arxiv: 2604.13190 · v1 · submitted 2026-04-14 · 🧮 math.HO

Recognition: unknown

From Manipulation to Abstraction: The Impact of Flexible Decomposition on Numerical Competence in Primary School

Fabio Pasticci

Pith reviewed 2026-05-10 13:37 UTC · model grok-4.3

classification 🧮 math.HO

keywords decompositionrecompositionnumerical competenceprimary schoolCPA progressionmath educationquasi-experimentalretention

0 comments

The pith

A structured 12-week program teaching flexible decomposition of large numbers produces substantially larger gains in primary students' numerical competence than standard instruction.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines whether guiding year 4 and 5 students through concrete, pictorial, and abstract stages of decomposing and recomposing large numbers improves their performance on number tasks. Experimental classes gained an average of 34 points in year 4 and 29.6 points in year 5 out of 100, compared with 16.4 and 11.1 points in control classes, with the group difference remaining large after controlling for pre-test scores. These gains showed over 97 percent retention after four weeks. A reader would care because number sense forms the foundation for later mathematics learning and the method provides a practical classroom sequence for building it.

Core claim

The intervention, grounded in the Concrete Pictorial Abstract progression, enables students to move from physical manipulation of number representations to abstract flexible decomposition and recomposition, resulting in significantly higher post-intervention scores on numerical competence measures and strong retention of those gains.

What carries the argument

The Concrete Pictorial Abstract (CPA) progression applied to flexible decomposition and recomposition of large numbers, which moves students from hands-on materials through diagrams to symbolic understanding.

If this is right

Experimental students outperform controls by an adjusted average of 18.25 points on the 100-point numerical competence test.
The gains persist at over 97 percent retention after four weeks without additional instruction.
The effect appears in both year 4 and year 5 classes and holds after statistical control for baseline performance.
The approach can be delivered within ordinary classroom time over 12 weeks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the gains replicate in randomized trials, curricula could shift earlier emphasis toward decomposition sequences rather than rote procedures.
The same concrete-to-abstract scaffolding might extend to other topics such as place value or basic operations.
Longer-term tracking would reveal whether early competence gains translate into improved performance on standardized tests two or three years later.

Load-bearing premise

The quasi-experimental design with a control group fully isolates the effect of the decomposition intervention from other differences such as teacher skill or class composition.

What would settle it

A randomized controlled trial with equivalent teachers and random assignment that finds no difference in post-test gains between the decomposition program and standard teaching would show the gains are not caused by the intervention.

read the original abstract

This study examines the effectiveness of a structured instructional approach to decomposition and recomposition of large numbers in six primary school classes (three Year 4 and three Year 5, N = 120) using a quasi - experimental design with a control group. The 12 - week intervention is grounded in the Concrete Pictorial Abstract (CPA) progression. The experimental groups achieved average gains of 34.0 points (Year 4) and 29.6 points (Year 5) out of 100, significantly higher than the control groups (16.4 and 11.1 points; p < .001). The Time Group interaction in the mixed ANOVA reached {\eta}^2p = .931. The ANCOVA with the pre - test as covariate estimated an adjusted difference of 18.25 points (F(1,117) = 2,978.10, p < .001, \eta^2p = .962), confirming the robustness of the effect after controlling for baseline differences. Four-week retention exceeded 97% in the experimental group. Internal reliability of the instrument was satisfactory (Cronbach's {\alpha} = .735).

