According to the Dyslexia Association in Australia (2014), teachers need special training to enable them to accurately identify students who are struggling with reading and written language before attending to them using the special behaviour modification skills to improve their reading skills. This research was informed by Skinner’s theory of reinforcement. The theory was founded by Skinner (1957), as cited in Skinner (2011). Skinner’s reinforcement theory attempts to explain how behaviours of organisms are acquired and modified to improve a certain behaviour. Skinner believed positive reinforcement was more useful in modifying and reinforcing an already existing behaviour than when punishment was used. This theory was relevant to this study because it shows how reinforcement can be used with behaviour modification methods to improve reading abilities. From Skinner’s theoretical perspective, teachers can modify the reading ability of dyslexics by using behaviour modification practices such as coaching, shaping, prompting and modelling. Neitzel and Wolery (2010) define prompts as gestures, instructions, touches or things to make children correct responses independently. Mcleod (2015) defined shaping practice as a progressive way of achieving a desired behaviour by repeating a behaviour several times. Coaching can be defined as a means of transforming a research into a practice (Bowgren & Sever 2010). Cunningham and Allington (2010) define modelling as the technique of modifying reading ability by observing how teachers or models read the letters and their corresponding sound and then imitating the same.

Literature on various behavioural reinforcement practices and their effectiveness exists. For example, in Turkey, Backhaus, Jeske, Poinstingl and Koenig (2017) established that use of antecedent prompts was effective in transferring skills taught to young females. Similarly, Wong et al.’s (2016) study in the United States of America showed that the teachers’ conceptions of ‘Nature of Science’ improved significantly after two semesters of instruction. In New Zealand, Hayes (2010) found that prompting was effective in teaching autistic children. Allenger’s (2015) study in the United States of America showed that there were no significant differences between students’ word productions and sentence lengths with the teacher’s writing prompts. In India, Jeyasekaran’s (2014) study reported that there was a positive relationship between use of a visual, kinesthetic, auditory and tactile teaching method to improve the reading skill of children with dyslexia. Fonger and Malot’s (2018) study in the United States of America established that use of shaping was effective in teaching young children with autism (ASD). Mc Clurg and Morris (2014) in the United States of America reiterate that rewards are effective in motivating students to work hard.

In an experimental study with dyslexic learners, Ooko, Aloka and Koweru (2019) reported that there was a statistically significant positive relationship between shaping and reading abilities (R² = 0.109). Majcharazak, Wagner and Yatez (2013) established that shaping is affected by three knowledge resources, which include adding domain knowledge to a Wiki. Heineke (2013) agreed that coaching can lead to improved teacher learning. Dusenbury et al. (2010) found that coaching was effective in enhancing the quality of implementation of drug abuse prevention programmes. In the United States of America, Joseph et al. (2015) revealed that a variety of strategies were used to teach self-questioning to students. In another study, Loh (2009) found that there was a significant positive relationship between teacher coaching and reading outcome. In the United Kingdom, Holliman et al. (2010) established that prosodic sensitivity is important in models of literacy development. Piper and Zulkowski (2015) found a statistically significant relationship between coaching and reading effect. In Kenya, Chacha’s (2018) study indicated that use of picture prompts was useful in the teaching of oral skills in prescholars. On the contrary, Ambrose and Cheong (2011) in Malaysia showed that the Clay Modeling Program does improve the reading behaviour of dyslexic children. Similarly, in Germany, Schukajilow, Krug and Rakoczy’s (2015) study reported that prompting students to find multiple solutions does not improve their performance directly. Maguire’s (2015) study in New Zealand showed that the video self-modelling intervention did not influence the participants’ reading habits.

