Gender.

There is evidence to suggest that fewer girls are diagnosed with dyslexia than boys. It has been found that in most school systems, girls tend to be better at reading [24, 25]; whilst there is also evidence that more boys than girls have the phonological awareness difficulties commonly associated with dyslexia [26]. Miles, Haslum and Wheeler [27] present evidence which shows that more males have reading disabilities. Furthermore, results from the Organisation for Economic Co-operation and Development’s (OECD) Programme for International Student Assessment (PISA) [28] shows sex differences in reading skills across countries, suggesting that sex differences in reading are not down to educational practices or language orthography and that they are likely biological in nature. Further research has shown that sex differences in reading skill are more apparent at the lower end of the reading distribution [29, 30], suggesting that biological sex may be involved in reading difficulties such as dyslexia.

However, in studies of those that have been referred for a dyslexia assessment, the male:female ratio ranges from 3:1 to 5:1, whereas, in epidemiological studies of dyslexia, a smaller disparity has been found (1.5:1 to 3.3:1) [31]. This suggests an over-referral of males for dyslexia assessments compared to females. Shaywitz [32] spoke of the ‘myth’ of male vulnerability to reading disability (p.98). Shaywitz suggests that the reason more boys than girls receive a diagnosis of dyslexia is that girls are often less obtrusive and attention-seeking [32]. Therefore, boys are brought to the attention of teachers more easily, causing a disproportionate referral of boys. It could be argued that these characteristics may interplay with biological academic ability in leading to dyslexia identification.

This raises interesting questions about the biological differences found between boys and girls, and how this may interact with environmental norms with regards to displays of gender. What is unclear here is whether there are biological differences which explain why more boys than girls have been found to have dyslexia, or whether there are environmental aspects at play which influence the perception of difficulties in boys and girls differently. Arnett et al. [29] suggests that the overrepresentation of males can be due to two factors ‘one invalid part explained by referral bias, and one potentially valid residual part found in epidemiological samples’ (p.719). To follow this up they studied sex differences in a sample of 2,399 7- to 24-year-olds and found a small male:female disparity (1.15:1). They concluded that ‘the higher prevalence of males with reading difficulties can be explained by a combination of males’ slower and more variable processing speed and worse inhibitory control, although these are partly offset by males’ better verbal reasoning’ (p.726). These findings suggest that there are valid sex differences between males and females; however, the extent of these differences is not as large as the referral rates suggest. Therefore, there appears to be a complex system in the referral, diagnosis and labelling of dyslexia. While there may be some evidence for a biological underpinning in the differences between males and females, this does not entirely explain the large discrepancy in referrals for diagnosis.

Age in year group.

Donfrancesco et al. [33] found that in Italy, those who were younger in the year when starting school were more likely to be identified as dyslexic. However, the relationship between age in year group and the dyslexia label remains unexplored in the UK.

The impact of age in year group has become a growing area of interest; Crawford, Dearden and Greaves [34] produced a report on the effect of being born in the spring or summer and therefore being young in the academic year in England. They found ‘large differences in educational attainment between children born at the start and end of the academic year in England’ [34]. These differences decreased as the child got older, but they state that the gap remains significant throughout compulsory schooling. Large-scale longitudinal datasets are the best ways to follow these trends as they can assess the impact of the month of birth over an individual’s life course. Research using the MCS found that older children in the year group were significantly more likely to be placed in the highest set, whereas those who were younger were more likely to be in the lowest set [35].

Relationships have also been found between a child’s birthday and SEN. Crawford, Dearden and Greaves [34] show that ‘relative to those born in September, those born in August are 5.4% more likely to be labelled as having mild special educational needs at age 11’ (p.2). Furthermore, Zoega, Valdimarsdóttir and Hernández-Díaz [36] researched the relationship between attention deficit hyperactivity disorder (ADHD) and month of birth. They found that children in the youngest third of a class were 50% more likely than those in the oldest third to be prescribed stimulants between ages 7 and 14. They conclude that the effect of age should be considered in the diagnostic process of ADHD.

