answer

Grammar for writing? An investigation of the effects of contextualised grammar teaching on students’ writing

Susan Jones • Debra Myhill • Trevor Bailey

Published online: 14 September 2012

� Springer Science+Business Media B.V. 2012

Abstract The role of grammar instruction in the teaching of writing is contested in

most Anglophone countries, with several robust meta-analyses finding no evidence

of any beneficial effect. However, existing research is limited in that it only con-

siders isolated grammar instruction and offers no theorisation of an instructional

relationship between grammar and writing. This study, drawing on a theorised

understanding of grammar as a meaning-making resource for writing development,

set out to investigate the impact of contextualised grammar instruction on students’

writing performance. The study adopted a mixed-methods approach, with a ran-

domised controlled trial and a complementary qualitative study. The statistical

analyses indicate a positive effect on writing performance for the intervention group

(e = 0.21; p \ 0.001); but the study also indicates that the intervention impact

differentially on different sub-groups, benefiting able writers more than weaker

writers. The study is significant in being the first to supply rigorous, theorised

evidence for the potential benefits of teaching grammar to support development in

writing.

Keywords Grammar � Linguistics � Writing � Subject knowledge

Introduction

The instructional benefit of teaching of grammar in first language English curricula

is contested in both research and professional literature in Anglophone countries

(Gordon, 2005; Wyse, 2004) and in the Netherlands (van Gelderen & Oostdam,

S. Jones � D. Myhill (&) � T. Bailey

Graduate School of Education, University of Exeter, Heavitree Road, Exeter EX1 2LU, UK

e-mail: [email protected]

D. Myhill

University of Wollongong, Wollongong, NSW, Australia

123

Read Writ (2013) 26:1241–1263

DOI 10.1007/s11145-012-9416-1

2005; van Gelderen, 2006). Indeed, in the 1960s and 1970s, following the

Dartmouth Conference (Dixon, 1975), most Anglophone countries (for example,

England, Australia, New Zealand and the USA) abandoned grammar instruction on

the grounds that it was ineffectual in supporting language development, particularly

writing development (Locke, 2009). More recently, driven principally by policy

imperatives, grammar has been re-introduced into the English curriculum in

England, and a parallel process is currently occurring in Australia. However, there is

no clarity or agreement about the role of grammar in the English curriculum and it

remains a strongly contested issue (Myhill, 2011; Myhill & Jones, 2011; Myhill

et al., 2011). The uncertain role of grammar in the language curriculum is set within

an international context, in Anglophone countries particularly, expressing concerns

about the writing attainment of children (NCW, 2003; OFSTED, 2009; Salahu-Din,

Persky, & Miller, 2008). In England, for example, in 2011, 32 % of boys and 19 %

of girls entering secondary education had not achieved the baseline standard in

writing expected for their age group, compared with 20 % of boys and 13 % of girls

who had not achieved the baseline in reading (DfE, 2011). However, there have

been no systematic studies of whether making meaningful connections between

particular linguistic structures and particular writing tasks supports the development

of students’ writing. This paper reports on the outcomes of a randomised controlled

trial (RCT), funded by the Economic and Social Research Council, which

investigated the impact of contextualised grammar teaching on students’ writing

performance.

The effectiveness of grammar teaching

Empirical studies investigating the efficacy of grammar teaching provide little

evidence of any beneficial impact upon students’ competence in writing. Robust

meta-analyses by Braddock, Lloyd-Jones, and Schoer (1963), Hillocks (1986) and

most recently, by the Evidence for Policy and Practice Information and Co-

ordinating Centre (EPPI) (Andrews et al., 2006; EPPI, 2004) have concluded that

there is no evidence that teaching grammar is of benefit in supporting writing

development. Indeed, Hillocks and Smith (1991, 602) argue that ‘‘research over a

period of nearly 90 years has consistently shown that the teaching of school

grammar has little or no effect on students’’.

There are, however, several major difficulties with almost all of the research that

these reviews represent. The first is that studies repeatedly investigate whether

various forms of grammar teaching, such as learning transformational grammar,

grammar exercises and drills, or parsing sentences, improve writing. The emphasis

is on teaching grammar in the hope that it might have an impact on writing

outcomes. In many of the studies (for example, Bateman & Zidonis, 1966; Elley,

Barham, Lamb, & Wylie, 1975, 1979; Robinson, 1959) isolated grammar lessons

are taught as part of a curriculum programme in grammar, and the writing measures

used to draw empirical conclusions are produced in a different teaching context.

Robinson (1959) tested grammatical knowledge and compared this with the quality

of their composition—she correlated a grammar test with impression marking, and

looked only at word classes. Bateman and Zidonis (1966) taught a transformational

1242 S. Jones et al.

123

grammar course, with the purpose of ‘‘determining the effects of a study of

transformational-generative grammar on the language growth of secondary school

pupils’’ (Elley et al., 1979, 98). The Elley et al. study (1975, 1979) had three

treatment groups: the first was a course typical of English classes at that time in

New Zealand using a textbook addressing grammar, comprehension, and writing;

the second was a reading-writing course where students spent 40 % of their time

free reading, 40 % sharing a class reader, and 20 % writing; and the final treatment

group was a transformational grammar course with the intention of helping students

‘‘see how they can discover facts about their language and how they use it’’ (1975,

28). Students in this group were taught about such things as sentence combining,

subordination, participial modifiers, and deep and surface structures. A second

difficulty with the few existing studies is that many are small -scale. The Bateman

and Zidonis study, for example, had a sample of 41 students.

A further difficulty is that none of the studies theorise an instructional

relationship between grammar and writing, which might inform the design of an

appropriate pedagogical approach. The studies are all located in very different

educational jurisdictions, with differing pre-existing curricular emphases on

grammar. In New Zealand, for example (Elley et al., 1975, 1979) there was

growing unease about the efficacy of traditional grammar teaching amongst

educational professionals, but also a back-to-basics call at policy level which

appeared to advocate ‘‘strong doses of English grammar as a cure for some of our

educational ills’’ (Elley et al., 1975, 3). But none directly address the inter-

relationship of grammar and writing, or offer a theoretical account of such an inter-

relationship.

Contextualised grammar teaching

Thus, there are, to date, no large-scale studies which investigate the benefits or

otherwise of teaching grammar in the context of writing lessons, in which

connections are forged for the student writer between the grammar under focus and

the learning focus for the writing. However, Hudson (2001) draws attention to a

Finnish doctoral study (Laurinen, 1955) which reports improved punctuation scores

for primary students who have been taught clause structures. Hudson argues that the

benefits accrued are because the particular area of grammar taught correlates with

the learning focus for writing, punctuation. Effective punctuation is underpinned by

grammatical understanding and the teaching helped the students to make

connections between the two.

This synergistic relationship between writing and grammatical understanding is

also evident in Fogel and Ehri’s (2000) study. This is unusual in taking as its starting

point an identified writing problem, the tendency of some ethnic minority children

to use non-standard Black English vernacular (BEV) in their writing. The study set

out to ‘‘examine how to structure dialect instruction so that it is effective in teaching

SE forms to students who use BEV in their writing’’ (Fogel & Ehri, 2000, 215) and

found a significant improvement in avoidance of BEV in the group who were given

both strategies and guided support. They argue that their results demonstrated that

the approach used had ‘‘clarified for students the link between features in their own

Grammar for writing? 1243

123

nonstandard writing and features in SE’’ (2000, 231). The Fogel and Ehri study

moves the field forward by beginning to look at the pedagogical conditions which

support or hinder the transfer of grammatical knowledge into written outputs.

