Академический Документы
Профессиональный Документы
Культура Документы
In 2000–2002 an innovative early years curriculum, the Enriched Curriculum (EC), was introduced
into 120 volunteer schools across Northern Ireland, replacing a traditional curriculum similar to
others across the UK at that time. It was intended by the designers to be developmentally appropri-
ate and play-based with the primary goal of preventing the experience of persistent early failure in
children. The EC was not intended to be a literacy and numeracy intervention, yet it did consider-
ably alter pedagogy in these domains, particularly the age at which formal reading and mathematics
instruction began. As part of a multi-method evaluation running from 2000–2008, the research
team followed the primary school careers of the first two successive cohorts of EC children, compar-
ing them with year-ahead controls attending the same 24 schools. Compared to the year-ahead con-
trol group, the findings show that the EC children’s reading and mathematics scores fell behind in
the first two years but the majority of EC children caught up by the end of their fourth year. There-
after, the performance of the first EC cohort fell away slightly, while that of the second continued to
match that of controls. Overall, the play-based curriculum had no statistically significant positive
effects on reading and mathematics in the medium term. At best, the EC children’s scores matched
those of controls.
Introduction
Despite a vast literature on the benefits of play-based, developmentally sensitive ped-
agogy in the early years and its spreading influence on early years education in the
UK, the US and elsewhere, previous research has failed to investigate the effect of this
pedagogy in primary school, as distinct from pre-school, on medium-term reading
and mathematics outcomes. The opportunity to conduct such research arose when
the Enriched Curriculum (EC) was introduced into Northern Ireland in September
2000. This pilot project pre-dated the introduction of the statutory Foundation Stage
into the Northern Ireland curriculum in 2007. The research team was not involved in
the initial design or implementation of the EC. It was commissioned as the evaluation
team and provided formative feedback through annual reports from 2001 onwards
(e.g., Sproule et al., 2001, 2002, 2003) and through substantial summative reports at
the end of different phases in 2005 (Sproule et al., 2005) and 2009 (McGuinness
et al., 2009a, b; Sproule et al., 2009; Trew et al., 2009).
*Corresponding author. School of Psychology, Queen’s University Belfast, 18–30 Malone Road,
Belfast, BT9 5BN, UK. Email: L.Sproule@qub.ac.uk
and/or small sample sizes. In addition, all but one of these studies used DAP ratings
to assign classrooms to DAP or non-DAP groups rather than comparing explicitly
DAP curricula with more traditional programmes. Furthermore, studies have shown
that teachers’ stated beliefs in DAP are not necessarily reflected in their practice (Bur-
ts et al., 1992; Buchanan et al., 1998; McGovern, 2003), which remains open to pres-
sure from outside stakeholders to follow a formal approach. Finally, most studies
looked at short-term outcomes. Across the studies covered in the above review, the
findings on cognitive and academic outcomes were mixed and inconsistent. Van
Horn and Ramey (2003), tracking a large national sample over the first three years of
school in the US, concluded that ‘the effects of DAP as observed in classrooms
accounts for little or no variation in children’s academic performance’ (p. 961). How-
ever, the question of the longer term effects of a DAP curriculum on traditional
school-related outcomes remains open. Addressing these issues, the present study has
a large sample size, compares a DAP curriculum with a traditional curriculum using a
quasi-experimental method, and is the first to look at medium-term scholastic out-
comes.
selected a further 18 schools to join the evaluation, such that the sample of 24 schools
(original six schools plus additional 18) came to better reflect the population of
schools in Northern Ireland.
The focus of this paper comes from the first strand of the evaluation, directed to
reading and mathematics progress only. Although it is important to judge the EC
against its wider aims, the role of literacy and numeracy as significant gateways to
learning-across-the-curriculum will always draw attention to these outcomes. With its
longitudinal design and inclusion of more than one cohort of EC children, this paper
allows rich exploration of the far-reaching effects of the transition to a play-based
early years curriculum on progress in literacy and numeracy and uncovers important
issues pertinent to national play-based curricula.
The main research question was:
• Compared to the year-ahead controls, what are the effects of the EC on the chil-
dren’s mathematics and reading achievement (test scores) as they progress through
the primary school years?
More specifically:
• What are the effects of the EC in the short-term—during the first two years while
the children are experiencing the curriculum, and during the next two years while
they transition into more formal teaching?
• What are the effects of the EC in the medium-term as they children progress
through their upper primary school years?
• Are there any differences between the progress of the first EC cohort compared to
that of the second EC cohort?
Method
Design
The study employed a quasi-experimental design investigating outcomes compared
with matched controls over time. Appropriate multilevel statistical techniques esti-
mated the incremental differences attributable to the EC over time. Two successive
cohorts of children who experienced the EC curriculum in the first and second years
of implementation in each school were studied, allowing the stability or change in EC
practices and outcomes over time to be investigated. The study was not initially
planned as a full longitudinal study over eight years. However, as the early and
interim findings were reported to the stakeholders and the funders, additional groups
of schools joined the study at different times leading to a complex sample structure
(see below).
