Five principles of effective modelling
Why exemplars so often fail, and how to make modelling genuinely instructive
Modelling is one of those things everyone agrees is important. Students need to see what success looks like. Of course they do. And yet, in practice, modelling too often ends up undermining the very purposes it was meant to bolster.
Sometimes teachers provide a single exemplar and expect students to absorb its quality by osmosis. Sometimes they model too much, producing something so polished and complete that students can only admire it from a distance. Sometimes the model becomes a script to be copied, and what was intended as support turns into imitation. This is an example of the Anna Karenina principle: effective modelling depends on lots of things going right at once, but ineffective modelling can fail in all sorts of different ways; the model may be too close to the target task, too vague in what it demonstrates, too polished to seem attainable, or too poorly timed to be useful. The result is that students only learn how to reproduce the model’s surface features with little understanding of the underlying processes.
For novices, studying a fully worked example is often more effective than being asked to solve a problem from scratch. The reason is simple enough. When students are new to a domain, problem solving extracts a heavy toll on working memory. They must simultaneously hold the goal in mind, search for possible solutions, monitor errors, and try to remember relevant rules. A worked example reduces that burden, allowing attention to settle on the relationship between steps, the sequence of decisions, and the structure of the solution. Instead of floundering through a dark wood, students are given a bird’s-eye view of the terrain with which they can understand the topography before being required to navigate it.
There is, though, an important qualification. Worked examples are most helpful when students are still novices. As expertise grows, their usefulness begins to diminish. This is often described as the expertise reversal effect. What supports a beginner can become unhelpfully constraining for someone more fluent. Once students already grasp the structure of a task, fully worked examples should become redundant. At that stage, completion problems, partial models, critique, comparison, or independent practice will be more effective. The point is not merely to provide models, but to provide the right kind of model at the right moment.
The worked example effect is not an argument for showing students any old finished product but an argument for reducing unnecessary cognitive load so that attention can be focused on what matters right now. If the model is too close to the task, students will copy it; if it’s presented without commentary, students will notice the wrong things; if it’s treated as an object to admire rather than a process to analyse, much of its value is lost.
If modelling is be effective, it must sharpen judgement, direct attention, reveal process, and help students see what is and is not important. These requirements suggest five principles for making modelling genuinely instructive rather than superficial and distracting.
1. Comparison clarifies quality
Comparison is nearly always easier than holistic judgement. It is much simpler to ask, “Which of these is more effective, and why?” than to ask students to evaluate a single response in the abstract. When students compare, they begin to notice difference. One introduction is clearer. One sentence is more precise. One paragraph develops its point more convincingly. Quality emerges through contrast. Without that contrast, weaker students in particular are often left guessing.
In The Eye of the Beholder, and in his later writing on marking1, Donald Laming argues that absolute judgement is an illusion. “There is no absolute judgment. All judgments are comparisons of one thing with another.” Contrary to our beliefs, we are incapable of making accurate judgments of single items. The idea that we can do so reliably is, he suggests, a naive fallacy. In other words, we are much more secure when deciding which of two responses is better than when trying to judge one response in isolation against an abstract standard.
Ask someone whether a single glass of wine is any good and their answer will depend on mood, expectation, what they drank yesterday, and what they usually like. Ask them to compare two wines, though, and they are on much firmer ground. One is richer, one sharper, one more balanced. The same applies to student work. Ask a teacher whether an essay is effective and the answer may be vague. Ask which of two is more effective, and why, and the judgement becomes more precise. Comparison makes quality easier to see.
If a single example reduces the burden of search, comparison reduces the burden of evaluation. A worked example helps students by showing them a route through the task. Comparison helps by making quality easier to perceive. Instead of having to judge one response against an abstract and half-formed notion of excellence, students can inspect two responses side by side and notice which choices are doing the real work. Students are unlikely to be be able to conjure an abstract standard of excellence out of thin air but they can inspect differences side by side and ask which choices produce the best effects.
Consider these examples:
Task: How does Dickens present Scrooge at the start of A Christmas Carol?
Response A: Dickens shows Scrooge is mean and unkind. He does not like Christmas and he is horrible to poor people. This makes the reader think he is a bad man.
Response B: Dickens presents Scrooge as cold and inhuman, not merely through what he says, but through the harsh imagery used to describe him. The description of him as “solitary as an oyster” suggests both isolation and hardness, encouraging the reader to see him as emotionally closed off at the start of the novella.
What students can compare:
Response B is more precise.
It uses evidence more effectively.
It explains the effect of the quotation instead of simply asserting a view.
It moves beyond plot comment into analysis.
Task: Why did the Treaty of Versailles cause resentment in Germany?
Response A: The Treaty of Versailles made Germany angry because it was unfair. Germany had to pay money and lost land. This caused problems and made people upset.
Response B: The Treaty of Versailles caused deep resentment in Germany because many Germans saw it as a humiliating peace. Reparations placed a heavy financial burden on the country, while territorial losses and military restrictions made Germany appear weakened and dishonoured. This sense of national humiliation helped create bitterness towards the settlement.
What students can compare:
Response B gives reasons rather than labels.
It groups details into a coherent explanation.
It uses subject-specific vocabulary such as reparations, territorial losses and humiliation.
It explains why the terms mattered, not just what they were.
Task: Explain why increasing the temperature usually increases the rate of a chemical reaction.
Response A: Increasing temperature makes the reaction go faster because the particles get hotter and react more.
Response B: Increasing temperature usually increases the rate of reaction because the particles gain kinetic energy. This means they move faster and collide more often. It also means a greater proportion of collisions have enough energy to overcome the activation energy, so more collisions are successful.
What students can compare:
Response B is causally sequenced.
It uses precise scientific vocabulary.
It explains the mechanism rather than offering a vague generalisation.
It distinguishes between more collisions and more successful collisions.
In English, students can compare a response that merely says Scrooge is “mean and unkind” with one that analyses Dickens’ imagery. In history, they can compare a vague claim that the Treaty of Versailles was “unfair” with an explanation of how reparations, territorial losses and humiliation produced resentment. In science, they can compare a loose statement that particles “get hotter and react more” with an explanation rooted in kinetic energy, collision frequency and activation energy. In each case, comparison makes it easier to see what quality looks like.
A lone model can seem complete and self-contained. Two models invite analysis. They create the conditions in which students can begin to discriminate between stronger and weaker choices. And that capacity to discriminate is at the heart of improvement.
2. Model structure, not content
In our efforts to support students, we too often produce examples that are too close - or even, identical - to the task at hand. When students can lift content directly, the model ceases to function as a guide to structure and starts to function as a source of borrowed ideas and phrases.