Significantly, too, their study begins with a specific linguistic learning need around

which teaching is designed. Fearn and Farnan (2007) have also investigated

teaching grammar in the context of teaching writing, seeking to examine if there is

‘‘a way to teach grammatical structures that will satisfy high-stakes tests and

teachers’ needs, and at the same time, positively affect writing performance?’’

(Fearn & Farnan, 2007, 2). Their experimental study encouraged a problem-solving

approach, using oral language, and appears to be focused on the use of particular

linguistic structures or word class: there is no evident attempt to talk about the

construction of meaning or effect through form. Nonetheless, their study did find

positive impact of their intervention and their conclusion is that it is beneficial for

learners when ‘‘grammar and writing share one instructional context’’ (Fearn &

Farnan, 2007, 16). A recent meta-analysis by Graham and Perin (2007) looking at

effective strategies to teach writing did find that teaching sentence-combining,

helping students to construct more structurally complex sentences, had a positive

effect. In general, however, there is a dearth of studies which address contextualised

grammar teaching probably because ‘‘grammar has traditionally been taught and

learned in an environment that is devoid of context’’ (Mulder, 2010, 73) or not

taught at all.

Theorising grammar-writing connections

As noted above, a limitation in much of the existing research on grammar teaching

is that there is no clear conceptualisation of a theoretical rationale for why grammar

might support writing development (indeed much of the research is framed by

polemic and ideology). Educational linguists (Carter, 1990; Denham & Lobeck,

2005; Hancock, 2009) contend that a better understanding of how language works in

a variety of contexts supports learning in literacy. They draw particularly on the

principles of contemporary linguistic theories which are descriptive and socio-

cultural in emphasis, or as Carter describes them, ‘‘functionally oriented, related to

the study of texts and responsive to social purposes’’ (Carter, 1990, 104). This is in

contrast to the more prescriptive approach to grammar which traditional grammars

espoused (Hudson, 2004). In the US, there has been some emphasis on the notion of

grammar in context (Weaver, 1996, for example), but a theoretical relationship

between grammar and writing has never been adequately articulated, and the idea of

‘in context’ is problematic, often meaning in practice an isolated ‘mini-grammar

lesson’ within an English lesson (for a critique of this, see Myhill, 2010a).

The difference between prescriptive and descriptive views of grammar is central to

a consideration of a theoretical rationale for attention to grammar in the teaching of

writing. Prescriptive grammar sets out how language should be used, the rules of

language use; whilst descriptive grammar looks at language in use. Denham and

Lobeck (2010, 3) contrast linguists who ‘‘have sought to build a grammar that would

be adequate for describing the language’ with English teachers who ‘have sought to

apply a grammar that is already constructed’’. Public and political views of grammar

1244 S. Jones et al.

123

tend strongly towards the prescriptive view, maintaining that the role of the teacher is

to address grammatical accuracy in writing and eradicate error (Myhill, 2011; Myhill

& Jones, 2011; Myhill et al., 2011). Hancock, reflecting on the US educational

context, observed that ‘‘grammar is error and error is grammar in the public mind’’

(Hancock, 2009, 175). In England, the same tendency at public and policy level is

evident—grammar is frequently presented as a remediation tool, a language

corrective. The Queen’s English Society, whose remit is the preservation of the

English language, maintain that grammar is important for the ‘‘diagnosing of faults or

problems in one’s own writing and in that of others’’ (QES, 2011). Traditional school

grammar is prescriptive and is critiqued by Hudson for having ‘‘no roots in modern

linguistics’’ and for being ‘‘fragmentary, dogmatic and prescriptive’’ (Hudson, 2004,

116). Descriptive theories of grammar counterpoint the normative emphasis on

correctness, characteristic of prescriptive grammar, with a more socially-oriented

analysis of how language is used, including in different social, linguistic and cultural

contexts. A prescriptivist theory of a grammar-writing relationship would argue for

the importance of grammar in securing correctness in written expression; a

descriptivist theory of a grammar writing relationship would argue for the importance

of grammar in illuminating how written text generates meaning in different contexts.

The theoretical approach adopted in this study builds on descriptivist views of

grammar. Understanding and analysing how language works in different purposes

and contexts makes connections for learners between language as an object of study

and language in use, as realised in the act of writing. This is, in effect, a theory of

grammar centred upon rhetorical understanding. As a theoretical perspective, this

has at its heart the discussion and analysis of how meaning is crafted and created

through shaping language to achieve the writer’s rhetorical intentions (Kolln, 2002;

Locke, 2005; Micciche, 2004; Paraskevas, 2006). It aims to foster explicit

understanding and ‘‘conscious control and conscious choice over language which

enables both to see through language in a systematic way and to use language more

discriminatingly’’ (Carter, 1990, 119).

A theorised view of grammar teaching in the context of writing which builds on the

understandings outlined above, and which focuses on the teaching of writing rather

than the teaching of grammar, incorporates the following principles (Myhill, 2010a).

Firstly, writing is a communicative act supporting writers in understanding the social

purposes and audiences of texts and how language creates meanings and effects;

secondly, grammar is a meaning-making resource: supporting writers in making

appropriate linguistic choices which help them to shape and craft text to satisfy their

rhetorical intentions; and finally, connectivity, supporting writers in making

connections between their various language experiences as readers, writers and

speakers, and in making connections between what they write and how they write it.

Teachers’ grammatical subject knowledge

The absence of explicit grammar teaching in the English curriculum in Anglophone

countries for nearly 50 years has resulted in many present English teachers not

having the grammatical subject knowledge (GSK) needed to teach grammar

confidently. A survey of teachers in England in 1998 (QCA, 1998) revealed

Grammar for writing? 1245

123

considerable lack of confidence, in particular with clause structures and syntax. The

report noted a ‘‘significant gap … in teachers’ knowledge and confidence in

sentence grammar and this has implications for … the teaching of language and

style in texts and pupils’ own writing’’ (QCA, 1998, 35). In the US context, Vavra

(1996) observed the gap between modern linguistics and the prescriptive, rule-

bound grammar taught by most English teachers. Cameron (1997) argued that the

literature degree qualifications of most English teachers not only leaves them ill-

equipped to cope with grammar teaching, but also generates anxiety, hostility, and

lack of confidence towards grammar.

This lack of confidence plays out in English classrooms through inaccurate

teaching of grammar points (Myhill, 2000, 2003) and insecurity in dealing with

students’ questions (Burgess, Turvey, & Quarshie, 2000). In two studies investi-

gating pre-service teachers’ engagement with grammar, Cajkler and Hislam (2002,

2004) demonstrate how they struggle with GSK and how to use it appropriately in

the classroom. Hudson (2004) argued that without adequate grammatical knowledge

teachers cannot make the analysis of texts explicit, nor can they structure the

teaching context effectively.

Research questions

The over-arching research question that this study set out to investigate was: what

impact does contextualised grammar teaching have upon students’ writing and

students’ metalinguistic understanding? The qualitative study, not reported here,

provides evidence concerning students’ metalinguistic understanding, as well as

complementary evidence about the implementation of the intervention. The RCT

provides evidence concerning the impact of the intervention. Consequently, the

following hypotheses were formulated: (a) that contextualised grammar teaching

will be positively related to students’ writing performance, (b) that the quality of

teachers’ GSK will mediate the impact of contextualised grammar instruction.

Methodology

Participants

The participants were teachers (n = 32) and students (n = 855) of English in 32

different mixed comprehensive schools in the South-West and the Midlands regions

of England. In each school, a class of Year 8 students, aged 12–13, formed the

sample for the RCT. In order to avoid selection bias (Cook & Wei, 2002), the school

sample was secured by using local authority data to compile a numbered list of all

mixed comprehensive schools in the South-West and the Midlands. A random

number generator was used to determine a rank order, and each school was

approached in rank order until the desired sample of 32 was reached.