Instrument
Performance Indicators in Primary Schools (PIPS), run by the CEM centre at
Durham University, was chosen as a psychometrically robust measure with a substantial
history in the study of attainment (e.g., Tymms et al., 2004, 2009). The instrument
had not previously been in widespread use in Northern Ireland. PIPS was designed to
match the English National curriculum, which was very similar to the traditional pre-
existing curriculum in Northern Ireland followed by the controls. The PIPS baseline
and end of Year 1 tests yield picture vocabulary, reading and mathematics scores and
are highly predictive of later outcomes (Tymms, 1999). Thereafter, the yearly tests
include a non-verbal section. The yearly picture vocabulary and non-verbal scores
together were used to generate a ‘developed ability’ score, used here as a proxy for IQ to
demonstrate variation in the sample. Using data from several thousand English schools,
PIPS is standardised on a mean of 50 and an SD of 10. All tests were administered by
research staff.
Sample
Schools. Twenty-four schools participated, categorised into three groups, depending
on when they joined the evaluation. These groups were: the original Belfast high
deprivation group (Group 1, six schools); the first group of non-deprived schools out-
side Belfast (Group 2, six schools); and the second non-deprived group outside Bel-
fast (Group 3, 12 schools), as detailed in Table 1 (~36 classes in each year group).
The latter two groups of schools were divided between the Western (n = 2), North
Eastern (n = 7), Southern (n = 5) and South Eastern (n = 4) Education and Library
Boards. This breakdown is a reasonable reflection of the relative population distribu-
tions of the Education and Library Boards. Schools were not randomly selected to
participate in the evaluation; they volunteered. Nevertheless, Table 1 indicates that
the sample was reasonably representative of the population of schools in Northern
Ireland, in terms of type of location (urban/town/rural), social advantage/disadvan-
tage as indicated by percentage of free school meals (FSM) and by the average ability
Range of % of
Proxy IQ FSM in school Years schools Type of locality
Mean (SD)* group started EC of schools
Group: Cohort
Joined the evaluation in 2000 (6 Belfast high deprivation schools):
Year-ahead cohort (Control) 45.5 (6.4) 50–76 All in 2000 All in 2000
EC first cohort (EC1) 44.8 (5.2)
EC second cohort (EC2) 45.6 (6.4)
Joined the evaluation in 2001:
Year-ahead cohort (Control) 52.6 (7.4) 2–27 All in 2001 1 inner city
EC first cohort (EC1) 51.8 (7.1) 3 suburban
EC second cohort (EC2) 51.1 (6.6) 1 large town
1 rural or small town
Joined the evaluation in 2004 (12 schools outside Belfast):
Year-ahead cohort (Control) 50.5 (7.0) 0–30 1 suburban
EC first cohort (EC1) 51.5 (7.0) 2001 (n = 2) 4 large town
EC second cohort (EC2) 50.7 (6.8) 2002 (n = 10) 7 rural or large town
Comparison across cohorts:
Control 49.7 (7.5) 0–76 All of the above
EC1 49.9 (7.2) 0–76 All of the above
EC2 49.3 (7.2) 0–76 All of the above
of its pupils as measured by PIPS. In addition, a balance was sought between schools
with the three management types in Northern Ireland; integrated3 (n = 4), state con-
trolled (n = 13) and state maintained4 (n = 7) primary schools. This is a rough
approximation of the relative population of the three management types. As seen in
Table 1, demographic statistics clearly distinguished the high deprivation group in
Belfast from the rest in terms of FSM and intake ability characteristics. Appendix B
contains a flow chart that relates this complex sample to the 10 waves of data collec-
tion.
Participants: sample composition and controls. There were 2095 children (1025 boys)
who contributed to the evaluation at some point. For various reasons―cost, the roll-
ing nature of the evaluation, and absenteeism―not all children were tested on every
occasion and schools joined the evaluation at different times (see Appendix B). For
the Group 3 schools, outcomes data were not available for the early years as the chil-
dren had already progressed further by the time these schools joined the evaluation.
In addition, there were relatively fewer Year 7 children compared to Year 6 because
EC2 children in Group 3 schools had not reached Year 7 when the evaluation ended
(see Appendix A for explanation of year group notation). The average number of
observations per child was 3.2. However, missing data almost always conformed to
missing- at-random criteria. This determined the choice of model in the analysis (see
Appendix C).
There were three successive cohorts of children in each school; controls (one year
ahead of the first intervention cohort, the first intervention cohort itself (denoted
EC1) and the succeeding cohort (denoted EC2). Actual sample sizes in each year
group are given in Table 2. Use of year-ahead control groups minimises the influence
of potential confounding effects, such as differences in school intake, differences in
teachers, and so on.