According to Cassidy (2015), dyslexic learners can reach their full potential if remediated early as dyslexia is not a disability. In response to this, whole-word reading and phonetic methods have been employed to make a learner master reading in Kenya (Marima 2015). The phonetic method involves reading the sounds of letters and matching them to what each letter stands for. However, little has been realised on the efficacy of the phonetic method. According to Chitiyo and Wheeler (2009), behaviour can be modified by changing classroom routine and incorporating strategies such as shaping, modelling, coaching and reinforcement to improve behaviour. Dutta and Bhakta (2016) reiterate that behaviour modification is a way of using operant conditioning to replace undesirable behaviour with more desirable behaviour. The association between dyslexia and behaviour issues cannot be underscored, hence the need for behaviour modification to cater for such learners (Hulme &Snowling 2011). In Kenya, there are schools to cater for those learners with emotional and behaviour issues in approved schools, although the schools do not cater to the dyslexic learners (Muthee, Murugami &Tekle 2015).

It is estimated that about 10% of schoolgoing learners in Kenya could have dyslexia, although no official statistics are available for all regions in the country (Ooko, Aloka & Koweru 2019). Generally, there is lack of research in the area of dyslexic learners in Kenya and possible interventions are also not documented from the available research. One of the recent available studies by Ooko and Kaluyu (2016) found out that there is a significant weak positive relationship between writing dyslexia and academic performance amongst learners in upper primary public schools in Changamwe sub-county, Kenya. The study by Ooko and Kaluyu (2016) recommended a formulation of education policies catering to screening, teaching, learning, assessment and examination of dyslexic pupils in public primary schools in Kenya. Thus, the present study made follow-up to develop an intervention to enhance reading abilities amongst dyslexic learners in primary schools. The findings of the present study are significant because they would add to the body of literature and also be used by policymakers to provide early intervention programmes for pupils with dyslexia. The present study investigated the use of selected behaviour modification practices to enhance the reinforcement of reading abilities amongst dyslexic learners in primary schools in Kenya.

The research hypotheses of the study are stated as follows:

The study was based on a quantitative research paradigm. The paradigm was adopted because it helped to make comparisons (Babbie 2010; Creswell 2014) of the four groups to ascertain the effects of the intervention of behaviour modification techniques. Specifically, the Solomon four research design was adopted. According to Solomon (1949), this is a randomised experimental design consisting of two treatments versus two control groups. The sample size was subdivided into four groups. In the first group (experimental 1) a pre-test was given to test the reading ability of the learners but there was no post-test. In the second group (control) a pre-test was given and a post-test to check if the intervention had an effect on their reading ability. In the third group (experimental 2) no pre-test was given; intervention was given and a post-test was given. In the fourth group (control) there was no pre-test or intervention but there was a post-test to check if they could read the words in the reading test effectively. This design was chosen for the study because it makes it possible to examine both the main effects of testing and the interaction of testing and treatment; thus it enabled to assess the presence of pre-test sensitisation (Fain 1999). A diagrammatic representation of the Solomon four research design is presented in Table 1.

The study was carried out in four public primary schools in Changamwe sub-county, Kenya. The selected schools had a high number of dyslexic learners and had not been performing well in the final examinations (Ministry of Education 2017). The researcher carried out the research in the sub-county as some cases of dyslexia had been identified. The selected primary schools were mainstream and inclusive in nature, and the teachers had no structured interventions on literacy. The Bangor Dyslexia Test was used to screen those confirmed to have dyslexic characteristics. The Hardin Simmons was used to rescreen learners identified with dyslexic characteristics.

The target population was 3267 learners and 54 English language teachers from seven schools. The researcher with the help of the 54 English language teachers selected a purposive sample of 229 dyslexic learners using the Hardin Simmons University (2014) Dyslexia Characteristics tool. The learners were in grades 5–8 and in age range from 10 to 15 years and consisted of both male and female learners. Purposive sampling is the deliberate choice of an informant because of the qualities the informant possesses. This sampling technique was adopted because it helped the researchers decide the individuals who can and are willing to provide information by virtue of knowledge or experience (Lewis & Sheppard 2006). The researcher further screened the learners using a Bangor Dyslexia Test (Miles 1997) to make sure those whose poor reading ability was caused by other factors were not included in the final sample.

