An immersive, ‘Faster Read’: a pilot, mixed-method study, developing whole-text reading comprehension and engagement with adolescent struggling readers

ABSTRACT

Reading is fundamental to academic success, but international reading surveys indicate current pedagogy fails a fifth of adolescents, disproportionately from lower-socioeconomic groups. This UK, mixed-method study evaluated the impact of two whole-text reading approaches on comprehension, using standardised tests. Twenty teachers of English and 413 students (12–13 years) participated, 44% defined as ‘struggling readers’, in parallel classes per school, matched for reading ability. Both groups read two ‘challenging’ novels consecutively over 12 weeks. Ten teachers received no further treatment; ten teachers received training on reading-comprehension pedagogies. The mean comprehension for all students increased by 8.5 months with no significant differential effect of training condition. However, there was a significant differential effect between ‘struggling’ and ‘average+’ readers in both conditions: struggling readers’ mean comprehension improved by 16 months. Required to read whole texts, teachers in both groups altered practice, increasing print exposure, comprehension-strategies and supportive discourse-strategies, benefiting ‘struggling’ over ‘average+’ readers. This pilot study reframes understandings of ‘independent’ reading for struggling readers, indicating that the teaching of whole texts and extended reading must happen in classrooms, not outside.

KEYWORDS:

• Reading comprehension
• struggling readers
• adolescents
• teacher education
• discourse
• inference

Introduction

This article reports on the findings of an innovative, mixed-method, pre-test/post-test pilot study on developing adolescents’ reading comprehension, targeting ‘struggling readers’, here defined as having a reading age one year or more below chronological age. The study was conducted primarily in the under-researched context of mainstream ‘low sets’ in England, against a history of disadvantage for students with low literacy levels, disproportionately from lower socioeconomic groups (Myhill et al. 2022). For this paper, the study’s quantitative data have been re-analysed using multi-level modelling (for further discussion of the qualitative data, see Westbrook, Sutherland, Oakhill and Sullivan [2019]). Although conducted pre-Covid-19, this study’s findings are important as the pandemic has exacerbated gaps in reading, internationally. In England, secondary students’ reading declined, on average, by 2.4%, and 3.5% for the most disadvantaged students, measured after 18 months of pandemic (EPI 2022). Schools continue to seek evidence-based ways to ‘catch-up’ struggling readers. Both trialled interventions in this study implemented a ‘Faster Read’: 20 teachers in ten schools increased the volume and pace of reading with their Year 8 classes (12–13 years, n = 413, 44% defined as ‘struggling readers’), by reading two fictional texts ‘back-to-back’ in 12 weeks (Condition 1 - ‘Faster Read’ (FR)). Additionally, half the teachers, one in each school, received a theorised programme on evidence-based pedagogy to develop comprehension, particularly for struggling readers (Condition 2 - ‘Faster Read + Theory’ (FR+T)). The quantitative analyses showed that students’ mean comprehension increased by 8.5 months in both conditions, with no significant differential effect between groups. There was a significant difference between struggling readers and average readers in each condition: the mean comprehension scores for struggling readers improved by 16 months in both conditions. Average readers improved by 1.2 months (Condition 1) and 2.7 months (Condition 2).

The qualitative data found that, as a result of Condition 1, teachers in both groups altered their usual practice: for the FR+T group, this was as expected, given the additional training. However, teachers in the FR group also unexpectedly altered their practices by being required to teach whole texts, which benefited ‘struggling’ readers the most, compared with ‘average+’ readers. This paper explains the intervention and methodology that gave rise to the measured gains, and uses qualitative data to further account for them.

This study is also methodologically interesting: firstly, its interdisciplinary, mixed-method design combines the authors’ different expertise in cognitive psychology and education, drawing on their previous research and extensive prior experience of teaching English to adolescent struggling readers in ‘low sets’. McMaster et al. (2015) emphasise the imperative of reading research synthesising such expertise from theory and practice, and of being conducted in ‘real-world’ ‘low sets’, where teacher-participants encounter inherent challenges of low motivation and behavioural difficulties. Secondly, the paper illustrates the benefits and challenges of mixed-method research (see Conclusion). Qualitative data deepens understanding of why research-informed interventions do or do not achieve measurable gains in reading comprehension, illuminating complexity, raising questions for future research and addressing design limitations.

For 20 years, international reading surveys and other studies have identified a persistent fifth of adolescents, disproportionately from socioeconomically disadvantaged backgrounds, as having inadequate ‘reading literacy’, in Organisation-of-Economic-Cooperation-and-Development countries (OECD 2019), Europe (Dietrichson et al. 2020; EACEA 2011, 7) and the US (e.g. Biancarosa and Snow 2006; IRA 2007). These adolescents are prevented from high achievement in secondary school and, as adults, are more likely to be unemployed, have low incomes poorer health and ‘health literacy’ (Morrisroe 2014, 19). In England, as elsewhere (e.g. Scammacca et al. 2015), the reading-attainment gap between ‘good’ and ‘struggling’ readers widens in adolescence when texts become more challenging: 30% of adolescents, disproportionately from disadvantaged backgrounds, fail to meet national reading benchmarks (GCSE English, Level 4+, DfE 2019). Additionally, ‘struggling’ readers’ weaknesses are compounded by not reading independently, unlike ‘good’ readers (NLT 2020). COVID-19 has exacerbated disadvantage (EPI 2022; NFER 2020), widening the ‘reading gap’: in England, ‘lockdown’-schoolwork typically did not comprise extended reading (Green 2020).

However, research on reading comprehension with adolescents has still received relatively little focus, especially with whole texts (Brooks 2016; Vaughn et al. 2015). Rashid and Brooks (2010, 6) review of international research evidence from 1948 noted a plateau in reading-development in England since 2004, with 17% of 16–19 year-olds consistently having poor reading skills. Furthermore, secondary teachers are known to lack knowledge and practice in supporting ‘struggling readers’ (Moats 2009; Risko et al. 2008; Snow 2002).