Analysis
The regression model was required to estimate the effect of the intervention sepa-
rately on reading and mathematics, while controlling for other effects such as gender,
age in class, and social background (free school meals) that are known to impact on
scholastic outcomes. It also allowed us to take into account the effects of clustering in
the data—the fact that observations are from the same individuals over time clustered
in schools and that schools are grouped into areas (inner city vs. the rest). The analy-
sis employed a repeated measures random effects difference-in-difference model, esti-
mated by the GLIMMIX command in SAS 9.2 using quasi-maximum likelihood
criteria. Full details of the model are described in Appendix C, specifically how it
deals with the absence of certain baseline data and makes maximum use of the data
that is available.
Variables used in the analysis. Gender and month-of-birth were treated as main
effects by the model.5 FSM was only available at the school level as a covariate,
as several schools did not release FSM data at the individual child level. Conse-
quently, the percentage of free school meals (FSM) at the school level was used
to characterise differences in intake. A dummy variable (inner city) denoted those
schools in the high deprivation group, in order to estimate whether there was any
extra disadvantage in belonging to this group after allowing for social deprivation
as a linear function.
Results
Two models were estimated, one for mathematics and one for reading. The full set of
parameter estimates are shown in Appendix D along with their standard errors, effect
sizes6 and hypothesis tests.
Main effects
Main effects were estimated for pupil gender, birth month (deviation away from Janu-
ary birth), percentage school-level FSM entitlement and school inner-city location
and were considered time invariant, that is, they were estimated to have an impact
that persists over time without interacting with the curriculum choice. The impact of
such variables was thus limited to vertically shifting the expected transition of scores
over time for a pupil either up or down by a fixed amount depending on the sign and
magnitude of the coefficient.
In general the estimated main effects were of the a priori hypothesised sign and in
accordance with the previous literature. However school-level variables were not
found to be statistically significant in either reading or mathematics models.
Pupil-level main effects. The gender variable was found to be negative and statistically
significant in the reading model indicating that, all other things being equal, girls out-
performed boys at each time point. The effect size (ES) was 0.23. For mathematics
the picture was reversed with boys outperforming girls with an ES of 0.03, but this
was not statistically significant.
The month of birth variable was statistically significant in both models, with coeffi-
cients being 0.32 in reading per month and –0.36 in mathematics. The estimated
expected effect sizes at each year between a pupil born at one end of the age spectrum
against another born at the other (i.e., a difference of 12 months) was thus 0.38 in
reading and 0.48 in mathematics.
School-level main effects. School level main effects were variables measured at the
school level which were applied to individual pupils in the pupil level regression
model, i.e., for any given school, each pupil in that school will have the same value of
percentage of pupils eligible for FSM in the regression model. Such variables were
thus intended to capture the systematic impact of school factors on the pupil beyond
the individual circumstances (gender, etc.) of that pupil. Therefore the estimated
parameter on the FSM variable did not capture the impact of the FSM status of the
pupil; rather it was the impact of a school having higher levels of FSM after allowing
the individual pupil characteristics.
In terms of the difference of the impact of school-level FSM between the school
with the highest level centred FSM (51%) and the lowest (–25%) on the individual
pupil, the impact was similar in magnitude to that of the maximum impact of birth
month, with effect sizes of 0.43 and 0.55 points difference between the lowest and
highest FSM schools in mathematics and reading respectively. However the differ-
ences were not statistically significant. Similarly, the inner city school dummy had a
reasonably large effect (approximately 0.4 ES in both reading and mathematics) but
was not statistically significant.
Random effects
The random effects contained in the model were intended to capture not only the
random variation in results that occurs at each time point but also the time-invariant
school and pupil effects which were not captured in the school or pupil variables used
in the regression model, i.e., each school and pupil will have an associated time-
invariant constant term, which similar to the main effects was assumed to have a
systematic impact on the pupil attainment over time. Appendix D shows the model
estimates as covariances whereas we discuss them as standard deviations in this
section. All random components were statistically significant and have broadly the
same values across both reading and mathematics models, thus the following discus-
sion applies to both models.
The school level random effect showed a modest variation between schools that
is independent of the observable characteristics, with on average an approximately
ES of 0.25 difference between the pupils of any pair of schools even after allowing
for observed pupil and school characteristics. As expected, the variation at pupil
level was higher; for any randomly selected pair of pupils, even after allowing for
differences in scores due to their observable characteristics (gender and birth
month) and differences between schools (FSM proportion and school random
effect), then on average there is a difference of approximately 0.7 ES between
pupils in PIPS scores over time. Such a figure can be compared to the coefficients
related to observable characteristics and we can conclude that the influence of fac-
tors related to the unobserved characteristics of pupils outweigh those systematic
effects related to gender or month of birth.
