Big Ideas

Evidence-based teaching means using the best available research to inform instructional decisions, rather than defaulting to habit, received wisdom, or unverified claims. It does not mean rejecting professional judgement: it means grounding that judgement in what rigorous research actually shows. The key word is "informed": evidence rarely tells you exactly what to do in your specific classroom with your specific learners, but it does significantly narrow the field of approaches worth trying. Coe et al. (2014) describe six common misconceptions about teaching quality, many of which persist because they have surface plausibility but weak evidence bases: for example, praise, discovery learning, and learning styles.

Hattie (2009) identified 0.40 as the average effect size across all educational interventions and called it the hinge point: the level above which an intervention is doing more than the typical teacher does anyway, and therefore worth prioritising. The figure has been influential but also criticised. The main concern is that averaging across wildly different contexts, populations, and study designs produces a number that can be misleading: an intervention with effect size 0.60 in one narrow context may produce 0.10 elsewhere. The hinge point is useful as a rough benchmark for prioritisation, not as a hard threshold for decision-making. Read the evidence on any individual intervention before drawing conclusions from its effect size alone.

Rosenshine (2012) synthesised research from three sources: cognitive science, classroom observation studies of effective teachers, and studies of cognitive supports. His 10 principles include: begin each lesson with a short review of previous learning; present new material in small steps with practice after each step; ask a large number of questions and check all students' responses; provide models; guide practice; check for student understanding; obtain a high success rate; provide scaffolds for difficult tasks; require and monitor independent practice; engage students in weekly and monthly review. The principles are not a rigid script; they are a distillation of what consistently effective teachers do, translated into specific behaviours that any teacher can adopt.

The EEF Toolkit ranks teaching approaches on two dimensions: the estimated additional months of learning progress they produce for an average learner, and the cost per learner. A third dimension is evidence strength, rated from padlocks one to five. The highest-rated approaches combine strong evidence, significant impact, and low cost: metacognition and self-regulation (+7 months, very low cost), feedback (+8 months, very low cost), and reading comprehension strategies (+6 months, low cost). The toolkit is primarily designed for decisions about disadvantaged learners, because it was built from studies with that focus. However, the approaches identified as effective for disadvantaged learners tend to be effective for all learners.

Cognitive load theory, developed by Sweller (1988), is built on one key fact: working memory is severely limited. We can hold approximately four items of information in working memory at any one time, and we can only process information we are currently attending to. When a task places too many simultaneous demands on working memory, learning breaks down. The theory distinguishes three types of load: intrinsic load (the inherent complexity of the material), extraneous load (demands created by poor instructional design), and germane load (the mental effort invested in constructing schemas). Good teaching maximises germane load and minimises extraneous load. Worked examples, reducing split-attention effects, and using both visual and auditory channels are the main practical implications.

The learning styles hypothesis holds that individuals have preferred sensory modalities for receiving information (visual, auditory, kinaesthetic) and that matching instruction to these preferences improves learning. Pashler et al. (2008) reviewed the evidence and found that the hypothesis had not been adequately tested, and where it had been, the results did not support it. The critical test of the hypothesis requires showing that matching instruction to style produces better outcomes than mismatching it; no reliable studies have found this. The persistence of learning styles in education is a useful case study in how an appealing idea can survive in practice long after the evidence has failed to support it. All learners benefit from varied, well-designed instruction that uses multiple representations.

Hattie and Timperley (2007) identified three levels of effective feedback: task level (this answer is wrong because...), process level (the strategy you used here does not work because...), and self-regulation level (next time, check your answer against the original question). Feedback at the self level (you are a great learner) is largely ineffective. The most important finding is that feedback is only effective when learners act on it. This means feedback must be specific enough to enable a clear next action, and lessons must be structured to include time for learners to respond. Written comments that are never returned to, or that are given after the learning episode has ended, have minimal impact. The oral feedback that happens during a lesson in response to a learner's answer is often more effective than extended written marking.