The reading comprehension process

Reading comprehension is the active construction of meaning from text and involves readers using a set of language skills and cognitive processes distinct from, but additional to, the phonological skills, letter knowledge and short-term memory required for decoding (Muter et al. 2004). For comprehension, readers must use low- and high-level processes simultaneously, including inferential processing, using domain knowledge, metacognition, and working-memory (Cain, Oakhill, and Bryant 2004; Cromley and Azevedo 2007). Research has demonstrated a close relation between vocabulary knowledge and reading comprehension (e.g. Nagy and Herman 1987; Oakhill and Cain 2012), which can be improved by enhancing oral language, including listening comprehension (Clarke et al. 2010; Perfetti, Landi, and Oakhill 2005) and by exposure to print, as word-level reading skills are supported by encountering contextualised vocabulary in diverse texts (Nation 2017).

Stanovich (1986) identified the complex, reciprocal, or ‘bootstrapping’ relations between early reading acquisition, exposure to print and reading comprehension as the ‘Matthew Effect’: those with already good reading skills are likely to improve more rapidly than those with poorer reading skills. van Bergen, Vasalampi, and Torppa (2020) demonstrated, using PISA data, that at 15 years, exposure to print, earlier reading comprehension and fluency each contributed equally and statistically significantly to reading scores. Additionally, Oakhill and Cain’s (2012) longitudinal study found that whilst receptive vocabulary and syntactic knowledge were associated with reading comprehension in 7–11-year olds, these were not predictors of unique variance in comprehension across time. Instead, early abilities in inference and integration, comprehension-monitoring, and knowledge and use of story structure predicted performance on a later assessment of comprehension, independently of the contribution of earlier reading-comprehension skill. These findings confirm the centrality of inference in comprehension. Readers must make inferences and integrate meaning between sentences, paragraphs and chapters, accessing ‘text-based’ meaning and constructing a representation of whole-text meaning (a ‘Situation model’ [Kintsch 1998]), drawing on world and intertextual knowledge. Local-cohesion inferences are necessary for basic comprehension at sentence or paragraph level, while global-coherence inferences enable the reader to infer setting or characters’ emotions, motivations and goals, making sense of the overall text (van den Broek 1997). Struggling readers frequently have poor working memory, lacking background knowledge for inference-making, and finding inference-making hard, even with sufficient knowledge (Cain et al. 2001).

Which pedagogy supports the development of reading comprehension?

Extensive research identifies teachers’ inadequate theoretical knowledge of reading and associated pedagogy Compton et al. (2014; Moats 2009; Risko et al. 2008). Researchers advocate teaching metacognitive reading strategies flexibly, developing comprehension-monitoring (e.g.Vaughn et al. 2011), which poor comprehenders find challenging (Vellutino et al. 2007), in a content-rich unit with engaging texts (Guthrie and Klauda 2014; van Rijk et al. 2017) and reading aloud, to develop struggling readers’ comprehension (Brooks 2016; Vaughn et al. 2011). Several mainly pre-test/post-test studies in the UK, cited in Brooks (2016) review of literacy programmes, show that targeted interventions can enhance comprehension for KS3 struggling readers: ‘Boosting reading’; ‘Inference training’; ‘Literature programme’; ‘THRASS’; and ‘Toe by toe’. These are typically characterised by a focus on reading strategies and inference-training with 1/1 adult support or withdrawn small groups. Brooks concludes: ‘Children falling behind their peers need both carefully structured reading material and rich, exciting texts’ (Brooks 2016, 15).

Many studies show that dialogic whole-class discourses support comprehension, enabling readers to clarify meaning, elaborate responses and generate questions, developing mental models of the text; comprehension-monitoring; inference; and reading engagement (Asterham and Beck, 2015; McGeown et al. 2015; Soter et al. 2008). In RCTs in the UK, Howe et al. (2019) and Alexander (2018) found these ‘dialogic teaching’ strategies correlated with enhanced attainment in English lessons. Structured, collaborative peer talk also enhances reading comprehension (Fletcher et al. 2012; Okkinga et al. 2018; Palincsar and Brown 1984), interpretation (Applebee et al. 2003; Sutherland 2006, 2015) and engagement for reluctant readers (Guthrie and Klauda 2014).

Print exposure, text variety and motivation

Comprehension performance is related to print exposure (see Mol and Bus 2011 for a review), and to text variety, showing a stronger relation with extended, challenging, ‘print’ texts (e.g. Torppa et al. 2020). Some studies have found that adolescents’ independent, extended fiction-book reading is a stronger predictor of comprehension skill than other types of reading (Jerrim and Moss 2019; McGeown et al. 2015; van Bergen, Vasalampi, and Torppa 2020). Others have found that independent ‘reading for interest’ or curiosity is particularly related to reading skill (Retelsdorf, Köller, and Möller 2011). In either case, struggling readers are less likely to be motivated to read, and read less independently, thus lacking practice in comprehending challenging whole texts of any type (Allington et al. 2010; Clark and Akerman 2006; NLT 2020). However, according to Swanson et al. (2016) observational study of 137 classes in the US, lesson-time spent reading in Language Arts classes is only 14.8%, remaining constant since the 1980s.

Additionally, readers must tackle whole texts at adequate pace to enable recall of past events and details, making cross-text connections to engage in required inferential and integrative processes, including global-coherence inferences (Philpot 2005; Westbrook 2013). This also enables readers to experience enjoyment at the end of the book as threads are woven into an integrated whole. Intrinsically motivated readers, who enjoy reading independently, not only read more, which partly accounts for their enhanced skills (McGeown et al. 2015), but also have a ‘drive for coherence’ (van den Broek 1997), motivating their use of reading strategies (Taboada et al. 2009). Motivation also includes ‘expectancy value’ (Wigfield and Eccles 2000) or ‘self-concept’: Retelsdorf, Köller, and Möller (2011, 551) demonstrated that adolescents’ perception of their likely success at reading, present and future, is related to their comprehension skill. Cliff-Hodges (2016) and Cremin et al. (2014, 67) rich, qualitative studies have also demonstrated the importance of motivation and self-concept, recommending that English teachers create reading communities, where ‘teachers as readers’ and students collectively develop understanding, practice and reader identities.

Therefore, English teachers must explicitly teach all the above comprehension skills, while motivating struggling readers to engage with the cognitive challenge of reading difficult, whole texts, as their reader self-concept and reading experiences are likely to be imbued with failure (Guthrie and Klauda 2014).This challenge may explain why many robust interventions with adolescent readers have not demonstrated significant comprehension gains (e.g. Baye et al. 2019; James-Burdumy et al. 2012; Vaughn et al. 2015).

