The Playbook

High-quality project-based learning: a practical guide for secondary teachers

Project-based learning (PBL) can help students learn subject content while investigating a meaningful question and creating a product or response. Students learn important content through the investigation and production, rather than simply making something at the end of a unit. A worthwhile project needs clear learning goals, explicit teaching, responsive support, and checks on each student’s understanding. PBL is not a substitute for teaching. ^[1][2]

How effective is PBL overall?

There is no single reliable “effect size of PBL” for every project, subject or class. Different studies compare different kinds of PBL with different alternatives and measure different outcomes. Some find benefits; others find little difference. A positive average is not a guarantee for the next project. ^[1][3]

An effect size expresses the difference between two groups in units of the typical spread of scores. For example, +0.20 means that the PBL group’s average was one fifth of a standard deviation higher than the comparison group’s average. It is not a 20% increase in learning or a number of months gained. The sign and size depend on what was measured and what PBL was compared with. ^[4][5]

These findings cover different outcomes, programmes and comparisons. Read them as reference points, rather than estimates of one universal effect:

• Academic achievement, across subjects and education stages: Chen and Yang’s 2019 meta-analysis combined 30 articles and reported an average effect of +0.71 compared with traditional instruction. Subject, location, instructional time and technology support were associated with differences in reported effects. This average covers varied studies. It does not predict the effect in each secondary classroom. The detailed article was not available for close inspection here; the figures are from its abstract. ^[4]
• A mix of knowledge, thinking and attitudes: Zhang and Ma’s 2023 review of 66 experimental or quasi-experimental papers reported +0.44 overall, +0.65 for academic achievement, +0.39 for thinking skills, and +0.39 for affective attitudes. These are broad categories, not evidence that any particular project produces those gains. The authors report substantial variation between results and a statistical indication of possible publication bias; their 190 effects came from 66 papers, so treating results as entirely independent could also affect precision. The authors discuss differing subject averages, but those comparisons cannot establish that one subject inherently benefits more. ^[3]
• Mathematics outcomes: Wang, Zhang and Cheng’s 2026 review reported +0.53 across 33 studies, with a plausible statistical range of +0.38 to +0.68. Adjusting statistically for possible missing studies produced +0.28; this is a sensitivity check, not a known true effect. The review includes different education levels and mathematics outcomes, not exclusively secondary-school achievement. ^[6]
• Motivation and attitudes: Wijnia and colleagues’ 2024 meta-analysis reported +0.50 across project-, problem- and case-based approaches combined, compared with teacher-centred or lecture-based instruction. It included school and higher-education settings; only some studies were project-based. Results varied and the authors detected possible publication bias. This number is not a PBL-only effect or a guarantee that students will be motivated. ^[7]
• A large secondary-school science trial: Schneider and colleagues (2022) tested a particular supported chemistry and physics programme in a randomised study involving more than 4,000 students in California and Michigan. Students in the programme averaged +0.20 on an independently developed science assessment. The intervention included curriculum materials, professional learning and assessment; the result cannot tell us which ingredient caused the benefit. ^[5]

For comparison, Balemen and Keskin’s 2018 science-education meta-analysis reported +1.06 across 48 studies covering primary, secondary and tertiary education. This much larger estimate shows why you need to check the outcome, the comparison teaching and which studies were included. Do not convert it into the paper’s claimed percentage improvement, or assume it will replicate in your class; its full methods were not available for close inspection here. ^[8]

Results also differ between programmes. In a randomised allocation study of a US project-based high school, Craig and Marshall (2019) found broadly similar results to conventional schooling on several state mathematics and science tests, with an advantage on one science measure. Saavedra and colleagues (2022), by contrast, found positive effects on Advanced Placement exams in government and environmental science in a cluster-randomised study. The latter involved schools supportive of PBL and open-enrolment AP policies, so transfer to other settings is not automatic. Both findings can be true because the programme, assessment and setting matter. ^[9][10][11]

Seven priorities for planning a project

The order below is a practical priority. Research has not established a ranking of the separate effects of PBL ingredients. Whole programmes usually change several things at once. We cannot rank these priorities by measured causal impact or add their effect sizes together. Earlier items have stronger supporting evidence or help establish whether learning is happening. Later items still matter, but less research isolates their contribution within PBL. “Confidence” refers to support for the general teaching principle. It does not guarantee an outcome for every implementation. ^[2][1][5]

1. Teach and scaffold the difficult thinking

In class: Identify what students are unlikely to work out on their own. Before or during investigation, teach the relevant concept or method, model an example, provide useful sources and a worked example, then guide practice. Circulate, ask probing questions and adjust support. Remove supports when students are ready, and keep or restore them when needed. ^[12][13]

Evidence and confidence: comparatively strong, mostly from related inquiry research. Lazonder and Harmsen’s 2016 meta-analysis of 72 inquiry-learning studies found that guidance improved learning outcomes compared with less guidance, although the best form varied and age-specific conclusions were limited. Alfieri and colleagues’ 2011 discovery-learning analyses found that explicit instruction generally outperformed unassisted discovery, whereas supported discovery compared favourably with the mixed alternatives in the second analysis. These are not tests of every project design or proof that a single scaffold is best. ^[12][13]