Baseline data about participants were collected at school, teacher and student

level. At school level, data were compiled on national examination performance,

school inspection outcomes, ethnic diversity, Special educational needs (SEN) and

1246 S. Jones et al.

123

number of students entitled to free school meals (FSM) (as a proxy for socio-

economic status). At teacher level, data were collected on years of teaching

experience, degree subject studied and gender. In addition, teacher participants

undertook a test of GSK at the start of the study and the scores were used as further

baseline data. Student data comprised gender, whether they were students with

English as an additional language (EAL) or entitled to FSM. Attainment data in

English were also collated, drawing on standard national test results in English at

age eleven (including both a writing raw score and writing level) and school

predictions of English results in national tests at age fourteen.

Design

The study sought to investigate whether the use of teaching materials which

embedded grammar teaching within teaching units for writing improved students’

performance in writing. Additionally, the study sought to examine whether teachers’

GSK was a factor upon the efficacy or otherwise of the intervention.

The sample was first stratified at teacher level according to their GSK scores to

ensure that the two groups were matched, given that GSK is known to be a factor in

the teaching of grammar. The classes were then randomly assigned, using a random

number generator, to either a comparison or an intervention group. Because it is not

possible in a naturalistic educational setting to prevent any teaching of grammar

naturally occurring, we have consistently used the term ‘comparison’ rather than

‘control’ group. The study was blind as participant teachers were not told the

research focus was grammar; instead they were told the focus was on the teaching of

writing (see also ‘‘Ethical considerations’’ section). Full details of the intervention

and comparison group teaching are provided further below.

Because causal relationships are rarely deterministic, ‘‘to different degrees, all

causal relationships are context dependent’’ (Shadish, Cook & Campbell, 2002, 5), a

complementary qualitative data set was collated alongside the experimental study to

provide in-depth understanding of the theoretical, pedagogical and contextual

implications of the statistical data. This mixed method approach is important for

RCTs conducted in educational contexts. Indeed, Moore, Graham, and Diamond

(2003) argue that ‘‘to undertake a trial of an educational or social intervention without

an embedded qualitative process evaluation would be to treat the intervention as a

black box, with no information on how it worked, how it could be improved, or what

the crucial components of the intervention were’’. Likewise, Shadish et al. (2002, 71)

recommend ‘‘the addition of qualitative methodologies to experiments’’ to provide

better interpretation and avoid errors in applying research outcomes to practice. In this

study, the data comprised lesson observations, teacher and student interviews, and

writing samples. This paper reports principally from the statistical analysis of the RCT

(for an overview report of the full study, see Myhill et al., 2012).

Procedure

Three members of the research team (all former secondary English teachers) devised

three teaching units on writing, each addressing a different genre: fictional narrative,

Grammar for writing? 1247

123

argument, and poetry. These teaching units were in harmony with the requirements of

the National Curriculum for English (DCSF, 2007), a statutory instrument, and

addressed teaching objectives for Writing as set out in the Framework for Teaching

English (DfES, 2001), the recommended policy document guiding English teaching.

Three learning objectives which specifically addressed linguistic knowledge were

common to all three teaching units, giving the opportunity to explore, for example,

how sentence variety may fulfil different purposes in the three genres; the remaining

objectives were chosen for their relevance to the genre under study. Each teaching unit

was designed to take approximately 3 weeks of timetabled English lessons, and the

study period spanned a school year. In this respect, the intervention was wholly

aligned with curriculum and teaching norms and designed for implementation in a

naturalistic setting. The teaching units adopted many of the pedagogic practices

common in secondary English classrooms, such as the use of text models, group work

and discussion, opportunities for planning and drafting, peer assessment. To that

extent, they reflected typical practice in the teaching of writing. However, in addition,

within each unit, grammar teaching was embedded, making connections between a

linguistic feature and its effect in writing. Some of these were genre-specific. For

example, the fictional narrative unit looked at how first or third person are used to

create different voice or viewpoint. Others were more generically related to

improving writing: for example, varying sentence lengths to create textual rhythm

(For more detailed explanation of the teaching activities, see Myhill, 2010b, 2011;

Myhill & Jones, 2011; Myhill et al., 2011 in professional journals). There was no

focus on grammatical error or accuracy: rather the focus was to help writers to

recognise how making grammatical choices could shape their texts for communi-

cative purposes. Table 1 below provides an overview of the learning objectives and

written outcomes for each scheme and provides examples of the embedded grammar

focus.

Both the intervention and comparison group were taught the three types of

writing over the same period, addressing the same curriculum teaching objectives,

and producing the same written outcomes (in other words columns 1, 2, and 4 of

Table 1 were the same for both groups). Both groups were given the same set of

stimulus materials and resources, but only the intervention group had detailed

teaching units, planned at lesson level, in which grammar was explicitly taught.

Thus it is reasonable, as far as is possible within a naturalistic context, to conclude

that any differences in writing performance are attributable to the intervention.

Testing materials

The impact of the intervention on students’ writing performance was determined by

a pre and post test sample of writing. Both the pre and post test writing sample were

first person narratives, drawing on personal experience, and written under controlled

conditions (see ‘‘Appendix 1’’). In order to avoid any possible bias created by the

precise choice of writing task, the topic was selected to avoid known gender

preferences in writing and to avoid the need of any specific topic knowledge or

experience. To minimise any possible test effect, a cross-over design was used: half

the sample took Writing Task 1 as the pretest, and the other half took Writing Test 2;

1248 S. Jones et al.

123

this was reversed for the post test. Both tests were independently marked by

Cambridge Assessment, an independent organisation who was responsible for

setting and marking the national test of writing for 14 year olds until 2006. The

mark schemes, one for each task, were developed based on the generic framework

underpinning the mark schemes used for all national curriculum writing tests. The

model was based specifically on the mark scheme used for the longer writing task in

the former Key Stage 3 English test. This criterion-based mark scheme has three

strands, with a maximum score of 30 marks: sentence structure and punctuation

(max. 8 marks), text structure and organisation (max. 8 marks), and composition and

Table 1 An overview of the three teaching units

Genre Learning objectives Grammar focus

examples (intervention

only)

Written outcomes

Fictional

narrative

Varying sentences and

punctuation for clarity and

effect

Developing varied linguistic and

literary techniques

Improving vocabulary for

precision and effect

Developing viewpoint, voice and

ideas

Using grammar accurately and

appropriately

Using sentence

fragments or short

sentences for

emphasis

Use of first/third

person to establish

viewpoint

Expanding noun

phrases to create

character description

A story plan, plus writing the

opening, the climax, or the

resolution of the story

Argument Varying sentences and

punctuation for clarity and

effect

Developing varied linguistic and

literary devices

Improving vocabulary for

precision and effect

Developing viewpoint, voice and

ideas

Structuring, organising and

presenting texts in a variety of

forms on paper and on screen

Using grammar accurately and

appropriately

Using modal verbs to

create different

degrees of assertion

Using co-ordination to

create counter

arguments

Using subordination to

articulate

propositions

A written persuasive speech

Poetry Varying sentences and

punctuation for clarity and

effect

Developing varied linguistics

and literary techniques

Improving vocabulary for

precision and impact

Generating ideas, planning and

drafting

Using expanded noun

phrases to create a

picture poem

Comparing line length

in poetry with

sentence variety in

prose

Using sentence

patterning for effect

A portfolio of three poems,

plus reflective annotation

Grammar for writing? 1249

123

effect (max. 14 marks). Each strand is linked to two of the assessment focuses for

writing and the criteria are derived from national curriculum levels. For each set of

scripts, Cambridge Assessment provided a first marker’s set of marks, a second

marker’s set and a ‘resolution mark’, adjudicated by a third senior marker if the first

two marks were very different. Cambridge Assessment devised training materials

for marking; delivered a training day for each marking round; and ensured the usual

standardisation checks during the marking. The markers did not know from which

treatment group the writing was derived, and the research team was not involved in

the process of marking and assessment at all. A full report was provided by

Cambridge Assessment on the marking process, including a brief commentary on

the writing itself. Markers reported some evidence of carelessness in the post-test

writing, which led to speculation that some students might have been less motivated

than in the first round, perhaps because the task was completed at the end of the

academic year.