Therefore, it is of interest to investigate whether children in England and Wales with the dyslexia label are more likely to be younger in their year group. As there is unlikely to be a biological reason why someone who is younger in their year group would be more likely to have dyslexia, should this be the case, it will highlight the importance of social processes in who is labelled with dyslexia.

Ethnicity.

Evidence from the USA shows ethnic disproportionality in SEN labelling [37], whereby those from minority ethnic groups are more likely to be labelled with a SEN. Yet Morgan et al. [38] found that when controlling for confounding variables, including individual child-level academic achievement and behavioural functioning as well as family-level socioeconomic status, minority children are less likely than otherwise similar White, English-speaking children to be identified as disabled and receive special education services in the USA. Strand and Lindsay [39] conducted a study of SEN and ethnicity in England using school census data and found that while being a member of a minority ethnicity was a predictor of SEN, gender and poverty were stronger predictors. They also found differences between SEN types and minority ethnic groups. While they did not look specifically at dyslexia, analysis of those with a specific learning difficulty (SpLD) showed that when controlling for gender and poverty, those from Mixed White and Asian, Indian, Pakistani, Bangladeshi, Other Asian, and Chinese ethnic groups were significantly less likely to have a SpLD than White students [39]. In addition, Parsons and Platt [11] found that white children in the MCS were more likely to be considered dyslexic when controlling for indicators of deprivation. This, therefore, suggests an overrepresentation of white people being considered as dyslexic.

Social class.

A highly debated topic is the impact of social class on the identification of dyslexia. Media reporting in the UK has often promoted a notion that dyslexia is a ‘middle class’ phenomenon, present in fee-paying schools in particular [40, 41]. Yet, a 2010 policy paper on special education by the UK government noted the strong correlation between having SEN and being from a disadvantaged background [42]. However, within the wider SEN category, dyslexia was not found to be correlated with higher levels of disadvantage [11]. Tomlinson [43] states that

a relatively undocumented theme is that much of the expansion of special education categories and demands for funding and resources have, from the 1980s, come from middle class and articulate parents who […] have seized on expanding ways of defending children in need of special education and support

(p. 273–274).

Therefore, it is relevant to consider the impact of social class on dyslexia identification, and in particular, the impact of belonging to the middle class.

Reay, Crozier and James [44] claim that ‘above all, the distinguishing feature of the middle classes is a particular set of values, commitments and moral stances […] such as ambition, sense of entitlement, educational excellence, confidence, competitiveness, hard work and deferred gratification’ (p.12). These values suggest that great emphasis is placed on education in middle class children. In an interview survey of working class and middle class parents, Reay [45] found that parents’ involvement in schooling was largely class-based; she argued that the middle classes are able to utilise their economic and cultural resources to ensure that there is a continuation of their children’s educational advantage.

Gillborn [46] researched the experiences of black middle class parents with children who had SEN. He found that within his participants it was usually the parent who identified the problem and sought assessment for the child. He stated that ‘this involves drawing on both their economic capital (the financially expensive specialist assessments) and their cultural and social capital (often using friendship and professional networks to help ‘negotiate the system’) (p.280); this discussion of capital builds on the work of Bourdieu [47]. In brief, Bourdieu suggested that a key divide between the middle and working class was the volume and composition of habitus possessed by those in the different social classes. Bourdieu proposed that habitus was made up of the type and amount of capital that a person has. Capital can be made up from economic capital (financial resources), social capital (the individual’s social network) and, cultural capital (the knowledge of the right cultural codes of how to behave in various social contexts). This habitus then affects how the individual interacts with the field (in this case education). Reay [45] describes how middle-class parents have cultural capital in a combination of ‘requisite skills and competencies, confidence in relation to the educational system, a previous history of being supported educationally in the home, educational knowledge and information about schooling’ (p.111). Relating to dyslexia, findings from Gilborn’s [46] study suggest that parents in the middle class are able to mobilise their economic resources in order to obtain a diagnosis for their child. Furthermore, they are able to use their cultural capital in the form of social networks and knowledge of the field to get the most out of the education system for their children.