2. Start with precise subject learning, not the product

In class: Decide what students should know and be able to do before choosing a final product. Write a small number of subject-specific goals. Ask: “What must students understand to produce a good response?” Plan a task that requires those ideas, not just an attractive poster or model. Check that lessons, project work and assessment all point to the same goals. ^[2][1]

Evidence and confidence: moderate as a design principle; limited evidence isolating its effect in PBL. Barron and colleagues (1998) identified appropriate learning goals, scaffolds, revision and participation as connected principles in their long-term design work. In a randomised middle-school science study, Harris and colleagues (2015) found that students using project-based materials integrating science practices and disciplinary content outperformed a textbook comparison on aligned science measures. The study tested a package, not alignment alone. ^[2][14]

3. Check understanding during the project and use feedback

In class: Before a major decision or deadline, build in a short individual check: an explanation, sketch, calculation or paragraph. Give specific feedback linked to the learning goal, allow time to act on it, and reteach when a pattern of misunderstanding appears. Provide opportunities to improve an early draft, model or argument. ^[2][15]

Evidence and confidence: moderate for formative assessment generally; weaker for any specific PBL feedback routine. Yao and colleagues’ 2024 meta-analysis reported an average +0.25 for formative assessment across studies from kindergarten to the end of secondary school. A randomised programme in 140 English secondary schools reported +0.09 on national exam attainment. However, an umbrella review found that the certainty of much formative-assessment evidence was low or very low; it could not identify one best technique. Neither number is the effect of adding feedback to a PBL unit specifically. ^[15][16][17]

4. Assess individual learning as well as the finished group product

In class: Use clear criteria for the subject knowledge and reasoning you want to see. Assess the product, but also gather individual evidence through a brief explanation, independent task, oral question or reflection tied to a specific concept. Use these checks to identify students who contributed to a polished group product but still need help. ^[2][1]

Evidence and confidence: moderate for alignment and checking learning; limited direct evidence for a particular assessment mix in PBL. The PBL literature identifies aligned assessment and opportunities for revision as important design features. A group product can demonstrate what the group produced; it does not by itself establish what each student understands. The recommended individual checks follow from the need to assess each student’s intended learning. They are a practical inference, not a proven universal grading formula. ^[1][2]

5. Teach students how to collaborate

In class: Give the group a genuinely shared task, make each student responsible for explaining part of the learning, and teach students how to ask for reasons, listen, disagree and make decisions. Monitor participation and understanding; intervene when one student does all the thinking or another is excluded. Roles can help, but rotate or review them so everyone encounters the key learning. ^[1][18]

Evidence and confidence: moderate for structured cooperative learning; limited for the best group structure in PBL. A meta-analysis of cooperative-learning methods found positive achievement effects across several structured approaches; this is related evidence rather than a direct test of every PBL group arrangement. Kokotsaki and colleagues’ 2016 review identifies individual accountability, participation and group processes as important, but much of the PBL literature reviewed was not randomised. There is no secure universal “best” group size. ^[18][1]

6. Make the question meaningful and offer bounded choice

In class: Frame a question that students can investigate with the knowledge and time available, for example, “How should our community reduce summer heat?” Let them make consequential choices, such as which evidence to examine or which proposal to defend, while keeping the essential learning goals and quality criteria shared. A real audience or public presentation may be useful where feasible, but a simulated audience can also support a meaningful task. ^[7][2]

Evidence and confidence: promising but less certain for the individual ingredients. A meaningful question, inquiry, some student agency and a product are recurring features of PBL designs, not separately proven ingredients with known effect sizes. Wijnia and colleagues’ motivation review found a positive average for several related approaches, but did not establish that authenticity, student choice or a public audience independently caused that result. Avoid treating any one of them as a magic ingredient. ^[1][7]

7. Plan for access, time and teacher support

In class: Anticipate demands beyond the subject content: reading sources, language, planning, equipment, time outside class and collaboration. Offer accessible materials and alternative ways to participate without lowering the core learning goal. Make the timeline realistic, prepare materials and agree with colleagues on who will teach, assess and respond to difficulties. ^[5][14]

Evidence and confidence: moderate for the importance of implementation support; limited evidence identifying one best equity strategy. Successful trials have used coherent materials, teacher professional learning and assessments; this shows that supported PBL can work, not that every component is necessary or sufficient. Saavedra and colleagues observed positive results for both lower- and higher-income student groups in their particular AP setting; that does not establish that PBL automatically closes opportunity gaps. Check local participation and individual outcomes rather than assuming equity. ^[5][14][10]

How explicit teaching and inquiry fit together

The debate is real, but “direct teaching versus PBL” is often the wrong choice. Kirschner, Sweller and Clark (2006) argued that minimally guided instruction can overload novices and is less effective than strongly guided teaching. Hmelo-Silver, Duncan and Chinn (2007) replied that problem-based and inquiry approaches are often scaffolded, and should not be equated with unguided discovery. Their response concerns problem-based and inquiry learning, not a test proving that every project is well guided. ^[19][20]