Paradoxically, although the National Curriculum for England (DfE, 2014) requires Key-Stage-3 students (KS3, 11–14 years) to ‘read’ two texts a year, GCSE performance pressures for 16-year-olds have marginalised intensive, engaged, whole-text reading in English lessons. Year 8 current practice mistakenly focuses on written analytical GCSE skills, by-passing foundational reading-comprehension skills. Teachers typically read extracts or a text slowly, fragmenting it over a 12-week term and over-emphasising ‘micro’ language-analysis in written responses (Bleiman 2020). The new ‘mastery curriculum’ (Elliott et al. 2021) adopted by many Academies in England, has exacerbated these practices: whole texts taught are increasingly abridged or original Victorian novels with archaic syntax, purportedly preparing for GCSE-style questions on ‘unseen’ extracts. Neither research (Baye et al. 2019), nor GCSE examiners (AQA 2018) support this pedagogy as sophisticated reading requires motivated, whole-text comprehension of challenging and engaging texts, particularly for struggling readers.

Method

This inquiry addressed the following research questions:

1. Is there a measurable impact on adolescent students’ growth in reading comprehension under two conditions:

• (i) A faster read of two challenging texts (FR)?

• (ii) A faster read plus a training programme for teachers, providing a theorised knowledge of reading and its associated pedagogy (FR+T)?

2. What is the measurable impact on struggling readers and average + readers?

3. Which pedagogic approaches and other contextual factors contribute to any gains in students’ reading development?

The first two research questions are investigated by quantitative methods, the third by qualitative methods.

Based on previous studies, we proposed the following three hypotheses:

H1:

Students experiencing the FR+T condition will improve their reading comprehension to a greater extent than students in the FR-only condition.

H2a:

‘Struggling’ readers will improve their reading-age scores to a greater extent under the FR+T condition than under the FR condition.

H2b:

‘Average+’ readers will improve their reading-age scores to a greater extent under the FR+T condition than under the FR condition.

H3a:

‘Struggling’ readers will improve their reading-age score to a greater extent than ‘Average+’ readers under a FR condition.

H3b:

‘Struggling’ readers will improve their reading-age score to a greater extent than ‘Average+’ readers under a FR+T condition.

This was a mixed-method research design with two concurrent phases. Phase 1 involved a comparative analysis of two interventions with non-randomised, pre-test/post-test. A repeated standardised assessment was used to measure adolescent students’ reading progress (n = 413 recruited; 20 classes in 10 schools) over time (12 weeks) in two training conditions. Groups were matched for students’ initial reading ability and teacher experience/knowledge. The limitations of not having a third ‘comparison’ group, following ‘business as usual’ with no intervention at all, are acknowledged. It was originally judged, ethically, that the ‘Faster Read’ condition should not be withheld from vulnerable struggling readers, primarily in low sets, nearing public examinations and at risk of failure (Pring 2000). Therefore, a proxy control was used: the normal expectation of 12 months’ progress in a year and the normative concept of a Reading Age correlating with age, on which the curriculum in England and standardised reading tests are premised. The design was also based on using the same schools for both interventions, but not the same teacher, exploiting the typical binary structure of English State schools: two parallel, reading-ability ‘bottom-set’ or mixed-ability classes per school. Too few schools in the target area would have been sufficiently large to have three such parallel classes.

Phase 2 used qualitative research with all twenty classes and teachers implementing the two ‘Faster Read’ interventions. Data from lesson-observations and teacher interviews were used to monitor treatment fidelity and were triangulated with quantitative data to identify teacher practices, beliefs and knowledge, correlated with reading development.

Participants

A stratified-sample criteria was used to select the participants. The intervention catalysts were English teachers (n = 20) with their Year 8 mainstream classes (n = 20) in diverse, State secondary schools in south-east England (n = 10). Volunteer schools were recruited on the willingness of the Head teacher and teachers to alter their term’s curriculum, institute pre-/post-reading tests, and provide parallel classes, in terms of reading ability. Ten of the schools meeting these criteria were selected to be representative of the population in the target area (SE England), based on school characteristics: national examination data (average, above- and below-average); school-inspection rating (three schools were rated ‘Requires Improvement’, four rated ‘Good’, three rated ‘Outstanding’); and the proportion of students in receipt of the Pupil Premium (PP), a proxy for low socioeconomic status (SES). The school sample range for students in receipt of PP was 10.4% to 59.8%; the national average is 28.3% (DfE/EDF 2015).

Students (aged 12–13 years) in twenty Year 8 classes participated in the two interventions (n = 413 were recruited). Overall, 44% of students were identified as ‘struggling readers’, defined as having a reading age of ≥12 months below their chronological age. There were two parallel English classes per school: fourteen classes were classified as ‘low sets’; six were ‘mixed-ability’ (in schools with ‘mixed-ability’ grouping policies).

Data were also collected at the class and student levels for all the above categories, including reading age (RA). Although school data indicated that parallel classes were equivalent, in terms of students’ RA, the standardised pre-test demonstrated that paired classes were not matched in RA, although each of the groups in total (FR and FR+T) were matched (both had a mean reading age of 148 months, with no significant difference between them). Parallel classes also varied in size and student characteristics, affecting reliable comparison at the level of class (see below). At the student level, the overall sample included in the analysis (n = 365, that is, students taking both pre-/post- standardised tests), included a very slightly lower-than-average number of students in receipt of PP (94/365 = 26%); a higher-than-average proportion of students with SEND: 92/365 = 25% (national average is 15.4%, DfE 2015); and a lower-than-average proportion of students with EAL: 6%, compared with 15%, nationally (ibid).

Additional data were collected on teachers: teaching experience, qualifications and Continuing Professional Development (CPD) in reading. The range of teacher experience was 1–27 years. All participants (18 female; two male) had an undergraduate degree with a minimum of 50% content in English and a Postgraduate Certificate in Education. Six of the twenty teachers (three FR; three FR+T) additionally had postgraduate degrees in education or English, none related to reading comprehension; two teachers (one FR, one FR+T) had received brief CPD on reading once in their careers.

The study investigated the comparative results of two ‘Faster Read’ interventions: (i) teacher-implemented reading of two whole texts, back-to-back, in 12 weeks, to enhance students’ reading comprehension (Condition 1, ‘FR’). (ii) FR with CPD on theoretical knowledge of reading comprehension and pedagogy (Condition 2, ‘FR+T’).