The intervention on behaviour modification was developed on four practices, namely, prompting, shaping, coaching and modelling. Prompting was done by giving hints on letters or words where learners got stuck. For each word read correctly a tick was indicated in the score sheet. The learners were reinforced by use of praises and tokens when they performed the action correctly. For words which were prompted and they never managed to read correctly, a dot was used for marking on the scoresheet so that the following day they could be taken through the same process to read them. Shaping was done by progressively introducing words and sentences to the learners. Activities were picked from one part of the book Dyslexia Workbook by Morris (2012) daily and presented to the learners. The teacher then demonstrated to them how to read the words before they were asked to read them out on their own and observations were made on what they can read in a scoresheet. This was done by chunking the words into smaller portions and training the learners to read them out in smaller chunks. Where a learner had difficulty identifying the word from the list given, a ruler was used as a graphic organiser to help them locate the word in reference.

Coaching was done by the teacher acting as a coach and training the learners to read the words aloud. Every day activities were picked from one part of the book labelled Dyslexia Workbook by Morris (2012). The teacher then demonstrated to them how to read the words before they were asked to read them out on their own and observations made on what they could read in a scoresheet. Modelling was done by a teacher acting as a model to demonstrate how to read out the words by mapping out the sounds and the letters that they represented. Teachers carried out reading activities that were modelled to the learners. If a learner made a mistake in reading out the word aloud, the teacher corrected the learner by showing him or her how to read it out correctly and gave immediate feedback. Everyday activities were picked from one part of the book labelled Dyslexia Workbook by Morris (2012). The teacher then demonstrated to them how to read the words before they were asked to read them out on their own and observations made on what they could read in a scoresheet.

The Bangor Dyslexia Test was used to screen those confirmed to have dyslexic characteristics. The Hardin Simmons was used to rescreen learners identified with dyslexic characteristics. The Behaviour Modification Questionnaire had items on intervention for dyslexic learners, which included prompting, shaping, coaching and modelling. In addition, the researcher gave a short reading comprehension test to assess learners’ reading ability and a writing test to assess learners’ ability to organise ideas. External validity of the questionnaires was ensured by piloting the instruments in one non-sampled school using 20 learners and 4 teachers. The internal validity results of Bartlett’s test for sphericity were significant (p < 0.001, p = 0.000) and Kaiser–Meyer-hold Olkin indexes were 0.730, 0.685, 0.674 and 0.819 for all >0.6 for prompting, shaping, coaching and modelling subscales of the questionnaire, respectively. Reliability of the questionnaires was ensured by Cronbach’s alpha method and all the sub-scales met the required level of internal consistency of reliability with Cronbach’s alpha values ranging from 0.673 to 0.807.

Permission to carry out the research was first obtained from the National Council for Science, Technology and Innovation in Kenya. The instruments were pilot tested in one non-sampled school using 20 learners and 4 teachers. The researcher then sought consent from parents or guardians of the sampled dyslexic learners. The English teachers were trained for 1 month on behaviour modification practices and intervention practices. Learners who exhibited dyslexic characteristics were then screened by the researcher using the Bangor Dyslexia Test. The researcher subdivided the learners listed as dyslexics into four groups. The teachers helped in giving reading activities to the dyslexics to test the entry behaviour of the dyslexics for the pre-test scores. In one of the groups an intervention was given before the post-test. The trained teachers taught the dyslexic learners twice a week for 40 min a session for a period of 12 weeks using the interventions for the two experimental groups. After the intervention period, a post-test was given to the learners in two groups.

Descriptive statistics such as mean, frequencies, percentages and standard deviations were used to analyse the data. In addition, inferential statistics such as t-test, Pearson’s product moment correlation and regression analyses were used to analyse data. Shapiro–Wilk’s test (S–W) was used to interpret the normality of the variables for small and medium samples up to n = 2000 (Creswell 2014). It was evident that all the other variables followed normal distribution given that there were no statistically significant differences noted in any of the variables with their corresponding normal scores. Therefore, Pearson’s product moment correlation and regression analysis were used in the analysis.

This study was approved by the National Commission for Science Technology and Innovation in Kenya. The ethical clearance permit number is NACOSTI/P/19/5933/27531.