Attrition and fidelity

The principle of intervention fidelity is highly problematic in naturalistic educational

settings as it is very difficult to control the key variable of the teacher. There is always

a methodological trade-off between internal validity and ecological validity, and the

study privileged ecological validity. It is important to understand the ways in which

the intervention was mediated by the teachers. Thus, although each teacher in the

intervention had the same training and the same set of materials, it was neither possible

nor ethical to attempt to achieve identical implementation. Teachers were allowed to

adapt materials to suit the needs of their students, but they were asked to maintain the

specified learning focus and the intervention group were asked to remain as close as

possible to the teaching methods in the teaching units. The qualitative study,

particularly the lesson observations and teacher interviews, indicates that all bar one

teacher used the teaching units throughout the research period, and there was a high

level of commitment to the project. In all the lessons observed, the teaching units were

being used. In the teacher interviews just over half of the teachers specifically claimed

they had maintained high fidelity to the teaching schemes, in part because of their

commitment to the research project, or as one teacher put it, ‘‘I’ve tried to be faithful to

your project’’. They also appeared content to maintain high fidelity to the intervention

because of their approval of the quality of the schemes—‘‘I think the scheme of work

is really, really good’’; and the sense that they were being successful with the

students—‘‘I think the lessons are going really well and I think they’re really

learning’’. However, there were differences in the way they were implemented. Where

adaptations were made, a significant number were pragmatic adaptations in timing,

where tasks or activities had to be truncated because of time limitations. However, of

particular relevance to the research focus of the study, is the fact that some of the

adaptations were altering or omitting the focus on grammar at the heart of the research

because it was felt to be too difficult. One teacher reflected that she ‘‘found it really

difficult and I adapted it to something else’’, whilst another ‘‘just totally left out the

clause part of that exercise, the final element because actually, in the scheme so far we

haven’t taught, I haven’t taught them anything about clauses, so it would have just

1250 S. Jones et al.

123

completely have gone over their head’’. Another felt that her students ‘‘would really

struggle with ellipsis. I deliberately didn’t use the term, as with a lot of terms in this

scheme, saying a term that they’ve not come across before would scare them’’. These

adaptations to the grammar element of the lessons link to the statistical finding,

reported later in this paper, that teachers’ GSK was an important factor in mediating

the intervention.

However, one teacher in the comparison group showed such low levels of fidelity

that it was decided to exclude this class from the data analysis (for example, she

taught lessons which did not address writing, and had other teachers teaching her

lessons). Thus the final sample for data analysis was 31 teachers and their classes.

The initial student sample was 855: over the year of the study, due to students

leaving schools or being absent for the post-test, and due to the loss of one whole

class, as outlined above, the final student sample was 744.

Analyses

As is common in educational studies, the data arising from this RCT are conceptually

hierarchical, comprising individual students grouped within schools/teachers/classes.

However, in this case, multilevel modelling is not an appropriate or feasible statistical

approach because the intervention or non-intervention was restricted to only a single

teacher and their class in each school and applied to the whole of that class. Accordingly,

the data were analysed using a multiple regression framework at the single level of the

n = 744 individual students, with intervention, school and teacher covariate values

attributed to these individuals according to school or class membership.

The response or ‘outcome’ variable considered in the multiple regression

modelling was the difference between the post- and the pre-test scores (each marked

as a percentage of 30) for each participant (n = 744). The explanatory covariates

comprised the indicator of intervention or non-intervention and then the full set of

school, teacher and student baseline measures discussed in ‘‘Participants’’ section.

Interaction terms between these covariates were also considered. It should be noted

that since these covariates involve a mixture of continuous measures (e.g., teacher

GSK score) and categorical factors (e.g., EAL), one could more specifically refer to

the multiple regression modelling as a multi-way analysis of covariance.

A range of stepwise procedures was then used to determine which variables and

interactions were statistically significant in impacting the response from amongst

the intervention indicator and the other covariates. Final model selection was then

checked using full residual analyses and standard diagnostic procedures.

The final results from the model selection indicated that the most appropriate

reduced model involved the significant explanatory variables/factors or interactions

listed in Table 2 when judged at the 5 % significance level (p \ 0.05). Results relating

to this final selected model are reported and discussed in detail in ‘‘Results’’ section.

Ethical considerations

Blind randomisation, where participants do not know the purpose of the research,

requires careful ethical consideration, as it violates the principle of voluntary

Grammar for writing? 1251

123

informed consent. The British Education Research Association ethical guidelines

(BERA, 2004) recommend that all researchers should ‘‘avoid deception or

subterfuge unless their research design specifically requires it’’. In this case, and

particularly because of the contested nature of grammar teaching, knowledge of the

grammar focus could have been a significant bias. The BERA guidelines argue that

‘‘decisions to use deception or subterfuge in research must be the subject of full

deliberation and subsequent disclosure in reporting’’ (BERA, 2004, 6). Prior to

conducting the study, institutional ethical approval was sought and secured. At the

outset of the study, when participants were invited to give informed consent, they

were told that the precise research focus was being withheld from them, and assured

that they would be informed at the end of the data collection phase, at which point

they could withdraw their consent. At an end of project conference, teachers were

given the full details of the purpose of the study, alongside the preliminary findings:

in addition all teachers in the comparison group were given the detailed teaching

units used in the intervention. When data analysis was complete, all participating

teachers were given a summary of the findings of the study and recommendations

for classroom practice.

Results

As discussed in ‘‘Analyses’’ section, the outcome variable considered was the

difference between the post and the pre-test percentage marks for each participant.

The mean value of this variable over all 744 participants was 9.24 %, indicating an

overall improvement between pre- and post- writing tests in both the comparison

and intervention groups. In the comparison group (n = 332) the mean outcome was

6.41 %, lower than in the intervention group (n = 412) where mean outcome was

11.52 %. So overall, and ignoring the effect of any other covariates, a simple two-

sample t test suggests a highly significant (p \ 0.001) positive effect size of 5.11

percentage marks for the intervention in terms of improvement in writing

attainment. In other words, the embedded teaching of grammar relevant to the

writing being studied had an overall beneficial effect on students’ achievement in

writing.

Table 2 Showing most significant covariates at individual, teacher and school level

Student

level

Intervention or comparison

group

Factor: ‘comparison’, or ‘intervention’

Writing attainment level at age

11

Factor: ‘below average’, or ‘above average’

Interaction between writing

attainment and intervention

Factor: ‘comparison and below average’, ‘comparison and

above average’, ‘intervention and below average’ or

‘intervention and above average’

Teacher

level

Years of teaching experience Factor: ‘\5’, ‘5–10’, or ‘[10 years’

Linguistic subject knowledge Variable: pre study GSK test score

School

level

Latest inspection result Factor: ‘poor’, ‘satisfactory’, ‘good or excellent’

Special educational needs Variable: % of SEN pupils in school

1252 S. Jones et al.

123

However, the simple overall comparison above masks the more complex picture

which emerges when significant school, teacher and student explanatory covariates

are allowed for via the multiple regressions modelling approach discussed in

‘‘Analyses’’ section. Table 2 presents coefficient estimates (standard errors, t- and p-

values) relating to the student, teacher and school level variable/factors or

interactions for the final model selected from the stepwise regression modelling

as described in ‘‘Analyses’’ section. All the non-intercept terms in this model are

very highly statistically significant (at a 0.1 % level) with the exception of the

teacher GSK score which nevertheless is still significant at the 5 % level. Factor

levels, interactions or variables other than those reported in Table 3 were

insignificant at the 5 % level. Overall, the model explains only 9 % of the raw

variation in the pre-and post test differences (R2 = 0.09, adjusted R2 = 0.08), but

low overall explanatory value is not atypical in educational studies and taken overall

this model is very highly significant (F7,36 statistic = 10.42, p value = 0.00001).