Parallel classes, within each school, were assigned to Condition 1 or 2, based on teacher characteristics to achieve matched groups on covariates (range of teaching experience, qualifications and knowledge of reading) – paired randomisation (Torgerson and Torgerson 2007). Conditions 1 and 2 were operated simultaneously in each school, with the same texts and the instruction that participating teachers refrain from sharing any details of either condition. All classes received an equivalent amount of weekly lesson time (2.5 hours). As approved by the university’s Ethical Committee (ER/JCS23/1), ethical processes addressed continuing informed consent of teachers and students, anonymity and assurances that FR teachers (and thereby FR classes) would receive the study’s training at the project end.

Procedure

Condition 1

All teachers implemented Condition 1 – a ‘Faster Read’ (FR) of two, whole, challenging novels in a 12-week term for 2.5-hour English lessons, weekly. Teacher pairs, per school, were asked to select two fictional texts from those typically taught with mainstream, Year 8 classes of average+ attainment. Texts should be sufficiently challenging to support comprehension development, linguistically, thematically and structurally. Teachers selected texts likely to engage their students, including: Private Peaceful (Morpurgo 2003) and Once (Gleitzman 2006); Animal Farm (Orwell 2013) and The Curious Incident of the Dog in the Night-time (Haddon 2004) and culturally diverse texts (Now is the Time for Running, Williams 2012), with some texts selected by more than one school. FR-only teachers were briefed on Condition 1, informed of the focus on reading comprehension and asked to follow customary practice, apart from reading both novels, receiving no further input until the end of the project.

Condition 2

FR+T teachers additionally attended training (1.5 days), incorporating Conditions 1 and 2, with three aims: to develop theoretical understanding of reading comprehension; identify typical causes of comprehension difficulties with struggling readers; and explore why particular pedagogies are effective in enhancing comprehension, based on the literature summarised above. Although focusing on comprehension, training identified essential lower-level reading components (decoding, fluency, syntactic awareness and gist), demonstrating where recommended pedagogy also supported their development.

Guidelines for FR+T teachers to relate theory to pedagogy reflected programme content: reading strategies (predicting, clarifying, questioning, summarising, text-connections), comprehension-monitoring and inference (applying knowledge: ‘self’, ‘world’ and ‘text’) in guided-reading groups (GR) and whole-class, modelling and ‘think-alouds’. Additionally, guidelines required: establishing regular ‘dialogic talk’, class and group, with ground-rules; teaching story-structure with graphic organisers; strategies for vocabulary-development; and reading entire texts, combining teacher reading aloud (in class) with students reading aloud (in groups) for fluency and engagement, enabling students to form Situation models of the text. Teachers planned schemes of work independently, drawing on this training, so that the impact of theorised CPD on teachers’ knowledge and practice could be evaluated.

Measures of reading comprehension

The Hodder Access (McCarty and Crumpler 2014) assessed students’ reading comprehension at Test points 0 (start) and 1 (end) of the project. This standardised assessment is a timed (one-hour) test, comprising 60 multiple-choice questions on fiction and non-fiction texts. The Hodder Reading Test is a written assessment, enabling whole classes to be tested quickly with minimal disruption to learning, a key ethical consideration, and chosen because it has robust reliability and validity.¹ Three team members administered the tests in the same conditions (Weeks 1 and 12). Two standardised Hodder tests were used: they were administered with an AB/BA design so that half the sample received the first standardised test at pre-test and the other half the second, reversing the order at post-test.

The two researchers collecting the qualitative data were experienced teacher-educators, who drew on common practices in this context to design the semi-structured observation and interview schedules, jointly piloting instruments to strengthen reliability. The 40 observations of one-hour lessons and 40 follow-up teacher interviews (20–30 minutes’ duration) were conducted in Weeks 2–3 and 9–10. Data were audio-recorded and transcribed by the same researchers, enhancing reliability. Additional documentary evidence included units of work, lesson-plans and resources, and ‘student characteristics’ data.

Methods of data-analysis

A multi-level, mixed model procedure was used to interrogate the research propositions. The data have a hierarchical 3-level structure: schools (n = 10), class (n = 20) and students (n = 392) and a repeated measure (pre-test = 0; post-test = 1). The dependent variable was comprised of the pre-test and post-test Hodder Access scores. The fixed factor effects include: (i) the Condition group (FR and FR+T); (ii) Reader type (‘Struggling’ Reader and ‘Average+’ Reader); and two interactions (iii) Condition*Reader type (iv) Condition*Reader type*Time. The fixed factor effects of the Condition group tests H1, the Condition*Reader type interaction tests the H2 and Condition*Reader type*Time interaction tests the H3. There are two random effect blocks with the first block accounting for possible correlations between classes within the same school and the second block accounting for possible correlations between students within the same class.

The qualitative data were explored using ‘theoretical’ thematic analysis (Braun and Clarke 2006, 13) with NVivo 10. Deductive codes, based on intervention pedagogy, were used to assess treatment fidelity. Primarily deductive, with some inductive, codes identified the pedagogy, discourses and contexts relating to comprehension-development. Longer discourse-sequences were coded as ‘enabling’ comprehension-development (Sutherland, 2015; Soter et al. 2008) if student responses showed evidence of lower- or higher-level comprehension occurring. Sequences contained a high proportion of these indicators: ‘clarifying’ and ‘open’ teacher or student questioning at vocabulary, sentence or text-level; and probing follow-ups with uptake, supporting clarification or deepening of comprehension, by prompting response, development or cross-text connections, using textual reference. Final themes enabling comparison across groups are: whole-text reading engagement, comprehension processes, inference, metacognition (teacher and student), ‘enabling discourses’ and teacher knowledge. Data quotations from the rich corpus were selected as best illustrating key themes to address research questions.

In terms of treatment fidelity, seven of ten FR teachers met criteria for Condition 1, but three failed to complete reading two whole texts. All FR+T teachers met criteria for Condition 2, but varied in pedagogic skills to execute its more challenging elements.

Results

The descriptive results are presented in Table 1. Of the 413 participating students, 392 students (94.9%) completed the pre-test assessment and 365 students (88.4%) completed the post-test assessment test. Although 365 students (88.4%) completed both assessments (all the students that carried out the pre-test also completed the post-test), multi-linear models allow for cases with missing data: therefore all 392 student scores are included in the model (Twisk 2006). The average pre-test reading age for both the FR Condition group and FR+T Condition group was 148.3 months (12.4 years). The average post-test reading age was 155.6 months (13 years) and 157.5 months (13.1 years) for the FR and FR+T groups, respectively. Adjusting the results to include only those students that participated in both the pre-test and post-test, the reading age for the FR group (n = 183) improved by 7.6% (unadjusted 5.3%) and FR+T group (n = 182) improved by 8.6% (unadjusted 6.3%).