Having a label of dyslexia can lead to adjustments in formal examinations [12]. As qualifications ‘are designed and delivered to differentiate pupils and students’ [48, p.107] and middle class parents have been shown to be motivated to ensure that their child exceeds in education [44, 49, 50] it is easy to see why gaining adjustments in examinations may be important to middle class parents. In contrast, working class parents may not have the economic resources to pay for an assessment. Furthermore, they may not have access to the networks and educational history that firstly, may inform them about the potential benefits of dyslexia identification, and secondly, may help them in obtaining this label.

Furthermore, other factors that may also be associated with the construct of social class. Parsons and Platt [11] and Anders et al. [51] show correlation between parents having lower educational levels and their child having a SEN, however this has been unexplored with dyslexia specifically. Furthermore, according to the British Dyslexia Association (BDA) the cost of assessment ‘is £540 (+ VAT) with a specialist teacher and £720 (+ VAT) with an Educational Psychologist’ [52] (prices correct as of July 2021). Given the cost of dyslexia diagnosis, the impact of parents’ income also needs to be taken into consideration.

It is of interest to understand how the norms and values of different social classes may impact whether or not a child is labelled with dyslexia and furthermore, how the current dyslexia system may be contributing to the class structures that we see in the current education system.

Dyslexia.

At age 11, teachers were asked to comment on whether the child has dyslexia. If the teacher responded ‘yes’ to ‘Does this child have Special Educational Needs (SEN)?’, they were asked to select the reason from a list which included dyslexia. The teachers identified 253 children as dyslexic.

While this variable allowed insight into whether or not the teacher believed that the child had dyslexia, it did not allow insight into whether or not the child had officially been diagnosed. Previous research has used administrative data on the National Pupil Database (NPD) in England linked with the MCS in order to establish whether the child had an officially recognised SEN [58]. However, the NPD does not provide information on the type of SEN. Rather, the NPD dataset provides and overarching category of SEN, broken down into the type of support that the cohort member has received. As this research was interested in the dyslexia label in particular, it would not have been appropriate to use this more inclusive category. Due to this, we cannot assume that the children identified as dyslexic by their teacher had been officially diagnosed as dyslexic. However, using the NPD data, it was possible to see how many of those identified as dyslexic in England, had also been officially identified with a SEN at age 11 on the NPD. Table 1 shows the number of cohort members identified as dyslexic in England at age 11, and those that are also identified on the SEN register. Of the cohort members with NPD data, 152 were identified as dyslexic at age 11. Of these 121 (79.6%) were officially identified as having a SEN on the Key Stage 2 (the legal term for the four years of schooling in maintained schools in England and Wales for pupils aged 7–11) NPD. This, therefore, suggests a fairly accurate reporting of dyslexia as a SEN by teachers at age 11.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Those labelled as dyslexic and on the SEN register at age 11, KS2.

https://doi.org/10.1371/journal.pone.0256114.t001

Age in year group.

An ‘age in year group’ variable was created by allocating those who would be the oldest in the year ‘12’ (i.e. those born in September), and those youngest ‘1’ (i.e. those born in August). However, this did not consider that parents of children who are young in their year group are able to choose to defer their school entry to the following year. Therefore, a second variable was created which provided information on how old the child was when they started school full-time. This was created using the child’s month and year of birth, along with the parents’ report of the month and year that the chid started school full time. The variable was created in months and ranged from children starting school at 35 months (2.9 years) to 70 months (5.8 years) with an average of 56 months (4.7 years). Using this information, those that started school younger than 4 years (47 months) and older than 5 years (60 months) were excluded from the analysis. This meant that age in year group, according to month of birth, could be examined.

Parents’ highest socio-economic class.