The interpretation of the ‘intercept’ in this model is that it is the estimated post- minus

pre-test difference for a ‘baseline’ student who has below average writing ability

(whether in the comparison group or the control group), who has a teacher with a low

GSK score and with teaching experience of either\5 years or of more than 10 years,

and who is in a school which received a ‘poor’ result at last inspection and which has a

low percentage of students requiring special education needs. Coefficient values other

than the intercept in Table 2 then reflect the estimated adjustments to this baseline which

apply for students of other types and which are statistically significant different from

zero. These effects are discussed in more detail in the following subsections.

Analysis of the raw scores provided by the markers, broken down into the sub-

components of the mark schemes, indicates that the greatest improvement was made

in composition and effect, though similar improvements were made at sentence and

text level. This suggests that the impact of the grammar teaching was not simply at

the syntactical level of the sentence but had an effect on overall effectiveness,

consistent with the pedagogical goals of the intervention (Table 4).

Table 3 Showing coefficient estimates (standard errors, t- and p-values) at student, teacher and school

level for final selected model

Model coefficient Estimate SE t value p value

Intercept -6.1727 (3.5285) (-1.739) (\0.100)

Student level

Above average writing level -7.4006 (1.9037) (-3.887) (\0.001)

Intervention group and above average writing level 8.1246 (1.3863) (5.860) (\0.001)

Teacher level

Teacher linguistic subject knowledge (GSK) score 0.5817 (0.2760) (2.108) (\0.050)

Teacher has 5–10 years experience 5.9832 (1.3091) (4.570) (\0.001)

School level

‘Satisfactory’ result in last school inspection 7.0165 (1.9258) (3.643) (\0.001)

‘Good or excellent’ in last school inspection 8.3394 (2.0151) (4.138) (\0.001)

% special educational needs (SEN) in School 0.6411 (0.1610) (3.983) (\0.001)

Grammar for writing? 1253

123

Student level effects

The data analysis indicates that the intervention benefitted some writers more than

others, and that prior writing attainment was a significant factor in this. For the

analysis, students were divided into two attainment groups, ‘more able writers’ and

‘less able writers’ based on prior writing attainment. This was done using the raw

score results for Writing in the Key Stage 2 tests, the most fine-grained data we

collected was available for all children and which had been externally assessed. The

analysis reported in Table 3 shows a significant positive effect on writing

performance for students in the intervention group whose baseline data indicated

that they were above average in writing attainment. The estimated effect for this

group is strong in an educational intervention context (Cohen, 1969, 23) and provides

robust evidence for the beneficial impact on this group of students of receiving the

contextualised grammar teaching. However, the analysis also indicates that it is only

those students who were above average in writing at baseline who benefitted strongly

from the intervention. The impact of the intervention was neutral on learners whose

baseline writing attainment was below average. Table 3 also indicates that more able

writers in the comparison group made significantly less improvement in their writing

scores over the year than less able writers, suggesting that the teaching of writing

may be insufficiently challenging to secure improvement in writers who are already

fairly successful. Overall then, more able writers in the intervention group improved

significantly more than less able writers who received no significant benefit from the

intervention, and more able writers in the comparison group made significantly less

improvement over the year than less able writers in either the comparison or

intervention groups. This strongly suggests that the intervention may have been

particularly well-matched to the learning needs of more able writers.

Teacher level effects

The length of the teacher’s experience and the quality of the teacher’s subject

knowledge of grammar were both significant in influencing student writing

Table 4 Showing improvement by sub-component in the mark schemes

Total score Sentences structure

and punctuation

Text structure

and organisation

Composition

and effect

Comparison

Raw scores 6,565 1,799 1,748 3,018

Improvement 806 228 241 337

% improvement 12.3 12.7 13.8 11.2

Intervention

Raw scores 6,446 1,764 1,704 2,978

Improvement 1,256 319 334 604

% improvement 19.5 18.1 19.6 20.3

1254 S. Jones et al.

123

outcomes, regardless of whether students were in the comparison or the intervention

group. The data suggest that where the intervention was beneficial to students it was

most effective with teachers who had between 5 and 10 years teaching experience

(ie neither inexperienced nor highly experienced). This may seem paradoxical but it

may be that inexperienced teachers lacked the confidence to use the intervention

teaching materials appropriately, whilst highly experienced teachers had more

deeply embedded pedagogical practices which they found harder to alter, despite the

intervention. Certainly, the qualitative evidence from the classroom observations

and interviews with teachers suggests that experienced teachers, in particular, made

more adaptations to the teaching materials than others, often substituting in their

own preferred ways of working. Conversely, the least experienced teachers often

followed the teaching materials more rigidly and inflexibly.

Similarly, there is a relationship between GSK and the impact of the intervention.

Students in classes with teachers with lower GSK made less improvement than

those with teachers with higher GSK. This is a more predictable finding, as the

teaching materials required confident mastery of grammar, and although they aimed

to provide good support for teachers, it is possible that the grammar knowledge

required by the intervention teaching materials was at too high a level for those with

lower GSK to teach effectively. Indeed, the lesson observations repeatedly show

teachers struggling to cope with student questions on grammar and sometimes

communicating incorrect information to students.

School level effects

At school level, the baseline data collected aimed to capture both social diversity

(ethnicity; number of students entitled to FSM) and school performance (national

examination results; school inspection outcomes and the percentage of students with

SEN). The analysis indicates that the school inspection outcomes and the percentage

of students with SEN were significant factors. The school inspection system in

England grades schools through a rigorous process of analysing school performance

data on a range of factors, and through observing lessons and grading the teaching

quality, and assessing the quality of school leadership. The data indicate that

students in schools graded Good or Excellent through school inspection benefitted

more from the intervention than those in schools graded Satisfactory. It is important

to note that high inspection grades do not correlate with ‘high ability’ students: all

the schools in the study were comprehensive schools, which do not select students

on ability, and the inspection assessments take into account the baseline

performance of students on entry to the school. So a high-performing school could

be achieving good inspection results with lower ability students. It is not surprising

that the intervention was more successful in high-performing schools as teaching

quality is generally lower in lower-performing schools, which is likely to influence

how the intervention materials were used. It is less easy to account for the influence

of SEN, particularly as the data indicate that it was schools with higher percentages

of students with SEN which related positively to the beneficial impact of the

intervention.

Grammar for writing? 1255

123

Discussion

The study set out to investigate whether contextualised grammar teaching impacts

beneficially upon students’ writing, and addressed two hypotheses: (a) that

contextualised grammar teaching will be positively related to students’ writing

performance, (b) that the quality of teachers’ GSK will mediate the impact of

contextualised grammar instruction. The analysis indicates that both hypotheses are

confirmed. However, the use of multiple regression modelling reveals that the

intervention was differentially experienced by learners, with more able writers

benefiting strongly whilst weaker writers benefited less.