Table 1. Hodder access scores, pre-test and post-test results (reading age in months).

Condition	Reader Type	Pre-test reading age		Post-test reading age
		Mean	Std. dev	Mean	Std. dev
FR	Average+ reader	176.7	25.35	177.9	30.82
	Struggling reader	112.9	20.44	128.3	28.80
	Total	148.3	39.37	155.6	38.75
FR+T	Average+ reader	177.8	21.53	180.5	25.78
	Struggling reader	111.7	22.69	128.1	32.93
	Total	148.3	39.62	157.5	39.04
Total	Average+ reader	177.3	23.46	179.2	28.36
	Struggling reader	112.3	21.55	128.2	30.81
	Total	148.3	39.45	156.5	38.86

n = 392.

The ‘struggling’ readers and the ‘average +’ readers were evenly distributed across the two Conditions: the FR group comprised 87 ‘struggling’ readers and 108 ‘average+’ readers and the FR+T group comprised 88 ‘struggling’ readers and 109 ‘average+’ readers. The mean difference scores from the pre-test assessment (time = 0) between the two reader type groups was 63.8 months in the FR group and 66.1 months in the FR+T group. The mean difference scores from the post-test assessment (time = 1) between the two reader types decreased: 49.6 months and 51.9 months in the FR group and FR+T group, respectively. This is because the struggling readers proportionately made greater progress than the average+ readers. The mean gain for the FR group, for the average+ readers, was 1.2 months and for the struggling readers, 15.4 months. For the FR+T group, for the average+ readers, the mean gain was 2.7 months and for the struggling readers, this was 15.9 months (see Table 1).

The linear mixed-model fixed effects are presented in Table 2 and the fixed coefficients are presented in Table 3. Table 2 reports the test results for the first hypothesis. Students from both the FT group and FT+T group increased their reading comprehension scores from time = 0 to time = 1. However, there was no significant effect of the two condition groups, F(7, 749) = .498, p = .480.The test failed to reject the null hypothesis for H1, that is, the additional training for teachers (FR+T) did not make a significant difference to measurable reading scores.

Table 2. Fast reading model: Tests of fixed Effects^a.

	F	df1	df2	Sig.
Corrected Model	532.007	7	749	.000
Condition group	.498	1	749	.480
Reader type	295.146	1	749	.000
Group * Reader type	.005	1	749	.942
Group * Reader type * Time	22.828	4	749	.000

^aLink function: Identity.

n = 392.

Table 3. Fast reading model: Fixed Coefficients^a.

Model Term	Coeff.	Std. Error	t	Sig.	95% Confidence Interval
					Lower	Upper
Intercept	132.95	5.86	22.685	.000	121.44	144.45
Condition FR group	−3.38	7.98	−.423	.672	−19.05	12.30
Condition FR+T group	0^b	.	.	.	.	.
Reader type=Average+	40.27	4.07	9.893	.000	32.28	48.27
Reader type = Struggling	0 ^b	.	.	.	.	.
FR group *Average+	.29	9.08	.031	.976	−17.55	18.10
FR group*Struggling	0^b	.	.	.	.	.
FR+T group*Average+	0^b	.	.	.	.	.
FR+T group* Struggling	0^b	.	.	.	.	.
FR group*Average+*[Time = 0]	−1.36	2.32	−.586	.558	−5.90	3.19
FR group*Average+*[Time = 1]	0^b	.	.	.	.	.
FR group* Struggling*[Time = 0]	−15.60	3.86	−4.026	.000	−23.21	−7.99
FR group* Struggling *[Time = 1]	0^b	.	.	.	.	.
FR+T group*Average+*[Time = 0]	−3.08	3.23	−.953	.341	−9.43	3.27
FR+T group*Average+*[Time = 1]	0^b	.	.	.	.	.
FR+T group*Struggling*[Time = 0]	−16.95	2.80	−6.049	.000	−22.45	−11.45
FR+T group*Struggling * [Time = 1]	0^b	.	.	.	.	.

^aLink function: Identity.

^bThis coefficient is set to zero because it is redundant.

n = 392.

Table 2 also reports the test results for the second hypothesis. The Reader type (‘struggling’ and ‘average+’) was a predictor to the model, F(1, 749) = 295.15, p < .001, but this effect is definitional (‘struggling’ readers were classified to have a reading age one year or more below chronological age): ‘average+’ readers held a 64 month (time = 0) and 51 months (time = 1) reading age differential above ‘struggling’ readers.² However, the interaction between the two interventions and the two reader types was not significant, F(1, 749) = 01, p = .942.The interaction between the reading intervention and reader type groups is not a predictor for the reading comprehension of ‘struggling’ readers nor for ‘average+’ readers. The null hypothesis for H2 is not rejected, that is, additional training for teachers (FR+T) did not make a significant difference to the measurable reading scores of the ‘struggling’ readers (H2a) or for the ‘average+ readers (H2b).

Table 2 and Table 3 provide the test results for the third hypothesis (H3a and H3b). First, the interaction between the interventions, reader types and time is significant, F(1, 749) = 22.83, p < .001 (Table 2). From the test results presented in Table 3, we see a significant improvement in reading comprehension of ‘struggling’ readers in the FR condition from time = 0 to time = 1, b=−15.6, t(67.1)=-4.01, p < .001; ‘struggling’ readers also saw a significant improvement in reading comprehension in the FR+T condition from time = 0 to time = 1, b=−16.95, t(71.7)=-6.05, p < .001.However, the interaction of the FR group, ‘average+’ readers and time (t = 0, t = 1) was not significant (b = −1.36, t(40.49)= −.586, p = .558). The interaction of the FR+T group, ‘average+’ readers and time (t = 0, t = 1) was also not significant (b = −3.08, t(41.92)= −.953, p = .341). The reading comprehension scores for ‘struggling’ readers improved significantly in both reading conditions, while the reading comprehension for ‘average+’ readers’ improved, but not significantly. The third hypothesis (H3a and H3b) is accepted.