Consent was sought from the parents before screening of the learners was done. The parents then signed the letters of informed consent and returned them to the school before the learners participated in the study. The learners’ real names were not revealed to ensure anonymity, whilst those taking the Bangor Test were allowed to opt out voluntarily if they did not want to participate in the study. The interventions were carried out in classes to ensure confidentiality.

Information on the age distribution of respondents was sought and the results are presented in Figure 1.

The findings of the study show that the majority of the learners who took part in the survey were aged between 12 and 13 years; this was not surprising because although the study targeted learners from grades 5 to 8, most of the sampled learners were in grades 6 and 7. In Kenya, learners in these classes are mostly aged between 12 and 13 years. However, other age groups were also represented in the study.

The study examined differences in pre-test and post-test scores on English language reading ability between the dyslexic learners in the experimental and control groups. To achieve this, a paired sample t-test was used to determine the difference in English language reading ability between the experimental and control groups. The different combinations of pre-tested and un-pre-tested groups with treatment and control groups allowed the researcher to ensure that confounding variables and extraneous factors did not influence the results on English language reading ability. Four groups of learners were considered and were labelled 1, 2, 3 and 4, respectively. Out of the four groups, group 1 and group 3, the experimental groups, were given treatment by training them on behaviour modification skills. On the contrary, group 2 and group 4 were not treated but only received the traditional teaching of English language reading. However, group 1 and group 2 were pre-tested before and post-tested after, whereas group 3 and group 4 were only post-tested. The results in Table 2 show the descriptive statistics analysis.

From Table 2, it is evident that the highest mean score recorded was 33.1 (SD = 8.1) in the post-test reading by group 1 learners who received special treatment on behaviour modification strategies after a pre-testing. It was followed closely by the mean score of group 3 learners, who received treatment but were post-tested without pre-testing, at 32.4 (SD = 7.9) in a post-test examination. The least score recorded was pre-test results (n = 54, M = 24.13; SD = 9.2) for group 2 students, those who did not receive any special treatment on behaviour modification strategy. Figure 2 further shows the relative difference in mean scores of the various groups.

It is evident from Figure 2 that groups that received treatment reported relatively higher abilities in English language reading than their counterparts, who did not receive treatment. However, to investigate whether there is any statistically significant influence of behaviour modification strategies in English language reading abilities, three different steps involving the use of t-test were applied and findings were compared. Table 3 shows a comparison between the post-test scores attained by group 3 and group 4 learners.

The results in Table 3 show paired sample t-test investigating solution with the post-test only design with non-equivalent control groups. From the results, it can be concluded that there is a significant difference between the experimental group (group 3) and control group (group 4), (t (48) = 3.438; p = 0.001 < 0.05). Given that the difference is statistically significant at 5% level, it was concluded that the behaviour modification strategy is effective in improving English language reading abilities amongst the primary school learners. However, it is not known whether the existing difference in English language reading abilities is exclusively because of use of behaviour modification strategies or any other superseding variable that is not included in the survey. Therefore, the study further explored solution with the two control group design, as refinement over the finding, as shown in Table 4.

From the results in Table 4, a paired sample t-test on pair 2 suggests that there was no difference established between before and after values in the control group (t [53] = –1.321, p = 0.192 [ns]), but a test on pair 1 reveals that there is significant difference (t [48] = –15.059, p < 0.001) between the pre-test and post-test score of the experiment group, which means a significant impact of treatment was established on the experimental group. Equally, test 4 further confirms that there is significant difference at 1% significant level between the experiment group post-test (group 1) and control group post-test (group 4) (t [48] = 3.263, p = 0.002).

In addition, pair 3 suggests that the randomisation process was successfully applied to get samples for the experimental and control groups. This was implied by the fact that there was no significant difference (t [48] = 435, p = 0.666 [ns]) established between the experimental group pre-test (group 1) and control group pre-test (group 2). Hence, assuming that pre-testing has no effect on post-test results, it can be taken that the use of behaviour modification strategies is effective in improving English language reading skills amongst primary school learners. However, there may be some kind of effect of pre-testing on post-test scores because the mean difference increased from –8.245 to 6.735 from pair 1 to 4. This was confirmed through the use of solution with the four control group design, whose results are shown in Table 5.

