Key findings

Secondary maths

Educator's summary

As a PDF document

As web pages:

- Key findings
Top-rated programmes
Programmes with limited evidence
Other programmes
Review methods

Full report

As a PDF document (113 pages)

What works in teaching maths?

Report summary as a PDF document

Instructional Process Strategies (IP)
(Changing the way the teacher teaches, eg co-operative learning)

The strongest evidence of effectiveness was found for instructional process programmes, as was true in the Slavin & Lake (2008) review of primary school-age maths programmes. Two forms of co-operative learning, STAD and IMPROVE, had a weighted mean effect size of +0.46 across 7 studies, and 4 of these, with a weighted mean effect size of +0.48, used random assignment to conditions. (An effect size of 0.20 or more is considered educationally significant).

The findings for these co-operative learning programmes are in line with those of the primary school-age review, which found a median effect size of +0.29 for cooperative learning (Slavin & Lake, 2008). However, outcomes varied considerably by type of approach, and across 22 qualifying studies the median effect size was +0.18. (An effect size of less than 0.20 is weak).

Information and Communication Technology (ICT) (Programmes based on ICT)

No programme stood out as having notably large and replicated effects. A total of 40 qualifying studies evaluated various forms of computer-assisted instruction. Overall, the weighted mean effect size was +0.08, a small impact. There were few differences among programmes categorised as core (weighted mean ES=+0.09 in 17 studies) and supplemental (weighted mean ES=+0.08 in 20 studies). Computer-managed learning systems (ES=-0.02 in 3 studies) had lower effect sizes.

Mathematics Curricula (MC)

Taken together, there were 40 qualifying studies evaluating various mathematics curricula, with a sample size-weighted mean effect size of only +0.03. This is less than the effect size of +0.10 for primary age-level mathematics curricula reported by Slavin & Lake (2008). There were eight randomised and randomised quasi-experimental studies, also with a weighted mean effect size of +0.03. Effect sizes for the US National Science Foundation -supported textbooks had a weighted mean effect size of 0.00 in 26 studies. However, the NSF programmes add objectives not covered in traditional texts, so to the degree those objectives are seen as valuable, these programmes may be adding impacts not registered on the assessments of content covered in all treatments.

*Overall, 102 studies met the inclusion criteria, of which 28 used random assignment to treatments. These included 40 studies of mathematics curricula, 40 studies of ICT, and 22 studies of instructional process programmes.