The findings arose from the fixed effects (conditions) and student variability. The variances of the intercepts for the residual effects (repeated measures) are presented in Table 4. The residual covariance parameters are statistically significant (p < .001). The positive AR1 rho indicates that student reading ages increased from the pre-test to post test period. The variances of the intercepts for the random effects are not significant for between -school variation (Var = 166.7, p = .088); and are not significant for classroom variation (Var = 40.3, p = .180) (Table 5). These estimates were small relative to the residual variance, which suggests that, in addition to the fixed effects, student variation can explain the variability in test scores rather than classroom or school variation. These results suggest that there was no knowledge leakage between classrooms or schools.

Table 4. Fast reading model: Residual Effect^a.

Repeated measures	Est.	Std. Error	Z	Sig.	95% Confidence Interval
					Lower	Upper
AR1 Diagonal Subj: Spec.: School * Class * Student	562.25	32.91	17.08	.000	501.30	630.60
AR1 Rho Subj Spec. School * Class * Student	.49	.04	12.03	.000	.40	.562

^aCovariance Structure: First-order autoregressive.

n = 392.

Table 5. Fast reading model: Random effect Blocks^a.

Random Effect Covariance	Estimate	Std. Error	Z	Sig.	95% Confidence Interval
					Lower	Upper
Var(Intercept) Subject Spec:School	166.67	97.62	1.707	.088	52.88	525.32
Var(Intercept) Subject Spec: School * Class	40.25	29.99	1.342	.180	9.34	173.37

^aCovariance Structure: Scaled Identity.

n = 392.

Addressing RQ1 (Is there a measurable impact on adolescent students’ growth in reading comprehension under each of two conditions?), the findings from the Hodder test, measuring the impact on students’ reading development of the two ‘Faster Read’ interventions, did not reveal any substantive differences. While reading scores from the FR+T group improved 1% more than FR group scores, the statistical differences between the two interventions were not significant (H1). Therefore, as a comparison factor, the two interventions were not a predictor for reading-comprehension improvements in this model. Furthermore, the FR+T condition did not improve significantly the reading scores of the ‘struggling’ readers to a greater extent than the FR condition (H2). Again, the impact of the two conditions was indistinguishable.

This was undoubtedly disappointing to the research team. However, addressing RQ2 (What is the measurable impact on subgroups: struggling readers and average + readers?), what was revelatory to the researchers were the findings from the third hypothesis (H3a and b): ‘struggling’ readers improved their reading comprehension by 16 months in 12 weeks, and they improved significantly more than the ‘average+’ readers in both FR and FR+T conditions. The national benchmark for Year 8 (12–13 years) students in England, enabling them to achieve the reading age of 15 years required by GCSE examinations two years later, is that they should make approximately 12 months’ progress in their reading scores in a year: three-four months’ progress per term. Given that ‘struggling readers’ in this study were already at least one year behind their chronological age, mean progress in standardised tests of 16 months in a 12-week term, is notable. This finding suggests that, although, unpredictably, the training programme in the FR+T group did not impact on students’ measurable reading scores, the FR intervention alone, reading two challenging novels ‘back-to-back’ over 12 weeks, even without additional teacher training, is highly supportive of ‘struggling readers’. They benefitted more than average+ readers from the rapid acceleration in volume of words read and the facilitation of inference skills, practised over long chunks of texts. This reading practice enabled a ‘catch-up’ towards students’ expected reading age.

Qualitative findings and discussion

The FR+T intervention was founded on strong international evidence of best pedagogy to develop comprehension. Therefore, the qualitative data were essential to address RQ3 (Which pedagogic approaches and other contextual factors contribute to any gains in students’ reading development?) and explain why the third hypothesis was accepted and the first two were rejected. That is, why was the FR-condition alone, unpredictably, sufficient to accelerate struggling readers’ measurable reading comprehension, without additional training? Observations identified key practices associated with comprehension-development common to both groups and, importantly, also revealed that the FR condition – reading two whole books, back-to-back in a condensed time-frame – had prompted experienced FR-only teachers to emphasise aspects of their pedagogy that mirrored the more metacognitively executed practices of the FR+T teachers, although some differences remained. The common practices were: first, extended time spent on reading, reading aloud, engagement and inference-making; second, a focus on ‘enabling discourses’ known to support inference although executed with differential skills (see below); and third, vocabulary-development. Importantly, these common practices were not identified as key components of ‘current practice’ in first interviews with any FR or FR+T teacher. Instead, teachers’ current aims in reading lessons were reported as ‘developing written analysis’ - not comprehension – primarily of text extracts, using scaffolded ‘PEE’ (Point-Evidence/quotation-Explain, or equivalent acronym) paragraphs.

Extended reading aloud, engagement

Seventeen of the 20 teachers (excluding three FR teachers) attributed enhanced comprehension to expressive reading of complete texts quickly and aloud by teacher or students, either whole-class or in peer Guided-Reading (GR) or other reading-groups, which simultaneously fostered engagement, fluency and comprehension (e.g. Guthrie and Klauda 2014):

Anna (FR+T): children have a stronger understanding […] and engagement with the text […] because I’ve allowed them to read it, rather than analyse it. They have been reading it themselves, they’ve immersed themselves in the text.

Georgina (FR+T) described as ‘transformative’ and ‘radical’ this ‘immersion’ in extended chunks of text (typically 30+ minutes of one-hour lessons), increasing time spent on engaged reading, and ensuring that students were consciously monitoring their comprehension and learning strategies to resolve comprehension ‘blocks’. Teachers and peers in groups also paused to ask questions, clarifying vocabulary and text meanings, and requested recaps and cross-text connections, but without excessively disrupting the concentrated reading. This process required struggling readers to practise whole-text inference-making, recalling previous details, relating and integrating these with the new section, while applying contextual knowledge (Oakhill and Cain 2012). All teachers differentially scaffolded contextual knowledge to support knowledge-based inferences, by using maps, images and non-fiction texts, including newspaper articles and YouTube clips, for example, to illustrate refugee movements from Zimbabwe to South Africa in 2008, or details of trench warfare in World War 1 for Now is the Time for Running (Williams 2012) or Private Peaceful (Morpurgo 2003) respectively.