54
Controls
EC1
52 EC2
50
PIPS mathematics
48
46
44
42
Y0 Y1 Y2 Y3 Y4 Y5 Y6 Y7
Year group
56
Controls
EC1
54 EC2
52
PIPS reading
50
48
46
44
Y0 Y1 Y2 Y3 Y4 Y5 Y6 Y7
Year group
3, both EC1 and EC2 cohorts began to catch up with controls. In Year 4, both
cohorts achieved parity with controls. Thereafter, the performance of the two EC
cohorts diverged. In Years 5–7, EC1 children again showed significant deficits
compared with controls of ESs in the range 0.18 (5.6 months) to 0.24
(7.5 months), shadowing but exceeding the decline in the performance of controls
at the same stage. For the EC2 cohort, there was a significant deficit in Year 5 of
ES 0.17 (5.3 months) compared with controls, followed by a positive shift in Years
6 and 7 to mirror the pattern shown by controls.
Discussion
Preliminary points
Before proceeding with the main discussion about the impact of the EC on the out-
comes, some preliminary methodological points will be acknowledged and some
comments on the findings from the main effects analyses will be made. This was not a
randomised control trial and schools were self-selecting, so we must be cautious
about generalising the findings. Also, because of the different times at which schools
joined the evaluation, longitudinal data were not available for the full seven years for
all schools. Nevertheless, the sample was reasonably large and representative of pri-
mary schools in Northern Ireland, and the year-ahead classes from the same schools
provided a well matched control sample. For example, picture vocabulary and non-
verbal reasoning test scores were available for all cohorts from Year 2 to Year 7, and
showed that the mean developed abilities or proxy IQ scores for the three samples
were almost identical (Control mean = 49.7; EC1 mean = 49.9; EC2 mean = 49.3,
see Table 1). Also, the robust nature of many of the findings across different sub-
groups of pupils, leads us to have confidence in the overall patterns that have been
reported. In addition, the three cohorts of classes were taught by a substantial number
of the same teachers (66%), with higher levels of stability across any two cohorts
(e.g., 88% of Year 1 teachers taught both control and EC1, 71% taught EC1 and
EC2). This level of teacher variation probably reflects what would happen in the roll-
out of any new curriculum.
Moreover, findings from the main effects analyses confirmed established patterns.
As found by Sammons et al. (2007) in the UK EPPE study, girls were better at read-
ing than boys and significantly so, while boys were marginally better, but not signifi-
cantly better, at mathematics. Being young in your class was shown to have negative
effects on attainment, a pattern that is now emerging from different investigations
(Bedard & Dhuey, 2006; Crawford et al., 2007). The effects of socio-economic
disadvantage, measured by free school meals at school level, were in the predicted
direction but were not statistically significant, revealing substantial individual varia-
tion between schools that was not captured by the variables we have included in this
analysis.
treating the EC as a single intervention. By Year 3, the pattern was markedly chang-
ing; a recovery for EC children was underway and by Year 4, the gap had almost
closed. This pattern is consistent with the lack of formal teaching of reading and arith-
metic in the EC classes in Years 1 and 2, at a stage when the PIPS tests were designed
to match the more formal approach in the English curriculum. As more structured
teaching was introduced in Years 3 and 4, EC children caught up quickly, as had been
expected by teachers and training staff, based on their perceptions of the performance of
children in other European countries who begin formal reading instruction only at
age six or seven. Nevertheless, problems were experienced by some Year 3 teachers
(McGuinness et al., 2009b) in terms of the mismatch between teacher expectations
and children’s actual performance on joining these classes; many felt unprepared to
teach children at such an early stage of reading development. This mismatch was
particularly acute for Year 3 teachers who taught the first EC cohort. Importantly, in
the second year of implementation, Year 3 teachers were more prepared for the devel-
opmental stage of the second cohort of EC children who arrived in their classroom
and the test scores probably reflect this. So it cannot be excluded that the process of
change itself, as well as the specifics of change, contributed to some extent to the out-
comes observed. Even highly experienced teachers can behave more like novices
when confronted with unfamiliar pedagogical situations (e.g., Rich, 1993, describing
the introduction of co-operative learning groups); certainly several teachers in the EC
sample reported feeling ‘deskilled’ in the first year. Almost all early years teachers
expressed the conviction that they would ‘do it better’ in subsequent years. Neverthe-
less, whatever the difficulties encountered, the reading test scores for both EC cohorts
matched the control cohort in Year 4.
not rule this out completely. Nevertheless, there are good reasons to make the link
and to examine the differences in implementation practices between the two EC
cohorts and their possible medium-term impacts. While there was year-on-year varia-
tion even in the control classes, the advantage (when there was one) was always statis-
tically with the control group. Despite reaching parity with the control group at Year
4, the EC cohorts (particular the first cohort) were particularly challenged by the Year
5 tests where, in reading, there was a shift in difficulty towards comprehension items
that required substantial reading under timed conditions. This would suggest that the
decoding skills of the EC children were not sufficiently automatic for more challeng-
ing comprehension tests (e.g., Thurlow & van den Broek, 1997) and that again, they
needed to catch up. In Figure 2, the catching up can be seen in the EC2 cohort but
not in EC1. Interview evidence with the teachers throughout the schools confirmed
that there was greater recognition about the implications of the EC in the second year
and teachers made efforts to adjust their practices accordingly—with more or less suc-
cess as the reading test scores indicate.