From Table 5, a paired sample test for Pair 2 suggests that there was no statistically significant difference in reading ability found between before and after values in the control group, control group pre-test (group 2) and control group post-test (group 2) (t [53] = –1.321, p = 0.192 [ns]). On the contrary, test results for pair 1 reveal that there is statistically significant difference at 1% significance level between pre-test and post-test scores of the experiment group 1 (t [48] = –15.059, p < 0.01)], implying that a significant effect was found in the use of behaviour modification strategies in improving learner English language reading skills. Furthermore, from the test in pair 3 it was concluded that the randomisation process was effective during sampling of the experiment and control groups because no significant difference was found between control group pre-test and experimental group pre-test (p = 0.666).

The study also examined the effect of selected behaviour modification practices to enhance reinforcement of reading abilities amongst dyslexic learners. This was done by the use of multiple regression analysis, where all four behaviour modification practices were factored in the model. The multiple regression also provided information about the relative contribution of each of the variables that make up the model. Each aspect of behaviour modification practices was evaluated in terms of its predictive power, over and above that offered by all the other behaviour modification practices. It enabled the researcher to know how much unique variance, in the dependent variable, each of the independent variables explained. This is shown by coefficients values in Table 6.

In Table 6, the variable for block 1 is group of the respondents (experimental or control), which was controlled for in the analysis, while block 2 represents the predictor variables (modelling, prompting, shaping and coaching) together with the control variable (respondent group). It is evident that the respondent’s group alone accounted for 10.9%, as signified by coefficient of R² = 0.109, of the variation in the level of English language reading abilities amongst primary school learners with dyslexia. However, after the aspects of behaviour modification practices were included in block 2, it is clear that the model as a whole explained 54.7% (R² = 0.547) of the variability in the level of English language reading abilities amongst primary school learners with dyslexia. R-square change (0.439) in block 2 indicates the additional amount of variance accounted for by the behaviour modification practices after that explained by the control variable (respondents group) was removed. This indicates that behaviour modification practices accounted for 54.7% (R² = 0.547), translating to addition of 43.9% in the variability of in the level of English language reading abilities amongst learners with dyslexia after the effect of the respondents’ group has been statistically removed.

The study further sought to develop a regression model for the relationship between the level of English language reading abilities and behaviour modification practices amongst learners with dyslexia amongst primary school learners. This model was appropriate because each of the explanatory variables was independent and non-mutually exclusive. Table 7 shows the coefficient values of each aspect of behaviour modification practices:

In this model: Y = β₀ + β₁X₁ + β₂X₂ + β₃X₃ + β₄X₄ + β₅X₅ + ε,

where: Y is English language reading abilities

X₁: group, X₂: prompting, X₃: shaping, X₄: coaching, X₅: modelling.

The predicated optimum level of English language reading abilities was presented by:

0.154 units – 1.927 X₁ units + 0.512 X₂ units + 0.294 X₃ units + 10.774 X₄ units + 1.384 X₅ units + error.

From the model presented in Table 7, the coefficients indicate how much English language reading abilities amongst primary school learners with dyslexia varies with each practice when other behaviour modification practices are held constant. A learner who was treated on behaviour modification practices would perform better in English language reading abilities than the ones who were in the control group by 1.927 units. It emerged that coaching behaviour modification practice had the highest influence on English language reading abilities. This was reflected by the unstandardised coefficient, X₄, which is equal to 10.774, meaning that for each one unit increase in coaching behaviour modification practice, there is an increase in English language reading abilities amongst the primary school learners of 10.774 units. On the contrary, shaping behaviour modification practice reflected the least effect on English language reading abilities. A unit increase in perceived behaviour control would only result in 0.294 units’ improvement in English language reading abilities amongst primary school learners. However, in general, it was concluded that the model was adequate to predict English language reading abilities amongst primary school learners as it accounted for a fairly large variability of 54.7% (R² = 0.547).