Liam (FR) described the ‘Faster Read’ as enabling his students ‘to become heavily invested emotionally in the book’, enhancing their persistence and drive for coherence (van den Broek 1997). This was evidenced in Liam’s and other classes by students’ sustained textual concentration, whole-class and small-group, and their comprehension, while using reading strategies, including spontaneous questions (‘Why do the South Africans hate the [Zimbabwean refugee] boys?’). Student, Tom (FR+T), echoed students across groups, saying he was ‘excited’ to read ‘because you don’t quite know what’s going to happen next and you can’t wait to find out’. Penny (FR+T) attributed improved comprehension to the reciprocal relation between print exposure, engagement and students’ developing comprehension skills (van Bergen, Vasalampi, and Torppa 2020) in Private Peaceful, a challenging, non-linear narrative:

Because [students] just don’t read […] they’re spending 3 hours a week reading […] 100% more than they would be doing otherwise! They’re definitely thinking about things at a much deeper level […] forming their own opinions and being able to follow more than one thing going on in the book […] it’s improved their attitude when they come into English because we read for pleasure and it is pleasurable [original emphasis]

‘Enabling discourses’ for reading, and developing reading strategies

All 20 teachers used teacher questioning to support vocabulary-development, sentence- and text-level inferences and comprehension. Fifteen teachers either encouraged, or modelled how students could ask clarifying, literal and inferential questions, which the FR+T training included. Typically, FR teachers encouraged questioning and reading strategies more implicitly. However, six FR teachers, spontaneously mirrored FR+T practices without additional training because they were ‘dialogic teachers’, using dialogue skilfully to deepen text comprehension (Howe et al. 2019). Therefore, Ellen (FR) displayed an unread section on the interactive-whiteboard and attached question-bubbles, modelling her authentic questions and response strategies. Students repeated this exercise independently, asking authentic questions, supporting low- and higher-level comprehension: ‘Who are “the mara-mara”?’,‘Why will Captain Washington “know what to do?”’. This generated rich discussion: students inferred contextual information about Zimbabwe and integrated cross-text details to build consistent character representations.

FR+T teachers all systematically modelled questioning and other strategies, whole-class and in peer GR or reading groups. For example, Lily’s (FR+T) students used key-rings with their chosen images to represent different strategies (inference-questions, applying world knowledge, inferring vocabulary, predicting, making whole-text connections). Within weeks, students could identify the strategies they had independently used, displaying their ‘strategies key-ring’, but, importantly, they did this while deepening their interpretation of a whole text, not as a discrete exercise in ‘strategies instruction’ (McGeown et al. 2015, 57). Predictably, the most extensive student questioning and independent use of reading strategies were observed in GR groups, supporting research (Okkinga et al. 2018).

Vocabulary development

All teachers supported vocabulary-development to different degrees, typically through class discussion, clarifying meanings, eliciting definitions of new words and consolidating these, or using drama to enact unfamiliar words to peers. All FR+T and some FR teachers were more systematic: Grace (FR+T) gave groups a four-step approach to tackling unfamiliar vocabulary, using the context of sentence/paragraph, phonology and morphology, and requiring students to use vocabulary learnt independently in a new context, supporting vocabulary ‘depth’ and ‘breadth’ (Nagy and Herman 1987). The class then collaboratively inferred the meaning of ‘pneumonoultramicroscopicsilicovolcanoconiosis’, spontaneously applauding themselves.

GR groups were observed using these strategies, supported in one FR+T class by use of a ‘vocabulary monitor’, who wrote unfamiliar words on a mini-whiteboard during the group’s reading, encouraged peers to infer meanings and took any remaining unknown words to the class for clarification.

Differences in practice: FR+T and FR

The main differences in practice between the two groups were in degree: FR+T teachers were typically more explicit, systematic and metacognitive in teaching whole-text reading to struggling readers, particularly all types of inference, self-monitoring using ‘think alouds’, and story-structure via graphic organisers (Oakhill and Cain 2012). FR+T teachers were typically also more consistent in using high-quality, ‘enabling’ discourse to develop comprehension and interpretation, although six experienced FR teachers also demonstrated this skill. While 15 teachers used some peer-reading-groups, only FR+T teachers used structured GR, with students demonstrating sustained ability to tackle comprehension challenges collaboratively.

Conclusion

This pilot study found that increasing print exposure to challenging, engaging, culturally diverse whole texts, using immersive, expressive reading aloud, comprehension-strategies and enabling discourse strategies – a Faster Read as practised by both groups – increased struggling readers’ comprehension significantly by an average of 16 months, across both conditions. Hypothesis 3 (a and b) is accepted. Hypotheses 1 and 2 are not accepted: there was no significant differential effect of training condition on ‘average+’ or ‘struggling’ readers; average+ readers’ comprehension improved across both groups, but not significantly.

It was disappointing that, unpredictably, the full version of the ‘Faster Read’ – FR+T – did not differentially impact on students’ comprehension-development. The qualitative data suggest why. The FR-only intervention altered the pedagogic practice of the FR teachers more than anticipated, directly benefitting struggling readers, confirmed by observations and interviews. In reading two texts quickly, FR teachers changed customary focus on textual micro-analysis to whole-text comprehension. They asked good inference questions, requiring students to recall and make connections across the text, recognising narrative structure and enabled students to use reading strategies, supported by small-group discussion.

There were some differences in pedagogy and knowledge between the two teacher groups. FR+T teachers consistently used – and taught – skilful discourses to support comprehension and engagement, especially guided reading; had a more knowledgeable, metacognitive approach so that students knew how to resolve comprehension-failures, choosing from a range of strategies; and had a systematic focus on inference. We hypothesise that there may have been a latent effect on students’ comprehension from FR+T teacher practices, which would have required further testing six+ months after the project’s end and needs investigation in future research.

The FR+T intervention targeted struggling readers: Average+ readers’ comprehension starting points may have required more sophisticated teacher and peer discourse strategies to enable measurable impact (Soter et al. 2008). Teachers’ discourse practices were inconsistent, with differences in effectiveness. The study’s relatively short duration may have been insufficient for teachers to enhance discourse to differentiate for already ‘good’ readers. With a multi-component intervention, it is not possible to identify whether average+ readers already used most/all reading strategies taught, or if an alternative, even more challenging text may have had greater impact on comprehension, by developing more sophisticated vocabulary, inference and interpretive skills.