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Large reported gains from the CPA decomposition intervention, but the class-level quasi-experiment leaves the big effect sizes open to teacher and selection confounds.

read the letter

The paper shows that three experimental primary classes using a 12-week CPA-structured program on number decomposition gained far more on a 100-point test than three control classes: 34 points versus 16 in Year 4 and 30 versus 11 in Year 5, with a Time x Group interaction at partial eta squared .931 and an ANCOVA-adjusted difference of 18 points after baseline control. Retention at four weeks stayed above 97 percent in the experimental arm, and the test had decent internal reliability at alpha .735. That is the core finding, and it is presented with straightforward pre-post and retention numbers plus the usual ANOVA and ANCOVA tables. The work applies the familiar Concrete-Pictorial-Abstract sequence specifically to decomposition and recomposition of larger numbers rather than the usual place-value or addition topics, and the effect sizes are large enough that anyone teaching primary math would notice them. The methods section appears to describe the intervention steps clearly enough for replication attempts. The main limitation is the design itself. Assignment happened at the class level with only three classes per arm and no randomization, so any stable differences in teacher skill, extra instructional time, or class composition could drive the observed interaction. The ANCOVA controls only for pre-test scores; it does not address those group-level factors. The enormous partial eta squared of .962 is consistent with the F-statistic but does not rule out alternative explanations. Without more detail on how the classes were selected or whether teachers were matched on experience, the causal claim stays provisional. This paper is mainly useful to math-education researchers who want concrete examples of CPA applied to decomposition or who are planning their own classroom studies. The numbers are eye-catching and the reporting is transparent, but the evidence is not yet tight enough to change practice on its own. I would send it to peer review so that reviewers can examine the full methods, check for any additional controls, and suggest a follow-up with randomized classes or teacher-level matching.

Referee Report

2 major / 3 minor

Summary. The manuscript reports results from a quasi-experimental study involving 120 primary school students in six classes (three Year 4, three Year 5) testing a 12-week CPA-based intervention on flexible decomposition of large numbers. Experimental groups showed mean gains of 34.0 (Year 4) and 29.6 (Year 5) points versus 16.4 and 11.1 in controls. Mixed ANOVA indicated a Time×Group interaction with η²p=.931, and ANCOVA yielded an adjusted group difference of 18.25 points (F(1,117)=2978.10, p<.001, η²p=.962), with >97% retention at four weeks.

Significance. Should the intervention effect prove robust to design limitations, the large and retained gains would indicate that CPA progression can substantially enhance numerical competence in primary grades, offering a practical framework for mathematics instruction. The high retention rate and internal reliability (α=.735) add to the applied value, though the effect magnitude invites replication in randomized settings.

major comments (2)

[Design and Methods] The quasi-experimental assignment at the class level (three classes per arm) without randomization leaves the large observed effects vulnerable to teacher-specific or compositional confounds. The ANCOVA with pre-test as sole covariate (F(1,117)=2978.10, η²p=.962) does not address these, as the Time×Group interaction (η²p=.931) could arise from unmeasured group-level factors.
[Results and Analysis] The exceptionally large effect sizes (η²p=.931 and .962) and F-statistic warrant additional analysis details; the manuscript should report intra-class correlations or justify the single-level ANCOVA given the clustered structure of classes within years.

minor comments (3)

[Abstract] Formatting artifacts in the abstract such as spaced 'quasi - experimental' and LaTeX fragments like ' {η}^2p ' should be cleaned for readability.
[Methods] Provide more explicit description of the numerical competence instrument, including sample items and scoring to reach the 100-point scale.
[Discussion] Expand the limitations section to discuss the quasi-experimental design and the need for future randomized trials.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our quasi-experimental study. We address the concerns about potential confounds in the design and the need for additional analysis details on clustering below. We propose targeted revisions to the manuscript to improve transparency while acknowledging inherent limitations of the class-level assignment.

read point-by-point responses

Referee: [Design and Methods] The quasi-experimental assignment at the class level (three classes per arm) without randomization leaves the large observed effects vulnerable to teacher-specific or compositional confounds. The ANCOVA with pre-test as sole covariate (F(1,117)=2978.10, η²p=.962) does not address these, as the Time×Group interaction (η²p=.931) could arise from unmeasured group-level factors.