The lack of impact of the intervention upon less able writers may be because the

grammatical features addressed in the teaching materials were poorly matched to

their learning needs as writers. One drawback of the RCT was the requirement for

standard teaching materials to be used across all groups, making it difficult to tailor

the teaching materials to particular classes needs or interest. The design of the

intervention materials in this study appear to have been better matched to the

learning needs of able writers. It is also possible that more able writers have clearer

communicative and rhetorical intentions for their writing than less able writers,

enabling them to make more appropriate use of their grammatical understanding to

shape text appropriately. Further research could usefully begin by analysing writers’

developmental needs, and then designing the teaching schemes around those

grammar features most relevant to these needs. It might also be the case that the

level of conceptual thinking required to understand grammatical concepts and

transfer that learning into their writing was too high a cognitive challenge or created

too high a cognitive load for this group. Cognitive load theory (Sweller, 1988)

acknowledges the constraints that can be placed upon working memory capacity

when tackling cognitively complex tasks, such as writing, and argues that teaching

materials need to be designed to manage cognitive load appropriately. Similarly,

Stark, Kopp, and Fischer (2011) argue that extraneous load is ‘‘attributed to sub-

optimal instructional design and interferes with productive learning’’. Certainly, the

classroom observations suggest that for some students the use of metalinguistic

terminology may have been a barrier, rather than a support. This study offers a

fruitful impetus for further research which explores the issue of transfer of

metalinguistic understanding into writing produced, and also whether appropriate

metalinguistic awareness could be generated without the use of metalanguage (for

example, by the use of patterns and examples to explain).

The multiple regressions modelling also allow us to use the data to construct

scenarios which predict outcome effects. A fuller set of these is provided in

‘‘Appendix 2’’; however, Table 5 illustrates some of the key predictive patterns

based on the significant covariates. The intervention was of little benefit to less able

writers so they are excluded from this scenario-building. For more able writers the

possible outcomes are more influenced by the covariates. For example, a student

with above average writing ability in a low-performing school with a teacher with

low subject knowledge and more than 10 or \5 years’ experience is less likely to

benefit from the intervention, with writing outcomes improving from -0.6 to 2.1

1256 S. Jones et al.

123

(or taking the model uncertainty into account, from somewhere in the range -10.7

to -1.3 to somewhere in the range -1.7 to 5.9).

As anticipated in the research design, the results and the predictive scenarios

indicate the complexity of interventions in educational settings, which defy

simplistic causal explanations. An inherent danger in the use of an RCT is its

tendency to be regarded as the ‘gold standard’ of research (Slavin, 2002; Torgerson,

et al., 2004), drawing on a medical model of research, and the widespread view that

it can answer the question of ‘what works’ in education. But, as Biesta points out, ‘‘a

student is not an illness just as teaching is not a cure’’ (2007, 8) and the closed,

deterministic system of the RCT can fail to acknowledge that cause and effect in

education are not linear and continuous but open, fluid and dynamic (Morrison,

2001). An RCT can provide important evidence and new understandings about a

given research ‘problem’ but it cannot provide evidence which translates directly

into classroom practice or policy. Caution is needed in implying ‘‘a causal model of

professional action’’ (Biesta, 2007, 7) which suggests that mandated action will be

implemented in practice and will result in similar outcomes for all. Understanding

‘‘how teachers take on research, adapt it, and make it their own’’ (Black & Wiliam,

2003, 633), is both more important and more challenging than simplistic notions of

‘what works’.

In the light of this, the results of this study are important not only in providing

robust empirical evidence of the potential benefit of teaching grammar in the

context of writing, but also in highlighting the complex interaction effects between

student, teacher, school and intervention. This is important in informing appropriate

recommendations for policy or professional practice. All interventions in educa-

tional settings are always mediated by teacher and school effects, and one-size-fits-

all policy imperatives are rarely an appropriate solution to the challenges of raising

student attainment. Whilst this study demonstrates the benefit of contextualised

grammar teaching on more able writers, it also signals that policy and practice need

Table 5 Showing predictive scenarios drawing on the multiple regression modelling

Scenario description Estimated post

minus pre test difference

95 % Confidence

interval

Comparison, high WL, low GSK,\5 or [10 years expr,

poor LIR, avg SEN

-6.0 (-10.7, -1.3)

Intervention, high WL, low GSK,\5 or [10 years expr,

poor LIR, avg SEN

2.1 (-1.7, 5.9)

Comparison, high WL, low GSK, 5–10 years expr,

poor LIR, avg SEN

-0.1 (-5.0, 4.8)

Intervention, high WL, low GSK, 5–10 years expr,

poor LIR, avg SEN

8.1 (4.0, 12.2)

Comparison, high WL, high GSK, 5–10 years expr,

good or excel LIR, avg SEN

11.6 (8.4, 14.8)

Intervention, high WL, high GSK, 5–10 years expr,

good or excel LIR, avg SEN

19.7 (15.9, 23.5)

WL writing Level, GSK linguistic subject knowledge, LIR latest inspection result, SEN special educational

needs

Grammar for writing? 1257

123

to consider the level of teachers’ GSK, their experience, different learners’ needs,

and the school context in which they operate. Moreover, it opens the way for further

experimental and qualitative research which will investigate the pedagogical

conditions which support or hinder the transfer of grammatical knowledge into

written outputs.

Conclusions

Theoretically, this study is important in reframing research in the relationship

between grammar teaching and writing. Contrary to previous studies (Andrews

et al., 2006; Hillocks, 1986, this study offers robust evidence of a positive

relationship between grammar and writing. Unlike previous studies, the intervention

was predicated upon a theorised pedagogical model for the role of grammar in the

teaching of writing which conceptualises grammar as a meaning-making resource

and which emphasises connectivity in instructional designs (Myhill, 2010a). The

study demonstrates that more able students’ learning about writing is enhanced by

explicit understanding of how grammar choices can be used to shape written text to

satisfy writers’ rhetorical goals, but importantly it also indicated no parallel benefit

for less able writers. The results of the intervention are important in indicating

future directions for research at the grammar-writing interface.

Acknowledgments We are grateful to the Economic and Social Research Council (UK) for the funding

which enabled this research to be undertaken.

Appendix 1

Writing task 1: challenges

Challenges come in many shapes and sizes, from adrenaline-pumping adventures

like rock climbing or bungee jumping to more everyday events like starting a new

school, overcoming a fear, getting into a team or learning a new skill. Often a

challenge is small to someone else but very big to you—for example, having to say

sorry for something you have done or meeting someone you don’t know for the first

time.

Write an account of a challenge you have faced in your life so far for a school

magazine feature on ‘challenging situations’

You could:

• Describe the challenge.

• Tell the story of how you tackled it and what happened in the end.

• Reflect on how you feel about this now.

• Choose language that will make your account vivid and interesting for someone

else to read.

1258 S. Jones et al.

123

You might want to start like this:

At the time, it seemed like the biggest challenge of my life …

Writing task 2: childhood fears

All of us get frightened about things from time to time—especially when we are

young children. Our fears may be based on something real that has happened or

something we only imagine will happen. For example, it could have been a fear of

something imaginary, such as serpents under the bed, or a very real fear, such as a

fear of flies.

Write about your childhood fears, real or imagined, for a school magazine

feature on ‘things that frighten us when we are small’

You could:

• Choose a time when you were frightened or nervous about something.

• Describe it in detail.

• Explain why you were afraid and what happened in the end.

• Reflect on how you feel about this now.

• Choose language that will make your account vivid and interesting for someone

else to read.

You might want to start like this:

I don’t think I have ever been as scared as when …

Appendix 2

See Table 6.