Limitations of the study contextualise the findings. This was a robust, pilot pre-/post-text study, not an RCT, atypically conducted in the naturalistic context of ‘low sets’ in ‘volunteer’ schools, and therefore designed without a third ‘comparison’ group. The FR condition (increasing print exposure by reading two books, but continuing business-as-usual), unpredictably triggered considerable change in teachers’ pedagogy, which, to an extent, mirrored that of trained FR+T teachers. Additionally, schools were diverse and volunteered to participate. Although FR+T teachers achieved treatment fidelity, the study took place in classrooms with some behavioural challenges so it was difficult for some FR+T teachers to implement the required changes fully. Some FR+T teachers would have benefitted from a longer period between the training and intervention to practise required pedagogy (Okkinga et al. 2018). While working in ‘real-world’ contexts strengthens external validity, these limitations could be addressed in a future, randomised design.

The two groups were matched for students’ reading ability, but there was variation in size and characteristics of parallel classes within and across schools. Given the study’s primary focus on comprehension, the weakest readers might have benefited from more teacher-training on lower-level reading processes. A limitation of group-administered tests is that they cannot differentiate between students whose comprehension difficulties are caused by decoding weaknesses and those with good word reading, who cannot understand the texts. A larger sample, enabling progress in the component skills of comprehension to be distinguished, would enable finer distinction between reader types and any gains. There was also no opportunity to re-test students after six or more months to assess sustained reading gains. This would further identify any Hawthorne effects of a 12-week study and longer-term effects of the training. Minimisation (Torgerson and Torgerson 2007) might also have achieved better baseline balance between the groups, in terms of teacher effect, than paired randomisation.

The study’s findings are, however, important for struggling readers, doubly disadvantaged in education, being disproportionately represented by those in low SES groups. These students need support to close the reading – and attainment – gap with advantaged peers, which the pandemic has exacerbated (EPI 2022). The FR intervention increased exposure to print and engaged reading of whole narratives, giving adolescent struggling readers the practice and experience of reading that extensive research has demonstrated is necessary (Torppa et al. 2020; van Bergen, Vasalampi, and Torppa 2020), but which, unaccountably, is not policy or practice in England. Instead, this intensive reading is relegated to the context of independent ‘reading for pleasure’, which struggling readers, typically from lower SES groups, do far less than advantaged peers (NLT 2020).

Occasional large-scale studies on independent reading have shown that disadvantaged students’ comprehension can be increased by enhancing access to books and encouraging greater reading alone (e.g. Allington et al. 2010), and use of online reading programmes (‘Accelerated Reader’, Gorard, Siddiqui, and See 2015). However, in this present study, teachers and peers in groups not only increased the amount of whole-text reading, especially aloud, they also used skilful talk to explore, question, infer and create mental models (Soter et al. 2008). Thus, while reading extended text, the teacher and peers guided struggling readers with weak reader identities to develop all the essential components of successful reading: comprehension strategies (e.g. EEF 2023), reading persistence, engagement and confidence (e.g. McGeown et al. 2015). Crucially, this development occurred in a community of similarly engaged readers, which is supported by Cremin et al. (2014) and Baye et al.'s (2019, 133) rigorous review on reading interventions for adolescents, which found that struggling readers benefit more from ‘socially and cognitively engaging instruction’ in mainstream classes than additional, daily independent ‘reading periods’.

The FR approach is a radical departure from current practice in England, confirmed by all teacher participants. Indeed, although FR-only teachers were instructed to use ‘business-as- usual’, they spontaneously adapted their pedagogy to read whole texts and focus on whole-text comprehension. However, FR teachers also reported finding the FR challenging without additional training and three failed to complete the texts. This clashed with what they saw, albeit unwillingly, as their primary objective: ‘teaching written analysis’ to prepare students for GCSE examinations three years later, a ‘Naming of Parts’ (Reed 2007) approach, not supporting whole-text comprehension, interpretation or engagement.

Current pedagogy in England is the product, firstly, of teacher accountability for attainment at 16 years, prompting appropriation of GCSE English assessment criteria for 11–14-year-olds, assuming that fundamental comprehension skills and reading habits can be by-passed in pursuit of higher, analytical skills. Secondly, pedagogy has been shaped by policy since the National Strategy (DfES 2001), recommending micro-analysis of decontextualised extracts without focusing on engaged, whole-text ‘reading for meaning’ as an end in itself. Similarly, Snow and O’Connor (2016) critique US policy for prescribing independent ‘close reading’, divorced from discussion of required background knowledge and new vocabulary.

Therefore, the findings respond to the ongoing emphasis in the literature on targeting independent reading as the solution to address adolescents’ reading weaknesses (Gorard, Siddiqui, and See 2015), by reframing the problem. English teachers need access to a theorised understanding of reading development to ensure that mainstream lessons not only provide struggling readers with strategies to develop comprehension independently, but also enable them to experience pleasurable, immersive, whole-text reading and overcoming of comprehension challenges. These are necessary conditions for motivating all reading, inside classrooms – not outside. Recent pandemic learning losses experienced by disadvantaged students (an overlapping category with struggling readers) make this study’s findings particularly urgent.

Acknowledgments

Many thanks to Dr Jeff Readman for his advice on multi-level modelling and to Head teachers, teachers and students of participating schools.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by a University of Sussex RDF fund.

Notes on contributors

Julia Sutherland

Julia Sutherland is a Senior Lecturer and Co-Director of Teaching and Learning, School of Education and Social Work, University of Sussex. Her research focuses on reading and comprehension; dialogic classroom discourses; teacher education and development; educational disadvantage and social justice.

Jo Westbrook

Jo Westbrook is Professor of International Education & Pedagogy, School of Education and Social Work, University of Sussex. Her research includes reading and comprehension in multilingual contexts; multiliteracies; inclusive & alternative pedagogies internationally; and teacher education.

Jane Oakhill

Jane Oakhill is Professor of Experimental Psychology, School of Psychology, University of Sussex. Her research focuses on all aspects of reading comprehension and recently on relations between children’s amount of reading and aspects of empathy.

Sue Sullivan

Sue Sullivan is a Senior Lecturer, School of Psychology, University of Sussex. Her research investigates reading comprehension, including children with hearing and visual impairments.

Notes

1. The Hodder test has a reported reliability coefficient (α value) of 0.95 and 0.94 for Forms [tests] A and B (McCarty and Crumpler 2014, 25).

2. Supplementary t-test result indicated that the reading age differentials between the two reader types was significant for both t = 0 and t = 1.