The reviews give reasons to take both concerns seriously. Learners often need explicit guidance, and inquiry can be structured around meaningful problems. The inquiry meta-analysis estimates are not a direct comparison proving that a particular PBL package beats excellent explicit instruction. More guidance is also not automatically better in every form, for every learner or at every moment. ^[12][13]

In practice: Introduce a motivating question. Teach and model essential knowledge, then let students use it in a supported investigation. Pause to check understanding and reteach where needed. Allow more independence as competence develops. Preserve student thinking and decisions without making students rediscover everything alone. This is an evidence-informed teaching sequence, not a tested universal recipe. ^[2][12]

A quick planning check

Before starting, work through these questions. They turn the principles above into a planning check. ^[2][1][14]

1. Learning: What specific subject knowledge and reasoning should every student gain?
2. Task: Must students use that learning to address the question and create a response?
3. Teaching: What will I explain, model and provide before or during the investigation?
4. Checking: When will I check each student’s understanding, respond and provide time to revise?
5. Participation: How will every student take part in the important thinking, and what barriers need removing?
6. Assessment: What individual evidence will show learning beyond the group’s finished product?
7. Feasibility: Do we have the time, materials and support to run and improve this project?

After the unit, compare students’ individual understanding with the intended goals, examine who participated, and note what to change. ^[2][1] A beautiful product and enthusiastic activity are worth noticing. Neither alone proves that every student learned the intended content.

Reading the evidence responsibly

The broad meta-analyses include different ages and countries, varied comparisons and many non-randomised studies; some reviews may include overlapping studies. Treat large averages with care when studies are small, results vary widely, outcomes closely match the new programme, or positive results are more likely to be published. A statistically adjusted estimate for missing studies is informative but not definitive. ^[1][3][6]

The evidence is stronger for some supported PBL programmes than for claims that PBL always wins, that any single design feature drives success, or that generic critical thinking and collaboration automatically transfer to new contexts. Subject-by-subject effects, ideal project duration, ideal group size, and exactly how much guidance each learner needs remain uncertain; cross-study patterns can be confounded by context and study design. Use the priorities above as a starting point and check their effects in your classroom. ^{[5][4][12][3]}

References

1. Kokotsaki et al., 2016. Project-based learning: A review of the literature. Improving Schools.
2. Barron et al., 1998. Doing with Understanding: Lessons from Research on Problem- and Project-Based Learning. The Journal of the Learning Sciences.
3. Zhang & Ma, 2023. A study of the impact of project-based learning on student learning effects: a meta-analysis study. Frontiers in Psychology.
4. Chen & Yang, 2019. Revisiting the effects of project-based learning on students’ academic achievement: A meta-analysis investigating moderators. Educational Research and Reviews.
5. Schneider et al., 2022. Improving Science Achievement—Is It Possible? Evaluating the Efficacy of a High School Chemistry and Physics Project-Based Learning Intervention. Educational Research.
6. Chunting Wang, 2026. Can project-based learning improve students’ mathematics learning outcomes: a three-level meta-analysis.
7. Wijnia et al., 2024. The Effects of Problem-Based, Project-Based, and Case-Based Learning on Students’ Motivation: a Meta-Analysis. Educational Psychology Review.
8. Balemen & Keskin, 2018. The Effectiveness of Project-Based Learning on Science Education: A Meta-Analysis Search. International Online Journal of Education and Teaching.
9. Craig & Marshall, 2019. Effect of project‐based learning on high school students' state‐mandated, standardized math and science exam performance. Journal of Research in Science Teaching.
10. Saavedra et al., 2022. The Impact of Project-Based Learning on AP Exam Performance. Educational Evaluation and Policy Analysis.
11. Knowledge in Action Validation Study, 2021.
12. Lazonder & Harmsen, 2016. Meta-Analysis of Inquiry-Based Learning. Review of Educational Research.
13. Alfieri et al., 2011. Does Discovery-Based Instruction Enhance Learning?. Journal of Educational Psychology.
14. Harris et al., 2015. Impact of project-based curriculum materials on student learning in science: Results of a randomized controlled trial. Journal of Research in Science Teaching.
15. Yao et al., 2024. The impact of formative assessment on K-12 learning: a meta-analysis. Educational Research and Evaluation.
16. Anders et al., 2022. The Effect of Embedding Formative Assessment on Pupil Attainment. Journal of Research on Educational Effectiveness.
17. Sortwell et al., 2024. A Systematic Review of Meta-Analyses on the Impact of Formative Assessment on K-12 Students’ Learning: Toward Sustainable Quality Education. Sustainability.
18. Johnson et al., 2012. Cooperative learning methods: A meta-analysis.
19. Kirschner et al., 2006. Why Minimal Guidance During Instruction Does Not Work: An Analysis of the Failure of Constructivist, Discovery, Problem-Based, Experiential, and Inquiry-Based Teaching. Educational Psychologist.
20. Hmelo-Silver et al., 2007. Scaffolding and Achievement in Problem-Based and Inquiry Learning: A Response to Kirschner, Sweller, and Clark (2006). Educational Psychologist.