Table 6 Scenarios showing effects of different variables as the success of the intervention

Scenario description Estimated post minus

pre test difference

95 % Confidence

interval

Comp, high WL, low GSK, 5–10 years expr,

poor LIR, avg SEN

-6.0 (-10.7, -1.3)

Comp, high WL, mid GSK, 5–10 years expr,

poor LIR, avg SEN

-3.7 (-7.9, 0.5)

Comp, high WL, high GSK, 5–10 years expr,

poor LIR, avg SEN

-1.4 (-6.1, 3.3)

Comp, high WL, low GSK, 5–10 years expr,

poor LIR, avg SEN

-0.1 (-5.0, 4.8)

Comp, high WL, low GSK, \5 (–2.1, 4.1) or

[10 years expr, good or excel LIR, avg SEN

1.0 (-2.1, 4.1)

Intervent or comp, low WL, low GSK,

5–10 years expr, poor LIR, avg SEN

1.4 (-3.3, 6.1)

Grammar for writing? 1259

123

Table 6 continued

Scenario description Estimated post minus

pre test difference

95 % Confidence

interval

Intervent, high WL, low GSK, 5–10 years expr,

poor LIR, avg SEN

2.1 (-1.7, 5.9)

Comp, high WL, mid GSK, 5–10 years expr,

poor LIR, avg SEN

2.3 (-2.0, 6.6)

Comp, high WL, mid GSK, 5–10 years expr,

good or excel LIR, avg SEN

3.3 (1.3, 5.3)

Intervent or comp, low WL, mid GSK,

5–10 years expr, poor LIR, avg SEN

3.7 (-0.7, 8.1)

Intervent, high WL, mid GSK, 5–10 years expr,

poor LIR, avg SEN

4.4 (1.0, 7.8)

Comp, high WL, high GSK, 5–10 years expr,

poor LIR, avg SEN

4.6 (-0.2, 9.4)

Comp, high WL, high GSK, 5–10 years expr,

good or excel LIR, avg SEN

5.6 (2.8, 8.4)

Intervent or comp, low WL, high GSK,

5–10 years expr, poor LIR, avg SEN

6.0 (1.0, 11.0)

Intervent, high WL, high GSK, 5–10 years expr,

poor LIR, avg SEN

6.7 (2.4, 11.0)

Comp, high WL, low GSK, 5–10 years expr,

good or excel LIR, avg SEN

7.0 (3.6, 10.4)

Intervent or comp, low WL, low GSK,

5–10 years expr, poor LIR, avg SEN

7.3 (2.1, 12.5)

Intervent, high WL, low GSK, 5–10 years expr,

poor LIR, avg SEN

8.1 (4.0, 12.2)

Intervent or comp, low WL, low GSK,

5–10 years expr, good or excel LIR, avg SEN

8.4 (4.6, 12.2)

Intervent, high WL, low GSK, 5–10 years expr,

good or excel LIR, avg SEN

9.1 (6.1, 12.1)

Comp, high WL, mid GSK, 5–10 years expr,

good or excel LIR, avg SEN

9.3 (6.8, 11.8)

Intervent or comp, low WL, mid GSK,

5–10 years expr, poor LIR, avg SEN

9.7 (4.8, 14.6)

Intervent, high WL, mid GSK, 5–10 years expr,

poor LIR, avg SEN

10.4 (6.7, 14.1)

Intervent or comp, low WL, mid GSK,

5–10 years expr, good or excel LIR, avg SEN

10.7 (7.5, 13.9)

Intervent, high WL, mid GSK, 5–10 years expr,

good or excel LIR, avg SEN

11.4 (9.0, 13.8)

Comp, high WL, high GSK, 5–10 years expr,

good or excel LIR, avg SEN

11.6 (8.4, 14.8)

Intervent or comp, low WL, high GSK,

5–10 years expr, poor LIR, avg SEN

12.0 (6.5, 17.5)

Intervent, high WL, high GSK, 5–10 years expr,

poor LIR, avg SEN

12.7 (8.2, 17.2)

Intervent or comp, low WL, high GSK,

5–10 years expr, good or excel LIR, avg SEN

13.0 (9.0, 17.0)

1260 S. Jones et al.

123

References

Andrews, R., Torgerson, C., Beverton, S., Freeman, A., Locke, T., et al. (2006). The effect of grammar

teaching on writing development. British Educational Research Journal, 32(1), 39–55.

Bateman, D. R., & Zidonis, F. J. (1966). The effect of a study of transformational grammar on the writing

of ninth and tenth graders. Champagne, IL: National Council of Teachers of English.

Biesta, G. J. J. (2007). Why ‘what works’ won’t work. Evidence-based practice and the democratic deficit

of educational research. Educational Theory, 57(1), 1–22.

Black, P., & Wiliam, D. (2003). ‘In praise of educational research’: Formative assessment. British

Educational Research Journal, 29(5), 623–637.

Braddock, R., Lloyd-Jones, R., & Schoer, L. (1963). Research in written composition. Urbana: National

Council of Teachers of English.

British Educational Research Association (BERA). (2004). Revised ethical guidelines for educational

research: 2004. Nottingham, UK: BERA.

Burgess, T., Turvey, A., & Quarshie, R. (2000). Teaching grammar: Working with student teachers.

Changing English, 7(1), 7–21.

Cajkler, W., & Hislam, J. (2002). Trainee teachers’ grammatical knowledge: The tension between public

expectations and individual competence. Language Awareness, 11(3), 161–177.

Cajkler, W., & Hislam, J. (2004). Teaching and understanding grammar: How trainee teachers make

sense of a new curriculum? In N. Bartels (Ed.), Applied linguistics in language teacher education

(pp. 295–312). Dordrecht: Kluwer.

Cameron, D. (1997). Sparing the rod: What teachers need to know about grammar? Changing English,

4(2), 229–239.

Carter, R. (Ed.). (1990). Knowledge about language. London: Hodder and Stoughton.

Cohen, J. (1969). Statistical power analysis for the behavioral sciences. New York: Academic Press.

Cook, R., & Wei, W. (2002). Selection effects in randomized trials with count data. Statistics in Medicine,

21(4), 515–531.

Denham, K., & Lobeck, A. (2005). Language in the schools: Integrating linguistics into K12 teaching.

Hillsdale, NJ: Lawrence Erlbaum Associates.

Table 6 continued

Scenario description Estimated post minus

pre test difference

95 % Confidence

interval

Intervent, high WL, high GSK, 5–10 years expr,

good or excel LIR, avg SEN

13.8 (10.4, 17.2)

Intervent or comp, low WL, low GSK,

5–10 years expr, good or excel LIR, avg SEN

14.3 (9.8, 18.8)

Intervent, high WL, low GSK, 5–10 years expr,

good or excel LIR, avg SEN

15.1 (11.7, 18.5)

Intervent or comp, low WL, mid GSK,

5–10 years expr, good or excel LIR, avg SEN

16.7 (12.7, 20.7)

Intervent, high WL, mid GSK, 5–10 years expr,

good or excel LIR, avg SEN

17.4 (14.5, 20.3)

Intervent or comp, low WL, high GSK, 5–10

years expr, good or excel LIR, avg SEN

19.0 (14.3, 23.7)

Intervent, high WL, high GSK, 5–10 years expr,

good or excel LIR, avg SEN

19.7 (15.9, 23.5)

WL writing level, GSK linguistic subject knowledge, LIR latest inspection result, SEN special educational

needs

Grammar for writing? 1261

123

Denham, K., & Lobeck, A. (Eds.). (2010). Linguistics at school: Language awareness in primary and

secondary education. Cambridge: Cambridge University Press.

Department for Children, Schools, and Families (DCSF). (2007). English in the national curriculum.

http://curriculum.qcda.gov.uk/key-stages-3-and-4/subjects/key-stage-4/english/index.aspx. Acces-

sed February 10, 2010.