Although the EC also eschewed mathematics work schemes, there were strong
indications in the interview data that mathematics presented fewer problems for
teachers. They were less anxious about mathematics because they believed that
although children were less advanced in recorded mathematics in Years 1 and 2,
their conceptual understanding and mental mathematics were superior to that of
controls. Despite this, the EC children did not significantly outperform the control
children at any point. A possible reason is that the poorer reading levels had indi-
rect and negative effects on mathematics performance. Since mathematics tests in
Key Stage 2 were taken under greater time pressure, subtle differences in reading
fluency could assume importance in the mathematics tests because lack of automa-
ticity in reading would detract from the cognitive resources available for complet-
ing the mathematics questions, thus slowing the children down (e.g., Vilenius-
Tuohimaa et al., 2008).
What are the implications of these findings for developmentally appropriate and play-based
curricula and associated pedagogies?
The implications of the findings will be considered with reference to the specific con-
text of the EC in Northern Ireland and, drawing from that, will identify the lessons to
be learned for other educational systems with regard to implementation and expecta-
tions about outcomes in the short and medium term.
In common with most developmentally appropriate and play-based curricula,
the EC in Northern Ireland had expectations for children beyond literacy and
numeracy outcomes. The research team have previously reported about the posi-
tive effects of the EC compared to the traditional curriculum on the quality of the
children’s immediate experiences, using a structured classroom observation schedule
(Walsh et al., 2005). This is the first publication to report on the literacy and
numeracy outcomes for the children and several others are in preparation examin-
ing different child outcomes (e.g., attitudes to learning, dispositions) and teachers’
experiences.
more than one year (and sometimes several years) to become ‘settled’ practice in their
Year 1 and Year 2 classrooms, and was subject to adjustment and refinement espe-
cially after the first year. Also, the school principals recognised that changes in the
early years classrooms could not be confined to the early years but had implications
for the whole school.
While acknowledging that the experiences of these 24 schools in Northern Ireland
who volunteered to participate in this innovative curriculum and evaluation are con-
text specific and indeed time-bound (especially in Northern Ireland, where the find-
ings have already informed the design of the Foundation Stage of new curriculum),
there are lessons here for other educational systems that are about to embark, or have
recently embarked, on similar innovations. Play-based curriculum designers need to
have clear expectations with regard to outcomes for literacy and numeracy—they
need to design the curriculum around explicit knowledge about literacy and numer-
acy developmental pathways (and even alternative pathways), to provide the class-
room teachers with extensive professional development and support, and to recognise
the consequences for the whole school.
Lastly, there is an important general message with regard to the evaluation of
educational interventions. The differences between the first and second interven-
tion cohorts, between the short- and medium-term outcomes and the emergence
of issues of intervention coherence and fidelity point to the danger of making
judgments about major interventions too quickly. By their nature, the effects of
such interventions are complex and far-reaching and their full impact cannot be
clear without a correspondingly wide-ranging and carefully designed evaluation. In
the Enriched Curriculum project, the other major strands of the evaluation pro-
vided corroborating data for many of the reports from parents and interviews with
teachers and school principals that allowed increased confidence in the quantita-
tive findings. For a more general overview of the lessons learned and new direc-
tions suggested from evaluating the Enriched Curriculum, see Walsh et al. (2010,
2011).
NOTES
1
See Appendix A for table of correspondence between the English and Northern Ireland year groups.
2
Equivalent to a Local Education Authority in England.
3
Schools specifically set up to cross the religious divide in Northern Ireland.
4
Schools managed by the Catholic Council for Maintained Schools.
5
At the time of the study, it should be noted that children of non-indigenous ethnicity were a very small propor-
tion of the population in Northern Ireland and did not tend to have English Language problems. Therefore
there was no EAL variable.
6
The effect size is the standardised mean difference between two groups and is expressed in terms of units of
standard deviation. For example, if an effect size = .20, it means that the difference between the two groups is
one fifth of a standard deviation.
References
Bedard, K. & Dhuey, E. (2006) The persistence of early childhood maturity: International evidence
of long-run age effects, Quarterly Journal of Economics, 121(4), 1437–1472.
Blenkin, G. M. & Kelly, A. V. (1994) The National Curriculum and early learning (London, Paul
Chapman).
Bredekamp, S. E. (1987) Developmentally appropriate practice in early childhood programs serving chil-
dren from birth through age 8 (Washington, DC, National Association for the Education of
Young Children).