Authors: We agree that the absence of randomization at the class level introduces risks of teacher or compositional confounds that ANCOVA on pre-test scores alone cannot fully eliminate. The manuscript already notes the quasi-experimental design; we will expand the Discussion to explicitly address these threats to internal validity, qualify causal interpretations, and recommend randomized replication studies. No post-hoc changes to the assignment are possible. revision: partial
Referee: [Results and Analysis] The exceptionally large effect sizes (η²p=.931 and .962) and F-statistic warrant additional analysis details; the manuscript should report intra-class correlations or justify the single-level ANCOVA given the clustered structure of classes within years.

Authors: We will add intra-class correlation coefficients for the primary outcome (calculated from the class-level structure) to the Results section and provide a brief justification for retaining the single-level ANCOVA, noting that the low ICC supports treating the data at the individual level after accounting for the pre-test covariate. If multilevel modeling is preferred, we can include it as a sensitivity check in a revision. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical reporting of measured outcomes

full rationale

This is a quasi-experimental study reporting pre-post score gains, mixed ANOVA Time×Group interaction, and ANCOVA results on observed data from 120 students. No mathematical derivations, self-referential definitions, fitted parameters renamed as predictions, or uniqueness theorems appear. Statistical outputs (F-values, η²p) are direct computations from the collected test scores and do not reduce to inputs by construction. The design and analysis contain no load-bearing self-citation chains or ansatzes.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The study relies on standard educational psychology assumptions about learning progressions and statistical assumptions for the ANOVA and ANCOVA analyses. No free parameters or invented entities are introduced as this is an empirical intervention study.

axioms (1)

domain assumption The Concrete Pictorial Abstract progression is an effective framework for teaching numerical concepts.
The intervention is grounded in CPA, assuming its validity for this context.

pith-pipeline@v0.9.0 · 5503 in / 1512 out tokens · 37544 ms · 2026-05-10T13:37:00.022051+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

5 extracted references

[1]

The ability to decompose and recompose numbers constitutes an essential prerequisite for mental arithmetic, written calculation, and estimation (Fuson, 1990; Ross, 1989)

Introduction Understanding the decimal number system is one of the fundamental competencies in primary school mathematics learning. The ability to decompose and recompose numbers constitutes an essential prerequisite for mental arithmetic, written calculation, and estimation (Fuson, 1990; Ross, 1989). However, many students exhibit persistent difficulties...

1990
[2]

Ross (1989) identified five levels of place-value understanding, noting that many students plateau at intermediate levels

Literature Review 2.1 Place Value: Cognitive Challenges Kamii (1986) demonstrated that many children, even by the end of primary school, conceive digits as independent entities rather than as indicators of relative value. Ross (1989) identified five levels of place-value understanding, noting that many students plateau at intermediate levels. Thompson and...

1986
[3]

explainable

as the error term, representing variability between individuals in the same group not explained by any factor. Within-subjects effects (Time and all its interactions) use MS_within = 6.19 (df = 116), representing residual variability of individual pre→post differences. The effect size η²p (partial eta-squared) indicates what proportion of “explainable” va...

1988
[4]

Source of variation SS df MS F p η²p — Between-subjects effects — Group (Exp

Mixed ANOVA — complete results. Source of variation SS df MS F p η²p — Between-subjects effects — Group (Exp. vs Ctrl.) 2,679.08 1 2,679.08 773.29 <.001 .870 Class (Y4 vs Y5) 1,366.88 1 1,366.88 394.54 <.001 .773 Group × Class 8.53 1 8.53 2.46 .131 .021 Between-subjects error 401.88 116 3.46 — — — — Within-subjects effects — Time (Pre vs Post) 62,289.63 1...

1988
[5]

Discussion and Conclusions 5.1 Interpretation of Main Results The results confirm the effectiveness of a structured approach to decomposition and recomposition of large numbers. The observed effect sizes (η²p = .931 in the mixed ANOVA; η²p = .962 in the ANCOVA) indicate an exceptionally high impact: values above .14 are already considered “large” accordin...

1988