3

Reading rate
The ANOVA for reading rate shows no main effect of time ( There is also no significant interaction between school, group and time (F[1,30]=0.18, NS, ƞ 2 p=.01). As seen in Table 3, the control groups' means do not significantly showing the effectiveness of computerised reading intervention programs (Gibson et al., 2011;Kim et al., 2006;Torgesen et al., 2010). These findings support the use of silent reading (Kuhn, 2004;Reutzel et al., 2012), repeated reading (Armstrong & Hughes, 2012;Karweit & Wasik, 1996;Therrien et al., 2006), controlled decodable vocabulary and vocabulary instruction (Karweit & Wasik, 1996;Lovett et al., 2003), reading along with the computer (Basil & Reyes, 2003;Swanson et al., 2011), increasingly difficult texts (Chard et al., 2002), question and answering (Karweit & Wasik, 1996) and immediate corrective feedback (Chard et al., 2002;Hall et al., 2000) as an effective combined method of improving reading comprehension. However, due to the nature of the intervention, it is not possible to ascertain which of these techniques is responsible for the improvement in reading accuracy and comprehension.
As expected, the intervention group in School 2, receiving two Comprehension Booster sessions per week, showed significantly bigger gains in reading accuracy and reading comprehension than the intervention group in School 1, who only received one Comprehension Booster session per week. This supports the finding of Galushcka et al.
(2014) in their meta-analysis of reading trials, that interventions with higher amounts of treatment are more effective in improving literacy skills. The only effects of age are that the younger age group (6-8 year olds) showed a bigger increase in reading accuracy, as a result of the intervention, whilst the older age group (9-11 year olds) showed a bigger increase in reading comprehension. Both groups remained stable on reading rate. The National Reading Panel report (NICHHD, 2000) shows that interventions are not Reading comprehension: A computerised intervention with primary-age poor readers 23 equally effective for different age groups and providing an intervention to children within a wide age span is not recommended (Galuschka et al., 2014). However, the Comprehension Booster program has been shown to be effective in increasing reading comprehension in both age groups within this study, suggesting that it is suitable for the 6-11-year-old age range.
It was expected that reading rate would not significantly improve for either the intervention or the control group. However, the results show that reading rate actually increased for the control group, but remained stable for the intervention group. The NARA manual states that parallel form reliability is lowest for reading rate (0.66), which may have impacted on these data. Nevertheless, Carver (1992) found that readers adjust their reading speed according to the goal -speeding up if they are scanning for a particular word and slowing down if they want to memorise concepts, with a middle gear which is optimal for comprehension. The participants in the intervention group may have become more familiar with completing comprehension tasks as a result of the intervention and, as they knew they were going to be asked questions on the passages, may have slowed down to improve their comprehension.
The effect sizes (Cohen's d) comparing the differences from pre to post intervention between the intervention and control groups are 0.72 (medium effect) for reading accuracy and 0.78 (medium effect) for reading comprehension. These exceed the effect sizes reported by Swanson (1999), who carried out a meta-analysis of intervention studies with children with learning difficulties and found average effect sizes of 0.62 for word recognition and 0.45 for reading comprehension for studies using standardised (rather than experimental) outcome measures. They also surpass Torgesen et al.'s (2010) reported effect sizes of 0.53 (word reading accuracy) and 0.40 (passage Reading comprehension: A computerised intervention with primary-age poor readers 24 comprehension) following a longer and more intensive intervention (four 50-minute sessions per week for a 7-month period). The effect size for reading comprehension in the current study also far exceeds the 0.50 effect size found by Kim et al. (2006) for passage comprehension and the small effect size (0.18) found by Galuschka et al. (2014) for the three reading comprehension trials within their meta-analysis.
However, Torgesen et al. (2001) suggest that effect sizes merely describe the advantage in reading growth for children in an intervention group compared to a control group -they do not tell us anything about rate of normalisation of reading skills. Additionally, the study is limited to one geographical location, with the schools both being in an area of disadvantage, and so the findings may not be generalisable to children from more affluent backgrounds. Furthermore, the study only involved English participants. The requirements for acquiring proficiency as a reader vary between Reading comprehension: A computerised intervention with primary-age poor readers 26 different orthographies (Marinelli, Horne, McGeown, Zoccolotti & Martelli, 2014). Marinelli et al. (2014) found English (opaque orthography) readers to be more variable in terms of reading speed than Italian (transparent orthography) readers. They suggest that this is a specific characteristic of English readers as the opaque orthography and complex syllabic structure mean that readers use a variety of different strategies, resulting in differential reading efficiency. Therefore, different results to those found in the current study, particularly with regard to reading rate, may be gained in studies with participants using transparent orthographies.
The second set of limitations involves the intervention itself. The study utilised a short six-week intervention and no long-term follow-up data were collected. Although the scores of the intervention group significantly increased on reading accuracy and comprehension, scores were still not at the level expected for chronological age. A longer intervention would be required to see if this is possible and follow-up data is necessary to establish whether the improvements remain beyond the intervention period.
Also the intensity of the intervention in the current study was not standard across the two schools and a more intense intervention (e.g. 2 x one hour sessions per week) could be utilised. Furthermore, the current study only investigated one intervention and so it is not possible to directly compare its effectiveness to other interventions.
The third set of limitations relates to the outcome measures utilised. The study only measured reading pre and post intervention and did not look at changes in other variables, such as word reading, working memory, vocabulary, spelling, listening comprehension, reading confidence and reading motivation. Spelling and reading difficulties are likely to be caused by the same underlying deficits, so an intervention targeting reading deficits may positively impact on spelling ability. Indeed, Saine et al.
Reading comprehension: A computerised intervention with primary-age poor readers 27 (2013) found that a computer-assisted reading intervention raised the spelling levels of at-risk children to that of mainstream children.
Future research is required to compare the Comprehension Booster program to other conventional and computerised intervention programs using a larger sample, conduct continuous testing throughout an increased intervention period in order to determine the optimal length of program through the use of growth modelling, continue testing beyond the intervention period in order to determine the length of intervention necessary to prevent a 'relative decline' in reading after the intervention has been withdrawn, and administer a wider range of outcome measures to determine the effect of the intervention on word reading, spelling, vocabulary, working memory, listening comprehension, reading confidence and reading motivation.

Conclusions
Computerised reading programs that are shown through empirical research to be effective in improving reading skills are useful as additional tools for pupils with reading difficulties, particularly in disadvantaged areas where resources are limited and family support in reading lower. The gains shown in reading accuracy and comprehension in the current study are significant in only a short intervention period of six weeks. The program used here can be integrated within the primary school day, carried out on a group basis in networked computer rooms with minimal supervision, and is individualised. Children at risk of reading difficulties can be identified at school entry and could benefit from the use of a computerised reading intervention program.
Reading comprehension: A computerised intervention with primary-age poor readers 28 However, this is not a replacement for good teaching, and regular monitoring of children with reading difficulties is required.