Why practice problems can overload beginners

In one line: for beginners, working a problem out from scratch can use up the mental room they need to learn from it. Showing them a worked example first often helps more.

What Sweller did

John Sweller wanted to explain a puzzle. Students could solve plenty of practice problems and still not get much better at the next one. Why?

He looked at the way beginners usually solve problems. They look at where they are, look at the goal, and keep asking “what step gets me closer?” Psychologists call this means-ends analysis. It works for getting an answer. Sweller’s hunch was that it takes so much mental effort that there’s little left over for noticing the pattern behind the problem.

To test this, he built a computer model of the process and ran experiments with students working on puzzles and school maths problems. The paper pulled this together with earlier work, including his study with Cooper on worked examples in algebra.

What he found

• Solving from scratch is hard work for the head. Keeping track of the goal, the steps so far and the next move all at once fills up working memory: the small part of memory we use to think about new things.
• Getting the answer isn’t the same as learning. With working memory full, beginners can solve a problem without learning much from it.
• Worked examples free up room. In the earlier algebra study, students who studied solved examples later solved similar problems faster and with fewer mistakes than students who just practised. Sweller’s 1988 paper explains why: studying an example skips the effortful search.
• Not all mental effort is useful effort. Teaching should cut the effort that doesn’t help people learn. This became the core of cognitive load theory.

How much should you trust it?

Strong, well-repeated evidence in the lab; little tested in healthcare.

• The experiments used puzzles and school maths problems: tidy problems with one right answer. Clinical reasoning, communication and practical skills are messier, and may behave differently.
• The worked-example gains were mainly on problems with the same structure. Evidence that they help with quite different problems came later and is more mixed.
• The benefit fades as people gain experience. For someone who already knows the basics, working it out alone can teach more than studying an example.
• The wider theory has a lot of support, but some parts are hard to measure. It isn’t always clear in advance which effort “belongs” to the topic and which is clutter.

If you need to convince someone

In our words (a paraphrase, not a quotation): for beginners, solving problems uses up the working memory that learning needs. Show them how first, then let them try.

What it means for you

1. Let watching an expert count as learning. For a novice, a well-narrated demonstration works rather like a worked example, though the evidence comes from studying solved maths problems, not from clinical settings. “Watch how I take this history, and I’ll say what I’m noticing” comes first. “Now you try one” comes after.
2. Hand over gradually. Show a full example, then one with gaps for them to fill, then let them do it alone. Step back as they get more confident.
3. Clear the space. If a student paramedic or new pharmacy technician is coping with noise, interruptions and nerves, their working memory is already busy. A quiet corner, one clear focus and one thing at a time isn’t being soft. It is design that fits how working memory works.
4. Start from the task. Decide what people need to do differently, then pick an example that shows exactly that.

How it appears in Teaching That Lands

Session 1 gives participants a feel for overload before it’s named: a crowded slide for 30 seconds, then “what do you remember?” The theory then comes in short bursts (working memory, the two kinds of load, two channels), each used straight away in a redesign round. The redesign uses a paper kit, so nobody is fighting the software, and everything they need is there: the skill is choosing. Peers’ redesigned slides act as worked examples. Short theory blocks limit how many new ideas arrive before people use them.

The small print

• Worked examples: the direct evidence comes mainly from Sweller and Cooper (1985). The 1988 paper adds the explanation and the model.
• Expertise reversal: support that helps novices can get in the way for experts (Kalyuga and colleagues, 2003).
• Three kinds of load: later work split load into intrinsic (built into the topic), extraneous (from how it’s taught) and germane (effort that builds understanding). Germane load is contested. Sweller (2010) redefined it as the effort spent on intrinsic load rather than a separate kind, and Kalyuga (2011) argued it isn’t needed at all. Teaching That Lands uses only the first two.
• Working memory model: the paper draws on Baddeley’s (1986) model of working memory, which has since been refined. The paper is a behavioural one: it doesn’t measure the brain.
• Attention span: the popular claim that attention collapses after 10–15 minutes of teaching is poorly supported by primary data (Bradbury, 2016).