Buchanan, T. K., Burts, D. C., Bidner, J. W. V. F. & Charlesworth, R. (1998) Predictors of the
developmental appropriateness of the beliefs and practices of first, second, and third grade
teachers, Early Childhood Research Quarterly, 13(3), 459–483.
Burts, D. C., Charlesworth, R., Hart, C. H., Fleege, P. O., Mossley, J. & Thomasson, R. H.
(1992) Observed activities and stress behaviors of children in developmentally appropriate and
inappropriate kindergarten classrooms, Early Childhood Research Quarterly, 7(2), 297–318.
Carr, M. & May, H. (2000) Te Whariki: Curriculum voices, in: H. Penn (Ed.) Early childhood
services: Theory, policy and practice (Buckingham, Open University Press), 53–73.
Crawford, C., Dearden, L. & Meghir, C. (2007) When you are born matters: The impact of date of birth
on child cognitive outcomes in England (London, report produced for the Centre for the Econom-
ics of Education).
Edwards, C., Gandini, L. & Forman, G. (1998) The hundred languages of children: The Reggio Emilia
approach―advanced reflections (Santa Barbara, CA, Greenwood).
Goldstein, H. (2011) Multilevel statistical models: 4th edition (London, John Wiley & Sons).
Jambunathan, S., Burts, D. C. & Pierce, S. H. (1999) Developmentally appropriate practices as
predictors of self-competence among preschoolers, Journal of Research in Childhood Education,
13(2), 167–174.
Laevers, F. (1994) The innovative project Experiential Education and the definition of quality in
education, in: F. Laevers (Ed.) Defining and assessing quality in early childhood education
(Leuven, Leuven University Press), 159–174.
Laevers, F. (2000) Forward to basics! Deep-level learning and the experiential approach, Early
Years: An International Journal of Research and Development, 20(2), 20–29.
May, H. & Carr, M. (1997) Making a difference for the under-fives? The early implementation of
Te Whariki, the New Zealand national early childhood curriculum, International Journal of
Early Years Education, 5(3), 225–236.
McGovern, D. (2003) Primary school teachers’ beliefs and practices in early childhood education. Unpub-
lished MSc thesis, Queens University Belfast.
McGuinness, C., Sproule, L., Trew, K. & Walsh, G. (2009a) The Early-years Enriched Curriculum
Evaluation Project: End-of-phase 2, report 1. Overview: evaluation strategy and curriculum
implementation (Belfast, report produced for Northern Ireland Council for Curriculum Assess-
ment and Examinations (CCEA)).
McGuinness, C., Sproule, L., Walsh, G. & Trew, K. (2009b) The Early-years Enriched Curriculum
Evaluation Project: end-of-phase 2, report 2. Inside EC classrooms and schools: Children, teachers and
school principals (Belfast, report produced for Northern Ireland Council for Curriculum Assess-
ment and Examinations (CCEA)).
McNess, E., Triggs, P., Broadfoot, P., Osborn, M. & Pollard, A. (2001) The changing nature of
assessment in English primary classrooms: Findings from the PACE project 1989–1997,
Education 3–13, 29(3), 9–16.
NAEYC (1997) Developmentally appropriate practice in early childhood programs serving children from
birth through age 8 (Washington, DC, National Association for the Education of Young Chil-
dren).
NAEYC (2009) Developmentally appropriate practice in early childhood programs serving children from
birth through age 8 (Washington, DC, National Association for the Education of Young Chil-
dren).
Paris, S. C. (2005) Reinterpreting the development of reading skills, Reading Research Quarterly, 40
(2), 184–202.
Quiery, N., McElhinney, S., Rafferty, H., Sheehy, N. & Trew, K. (2003) Empowering parents: A
two generation intervention in a community context in Northern Ireland, in: I. Katz, J. Pinker-
ton (Eds) Evaluating family support: thinking internationally, thinking critically (London, Wiley),
207–226.
Rich, Y. (1993) Stability and change in teacher expertise, Teacher & Teacher Education, 9(2), 137–
146.
Rinaldi, C. (2005) In dialogue with Reggio Emilia: Listening, researching and learning (London, Routl-
edge).
Sammons, P., Sylva, K., Melhuish, E., Siraj-Blatchford, I., Taggart, B., Grabbe, Y. & Barreau, S.
(2007) The Effective Provision of Pre-school and Primary Education 3-11 Project. Summary report.
Influences on children’s attainment and progress in Key Stage 2: Cognitive outcomes in Year 5 (Lon-
don, Report for the Department for Education and Skills, no. RR828).
Schmidt, H. M., Burts, D. C., Durham, R. S., Charlesworth, R. & Hart, C. H. (2007) Impact of
the developmental appropriateness of teacher guidance strategies on kindergarten children’s
interpersonal relations, Journal of Research in Childhood Education, 21(3), 290–301.
Siraj-Blatchford, I. (1999) Early childhood pedagogy; practice, principles and research, in: P. Morti-
more (Ed.) Understanding pedagogy and its impact on learning (London, Paul Chapman), 20–45.