As part of looking at social class, the socio-economic classification (SEC) of each parent was derived from their occupation, using categories provided by the Office of National Statistics (ONS), the UK’s largest independent producer of official statistics and its recognised national statistical institute. These were collapsed into five main categories: ‘Managerial and professional’; ‘Intermediate’; ‘Small employer and self-employed’; ‘Low supervisors and technical’; and, ‘Semi-routine and routine’. Using this information, the five-class structure was reverse recoded and the highest SEC for the main parent and partner parent was derived. This provided a household SEC level, using the highest SEC household member’s status.

Income.

The MCS collected information on the main and partner parents’ gross earnings at each sweep. From this information, the MCS calculate the OECD-equivalised weekly family earnings. This is done by ‘dividing the total net household income, with the number of household members, according to their weight on the OECD equivalised income scale (equivalised household size) to give net disposable income’ [59] (p.49). The current research made use of the continuous equivalised income scale; this was used in order to determine how a one-unit change in equivalised income impacted the likelihood of the dyslexia label.

Parents’ highest education level.

The parents’ highest academic or vocational qualification level was calculated. The qualifications are aggregated into a five-point scale from England and Wales’ National Vocational Qualification (NVQ) level 1 (no qualifications at GCSE level) to NVQ level 5 (higher degree and postgraduate qualification). However, due to a small number of parents of children with dyslexia holding to lowest level of education (n = 14), NVQ levels 1 and 2 were aggregated to make a four-point scale.

Fee-paying school.

In each survey, the main parent was asked to state whether or not the child attended a fee-paying school. This response allowed the creation of the ‘fee-paying school’ variable.

Control variables.

Cognitive profile. Cognitive profile variables were entered as control variables. Therefore, directional estimates of the likelihood of dyslexia identification, uniquely attributable to the variables of interest, were adjusted so as to make the children as otherwise similar as possible in ability. For example, it might have been found there was a relationship between dyslexia and social class, yet without controlling for ability this relationship could be spurious. For example, research has shown that children with higher non-verbal ability (i.e., an indicator of IQ) are more likely to be diagnosed with dyslexia because many clinicians and teachers still embrace an IQ-discrepancy model of dyslexia, which requires an average to above average IQ [60–63]. Higher non-verbal ability has also been found to be correlated with higher social class [64–67]. Therefore, it might appear that there is an association between higher social class and the dyslexia label, but this is in fact an artefact of a reliance on IQ-discrepancy definitions of dyslexia.

Furthermore, it is necessary to control for verbal ability (of which a low score may be directly attributable to dyslexia) so that the difference in those identified as dyslexia, and those who are not, could not be attributed to difference in verbal ability that may also cluster within the variables of interest. Therefore, controlling for verbal and non-verbal ability in the regression analyses provides a stronger test of whether social class, and other variables of interest, are independently associated with the dyslexia label than bivariate analyses alone.

These scores were derived from cohort members’ age 8 results on the British Ability Scales (BAS) tests for word reading and pattern construction. These tests are widely validated, age-appropriate tests that have been shown to be predictive of later child cognitive performance [68].

Country. In order to control for the effect of either living in England or in Wales, the child’s country at sampling and country at interview were used. Those who did not have the same country at both sampling and the time of the interview were coded as missing and were not included in the analysis. This ensured that we were able to control for the impact of consistently living in one country.

Bivariate analysis.

Following the initial selection of variables of interest bivariate analysis was carried out using Stata. The purpose of bivariate analysis was to establish whether there were significant relationships between the aforementioned variables and the group of children labelled with dyslexia. However, bivariate analysis does not take into account any other variable which may be influencing the relationship between the two variables examined, therefore, logistic regression was also conducted.

Logistic regression.

Logistic regression analysis allowed exploration of how the predictor variables related to the dependent variable, while taking the other variables in the model into account. As the dependent variable was binary and categorical (labelled as dyslexic/not labelled as dyslexic), binary logistic regression was used. In this study ‘stepwise backwards logistic regression’ was conducted. This procedure involved all possible variables being entered into the model initially and the insignificant variables being removed individually. The backwards method was chosen as forward approaches can result in a more significant variable suppressing the effects of less significant variables so that they do not appear to be significant.