Department for Education (DfE). (2011). Interim results for key stage 2 and 3 national curriculum

assessments in England, 2010/2011. London: DfE. http://www.education.gov.uk/researchand

statistics/statistics/allstatistics/a00196847/interimresults. Accessed August 20, 2012.

Department for Education and Skills (DfES). (2001). Framework for teaching English: Years 7, 8 and 9.

London: DfES.

Dixon, J. (1975). Growth through English. Sheffield: NATE.

Elley, W. B., Barham, I. H., Lamb, H., & Wylie, M. (1975). The role of grammar in a secondary school

curriculum. New Zealand Council for Educational Studies, 10, 26–41.

Elley, W. B., Barham, I. H., Lamb, H., & Wylie, M. (1979). The role of grammar in a secondary school

curriculum: Educational research series no. 60. Wellington: New Zealand Council for Educational

Research.

EPPI Review Group for English. (2004). The effect of grammar teaching (syntax) in English on 5 to

16 year olds’ accuracy and quality in written composition. London: EPPI-Centre.

Fearn, L., & Farnan, N. (2007). When is a verb using functional grammar to teach writing? Journal of

Basic Writing, 26(1), 1–26.

Fogel, H., & Ehri, L. C. (2000). Teaching elementary students who speak Black English vernacular to

write in standard English: Effects of dialect transformation practice. Contemporary Educational

Psychology, 25, 212–235.

Gordon, E. (2005). Grammar in New Zealand schools: Two case studies. English Teaching: Practice and

Critique, 4(3), 48–68.

Graham, S., & Perin, D. (2007). Writing next: Effective strategies to improve writing of adolescents in

middle and high schools—A report to Carnegie Corporation of New York. Washington, DC:

Alliance for Excellent Education.

Hancock, C. (2009). How linguistics can inform the teaching of writing? In R. Beard, D. Myhill, J. Riley,

& M. Nystrand (Eds.), The sage handbook of writing development (pp. 194–208). London: Sage.

Hillocks, G. (1986). Research on written composition. Urbana, IL: ERIC Clearing House on Reading and

Communication Skills.

Hillocks, G., & Smith, M. (1991). Grammar and usage. In J. Flood, J. M. Jensen, D. Lapp, & J. R. Squire

(Eds.), Handbook of research on teaching the English language arts (pp. 591–603). New York:

MacMillan.

Hudson, R. (2001). Grammar teaching and writing skills: The research evidence. Syntax in the Schools,

17, 1–6.

Hudson, R. (2004). Why education needs linguistics? Journal of Linguistics, 40(1), 105–130.

Kolln, M. (2002). Rhetorical grammar: Grammatical choices, rhetorical effects. New York: Longman.

Laurinen, I. (1955). The development of sentence sense in the light of the results attained in the teaching

of writing in Finnish primary schools. Unpublished PhD thesis.

Locke, T. (2005). Writing positions and rhetorical spaces. In B. Doecke & G. Parr (Eds.),

Writing = learning (pp. 75–95). South Australia: AATE, Wakefield Press.

Locke, T. (2009). Grammar and writing—The international debate. In R. Beard, D. Myhill, J. Riley, & M.

Nystrand (Eds.), The sage handbook of writing development (pp. 182–193). London: Sage.

Micciche, L. (2004). Making a case for rhetorical grammar. College Composition and Communication,

55(4), 716–737.

Moore, L., Graham, A., & Diamond, I. (2003). On the feasibility of conducting randomized comparison

trials in education. British Education Research Journal, 29(5), 673–689.

Morrison, K. (2001). Randomised controlled trials for evidence-based education: Some problems in

judging ‘what works’. Evaluation and Research in Education, 15(2), 69–83.

Mulder, J. (2010). Envisioning linguistics in secondary education: An Australian exemplar. In A. Denham

& K. Lobeck (Eds.), Linguistics at school: Language awareness in primary and secondary

education. Cambridge: Cambridge University Press.

Myhill, D. A. (2000). Misconceptions and difficulties in the acquisition of meta-linguistic knowledge.

Language and Education, 14(3), 151–163.

Myhill, D. A. (2003). Principled understanding? Teaching the active and passive voice. Language and

Education, 17(5), 355–370.

1262 S. Jones et al.

123

Myhill, D. A. (2010a). Ways of knowing: grammar as a tool for developing writing. In T. Locke (Ed.),

Beyond the grammar wars: A resource for teachers and students on developing language knowledge

in the English/literacy classroom (pp. 129–148). London: Routledge.

Myhill, D. A. (2010b). Harnessing grammar: weaving words and shaping texts. Better: Evidence-Based

Education, Winter 2010, pp. 12–13.

Myhill, D. A. (2011). Living language, live debates: Grammar and standard English. In J. Davison,

C. Daly, & J. Moss (Eds.), Debates in English Teaching (pp. 63–77). London: Routledge.

Myhill, D. A., & Jones, S. M. (2011). Policing grammar: The place of grammar in literacy policy. In

A. Goodwyn, & C. Fuller (Eds.), The Literacy Game (pp. 45–62). London: Routledge.

Myhill, D. A., Lines, H. E., & Watson, A. (2011). Making meaning with grammar: A repertoire of

possibilities. mETAphor, 2, 1–10.

Myhill, D. A., Jones, S. M., Lines, H. E., & Watson, A. (2012). Re-Thinking grammar: The impact of

embedded grammar teaching on students’ writing and students’ metalinguistic understanding.

Research Papers in Education, 27(2), 139–166.

National Commission on Writing in America’s Schools and Colleges (NCW). (2003). The neglected ‘R’:

The need for a writing revolution. New York: College Board.

OFSTED. (2009). English at the crossroads. London: OFSTED.

Paraskevas, C. (2006). Grammar apprenticeship. English Journal, 95(5), 65–69.

Qualifications and Curriculum Authority (QCA). (1998). The grammar papers. London: QCA.

Queen’s English Society (QES). (2011). www.queens-english-society.com/about.aspx. Accessed January

20, 2011.

Robinson, N. (1959). The relation between knowledge of English grammar and ability in English

compositions. Unpublished MEd thesis, University of Manchester.

Salahu-Din, D., Persky, H., & Miller, J. (2008). The nation’s report card: Writing 2007. Washington, DC:

Institute of Education Sciences, US Department of Education.

Shadish, W. R., Cook, T., & Campbell, D. (2002). Experimental and quasi-experimental designs for

generalized causal inference. Boston: Houghton Mifflin.

Slavin, R. E. (2002). Evidence-based education policies: Transforming educational practice and research.

Educational Researcher, 31(7), 15–21.

Stark, R., Kopp, V., & Fischer, M. R. (2011). Case-based learning with worked examples in complex

domains: Two experimental studies in undergraduate medical education. Learning and Instruction,

21(1), 22–33.

Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12,

257–285.

Torgerson, C., Brooks, G., Porthouse, J., Burton, M., Robinson, A., Wright, K., et al. (2004). Adult

literacy and numeracy interventions and outcomes: A review of controlled trials. London: National

Research and Development Centre for Adult Literacy and Numeracy.

Van Gelderen, A. (2006). What we know without knowing it: Sense and nonsense in respect of linguistic

reflection for students in elementary and secondary education. English Teaching: Practice and

Critique, 5(1), 44–54.

Van Gelderen, A., & Oostdam, R. (2005). Effects of fluency training on the application of linguistic

operations in writing. L1 Studies in Language and Literature, 5(2), 215–240.

Vavra, E. (1996). On not teaching grammar. English Journal, 85(7), 32–37.

Weaver, C. (1996). Teaching grammar in the context of writing. English Journal, 85(7), 15–24.

Wyse, D. (2004). Grammar for writing? A critical review of empirical evidence. British Journal of

Educational Studies, 49(4), 411–427.

Grammar for writing? 1263

123