Siraj-Blatchford, I. (2004) Quality teaching in the early years, in: A. Anning, J. Cullen, M. Fleer
(Eds) Early childhood education, society and culture (London, Sage), 137–148.
Siraj-Blatchford, I. & Sylva, K. (2004) Researching pedagogy in English pre-schools, British Educa-
tional Research Journal, 30(5), 713–730.
Siraj-Blatchford, I., Sylva, K., Muttock, S., Gilden, R. & Bell, D. (2002) Researching effective peda-
gogy in the early years (London, HMSO).
Sproule, L., McGuinness, C., Trew, K., Rafferty, H., Walsh, G., Sheehy, N. & O’Neill, B. (2005)
The Early Years Enriched Curriculum Evaluation Project: End-of-phase 1 report (Belfast, report
produced for the Northern Ireland Council for Curriculum Assessment and Examinations
(CCEA)). Available online at: http://www.nicurriculum.org.uk/research/ (accessed 5 July
2012).
Sproule, L., McGuinness, C., Trew, K., Rafferty, H. & Walsh, G. (2009) The Early Years Enriched
Curriculum Evaluation Project: End-of-phase 2, report 4: Outcomes for pupils over time (Belfast,
report produced for the Northern Ireland Council for Curriculum Assessment and Examina-
tions (CCEA)).
Sproule, L., Rafferty, H., Trew, K., Sheehy, N., McGuinness, C. & Walsh, G. (2001) The Early
Years Enriched Curriculum Evaluation Project: First year report (Belfast, report produced for the
Northern Ireland Council for Curriculum Assessment and Examinations (CCEA)). Available
online at: http://www.nicurriculum.org.uk/research/ (accessed 5 July 2012).
Sproule, L., Trew, K., Rafferty, H., Walsh, G., O’Neill, B., McGuinness, C. & Sheehy, N. (2002)
The Early Years Enriched Curriculum Evaluation Project: Second year report (Belfast, report
produced for the Northern Ireland Council for Curriculum Assessment and Examinations
(CCEA)). Available online at: http://www.nicurriculum.org.uk/research/ (accessed 5 July
2012).
Sproule, L., Trew, K., Rafferty, H., Walsh, G., O’Neill, B., McGuinness, C. & Sheehy, N. (2003)
The Early Years Enriched Curriculum Evaluation Project: Third year report (Belfast, report produced
for the Northern Ireland Council for Curriculum Assessment and Examinations (CCEA)).
Available online at: http://www.nicurriculum.org.uk/research/ (accessed 5 July 2012).
Stephen, C. (2010) Pedagogy: The silent partner in early years learning, Early Years: An Interna-
tional Journal of Research and Development, 30(1), 1–14.
Suggate, S. P. (2009) School entry age and reading achievement in the 2006 programme for
international student assessment (PISA), International Journal of Educational Research, 48(3),
151–161.
Suggate, S. P., Schaughency, E. A. & Reese, E. (2012) Children learning to read later catch up to
children reading earlier, Early Childhood Research Quarterly. http://dx.doi.org/10.1016/j.ecresq.
2012.04.004
Thurlow, R. & van den Broek, P. (1997) Automaticity and inference generation during reading
comprehension, Reading & Writing Quarterly: Overcoming Learning Difficulties, 13(2),
165–181.
Trew, K., Sproule, L., McGuinness, C. & Walsh, G. (2009) The Early Years Enriched Curriculum
Evaluation Project: End-of-phase 2, report 3: Parents’ views and perceptions of the Enriched
Curriculum over time (Belfast, report produced for Northern Ireland Council for Curriculum
Assessment and Examinations (CCEA)).
Tymms, P. (1999) Baseline assessment, value added and the prediction of reading, Journal of
Research in Reading, 22(1), 27–36.
Tymms, P., Jones, P., Albone, S. & Henderson, B. (2009) The first seven years at school, Educa-
tional Assessment, Evaluation and Accountability, 21(1), 67–80.
Tymms, P., Merrill, C. & Jones, P. (2004) Using baseline assessment data to make international
comparisons, British Educational Research Journal, 30(5), 673–690.
Van Horn, M. L., Karlin, E. O., Ramey, S. L., Aldridge, J. & Snyder, S. W. (2005) Effects of
developmentally appropriate practices on children’s development: A review of research and
discussion of methodological and analytic issues, Elementary School Journal, 105(4), 325–
351.
Van Horn, M. L. & Ramey, S. L. (2003) The effects of developmentally appropriate practices on
academic outcomes among former Head Start students and classmates, grades 1–3, American
Educational Research Journal, 40(4), 961–990.
Vilenius-Tuohimaa, P. M., Aunola, K. & Nurmi, J. E. (2008) The association between mathemati-
cal word problems and reading comprehension, Educational Psychology, 28(4), 409–426.
Walsh, G., McGuinness, C., Sproule, L. & Trew, K. (2010) Implementing a play-based and
developmentally appropriate curriculum in Northern Ireland primary schools: What lessons
have we learned? Early Years: An International Journal of Research and Development, 30(1),
53–66.
Walsh, G., Sproule, L., McGuinness, C., Trew, K., Rafferty, H. & Sheehy, N. (2006) An appropri-
ate curriculum for the 4- to 5-year-old child in Northern Ireland: comparing play-based and
formal approaches, Early Years: An International Journal of Research and Development, 26(2),
201–221.
Walsh, G., Sproule, L., McGuinness, C. & Trew, K. (2011) Playful structure: A novel image of
early years pedagogy for primary school classrooms, Early Years: An International Journal of
Research and Development, 31(2), 107–119.
Wood, E. (2007) New directions in play: Consensus or collision? International Journal of Primary,
Elementary and Early Years Education 3–13, 35(4), 309–320.
Wood, L. & Bennett, N. (1997) The rhetoric and reality of play: Teachers’ thinking and classroom
practice, Early Years: An International Journal of Research and Development, 17(2), 22–27.
Woodhead, C. (1999) Is the formal approach better? Early Years Educator, 1(6), 10–11.
According to Goldstein (2011) there are two types of models that may be consid-
ered for this type of data; (i) a conditional model which takes earlier measurements
and uses them as covariates on the right-hand side of any equation that explains a
later outcome measurement, or (ii), an unconditional model which recognises the
multilevel structure of the data and in which all measurements are treated as observed
responses appearing only on the left-hand side of any regression equation. In the con-
text of this particular analysis Goldstein shows that there is a clear preferred model,
the repeated measures model.
The repeated measure approach recognises the Year 0 observation as just one
more outcome which, in this case, contains information on the underlying pupil
ability without any influence of curriculum choice. For observations beyond Year
0, each observation will be a function of pupil ability and the influence of the cur-
riculum, so long as the regression design matrix is structured so that we can disen-
tangle these effects (and by interacting years by curriculum with the exception of
Year 0 it is), then the model can be used to identify the impact of curriculum tak-
ing into account differing pupil abilities without having to have a Year 0 observa-
tion.
In terms of missing data, a complete discussion of the impact of missing data in
repeated measures models is contained in Section 5.9 of Goldstein (2011). Provided
missing data conforms to missing at random (MAR) criteria, ‘balanced data are not a
requirement for efficient results’. There is thus no a priori need to impute data to fill
missing results provided the absent data is not related to any factor which might influ-
ence outcomes. If certain schools had been excluded because their Year 0 results were
especially high or low that year, we could not class the data as MAR. In this case, we
know the nature of mechanism causing the missing data—exogenous administrative
factors that are not caused by the outcomes of the trial. This knowledge then allows
us to define and understand the consequences of the missing data without having to
rely on untestable assumptions. By including school effects on the RHS of the regres-
sion equation we are explicitly conditioning on the source of any bias and thus obtain
unbiased results of the curriculum.
The full model is given by:
where:
pipsjkl refers to the PIPS score of observation j from individual k in school l,
b0 is the constant term applicable at all time points,
b1genjkl is the gender term for individual k, zero for females,
b2birth_monthjkl is the month-of-birth term for individual k, zero for January,
Reading Mathematics
Appendix D. (Continued)
Reading Mathematics
First EC cohort incremental effect (in addition to control 9 time effect): Control 9 EC1 9 time
Year 1 2.59*** 0.66 0.26 1.78** 0.64 0.18
Year 2 6.18*** 0.62 0.62 5.74*** 0.60 0.57
Year 3 1.92* 0.54 0.19 1.15* 0.52 0.12
Year 4 0.85 0.49 0.09 1.16* 0.47 0.12
Year 5 1.81*** 0.54 0.18 1.04* 0.52 0.10
Year 6 1.90*** 0.50 0.19 0.61 0.48 0.06
Year 7 2.39*** 0.55 0.23 1.42** 0.53 0.14
Second EC cohort incremental effect (in addition to control x time effect): Control x EC2 x time
Year 2 6.66*** 0.69 0.67 3.48*** 0.67 0.35
Year 3 1 .23*** 0.56 0.12 0.30 0.54 0.03
Year 4 0.17 0.52 0.02 0.53 0.50 0.05
Year 5 1.71** 0.55 0.17 0.73 0.53 0.07
Year 6 +0.35 0.57 +0.04 +0.02 0.55 +0.00
Year 7 0.41 0.73 0.04 1.38 0.70 0.14
Model variance statistics
r2m (school re) School level 6.10 2.33 7.12 2.59
r21 (pupil re) Pupil level 50.07 1.96 44.84 1.77
r2e (error term) Error variance 23.46 0.51 22.12 0.48
Notes: Significant coefficients are shown in bold type. Levels: *p < .05; **p < .01; ***p < .001.