Asking learners to solve realistic problems is an essential part of instruction. The difficult question is not whether learners should practise independently, but when that independence should begin and how much guidance should precede it.
For an experienced practitioner, a problem is rarely encountered as an undifferentiated collection of facts. Familiar features suggest a problem type, a likely method, relevant constraints, and a sequence of actions. A novice does not yet possess these efficient organising structures. Presented with the same problem, the novice may have to search among possible moves, test unproductive approaches, revisit earlier decisions, and hold numerous intermediate results in mind.
Independent problem solving from the outset can therefore demand a great deal of mental effort without producing a corresponding amount of learning. A learner may eventually obtain the correct answer while acquiring little understanding of why the method worked, when it should be used again, or which features of the problem were important.
Worked examples offer a different starting point. By showing a complete solution—or a carefully selected portion of one—they allow attention to be directed towards the structure of the task and the reasoning that connects each step. Properly designed, they are not substitutes for practice. They are a way of preparing learners to practise more intelligently and with progressively less support.
Why novices search instead of recognise
Effective problem solving depends heavily on schemas: organised knowledge structures that represent common problem types, principles, procedures, and relationships. Schemas allow several separate elements to be treated as a single meaningful unit. They help an experienced person recognise that a new problem is structurally similar to one encountered before, even when its surface details differ.
Novices have fewer of these structures available. When they are asked to solve a complex problem independently, they often rely on general search strategies. They may work backwards from the goal, compare the current state with the desired state, and try actions that appear to reduce the difference. This can be useful for obtaining a solution, but it consumes working-memory capacity that might otherwise be used to understand the principles underlying the solution.
The problem is not effort itself. Learning should require thought. The problem is effort spent on activities that contribute little to the knowledge the learner needs to retain. If most available capacity is occupied by keeping track of unsuccessful moves, recovering from errors, or reconstructing an unfamiliar procedure, too little remains for identifying patterns and building a reusable mental model.
The worked-example effect
The worked-example effect describes the finding that novices can learn more efficiently from studying well-designed solutions than from attempting an equivalent series of conventional problems without guidance. Learners who study examples may require less instructional time and make fewer errors while achieving equal or better performance on later problems.
A worked example reduces the need for broad solution search. Instead of asking, “What move could I try next?”, the learner can ask more productive questions: “Why was this method selected?”, “What changed between these two steps?”, “Which principle justifies the operation?”, and “Would the same approach work if this condition were different?”
This advantage is strongest when learners have limited prior knowledge and the task contains multiple interacting elements. It is not a universal preference for demonstrations over activity. A poorly designed example can be as difficult to learn from as an unsupported problem, particularly when important reasoning is omitted or the learner must search across several disconnected sources of information.
Cognitive load and the design of an example
Working memory can process only a limited amount of unfamiliar information at one time. Instructional design cannot remove the inherent complexity of a demanding subject, but it can reduce avoidable demands and organise essential information so that it can be processed coherently.
A strong worked example makes the relationship between the problem, the method, and the solution explicit. Explanatory notes should appear close to the steps they explain. Diagrams, labels, equations, code, or controls should be integrated where possible rather than forcing learners to alternate between distant sources. Each step should contain enough information to make the reasoning intelligible without burying the governing principle beneath incidental detail.
The example should also reflect the way competent performance is actually organised. A demonstration that lists actions without their conditions may teach imitation but not judgement. Learners need to see not only what was done, but the cues that made the action appropriate, the constraints that ruled out alternatives, and the checks used to confirm that the result was plausible.
Completeness does not require exhaustive commentary. Explanations should concentrate on decisions that a novice would not yet be able to infer. Repeating obvious interface descriptions or narrating every visible action can create unnecessary load and obscure the decisions that matter.
Self-explanation turns observation into learning
Simply exposing learners to a correct solution is not enough. Some learners inspect an example superficially, following the sequence without connecting it to prior knowledge or extracting a general method. Self-explanation prompts can make example study more active by requiring learners to explain the reasoning to themselves.
Useful prompts ask learners to identify the principle used in a step, explain why an alternative would be unsuitable, predict the next action before it is revealed, or state how a change in the problem would affect the solution. These activities encourage learners to connect procedures with their underlying conditions and to distinguish general structure from case-specific detail.
Prompts should be selective. Asking for an explanation after every minor action can interrupt the flow of the example and add more work than insight. The most valuable points are usually decisions, transitions, checks, and places where a common misconception is likely to arise.
Self-explanation does not have to take the form of a long written response. Learners can select a rationale, annotate a step, compare two possible approaches, predict an output, or briefly state the rule being applied. What matters is that they process the relationship between the current step and the broader solution.
Completion problems provide the bridge
Moving directly from a fully worked example to an entirely unsupported problem can create too large a change in difficulty. Completion problems offer an intermediate stage. Part of the solution is supplied, while the learner completes the remaining steps.
Early completion tasks may leave only one familiar operation unfinished. Later tasks can remove a decision point, a sequence of calculations, a diagnostic check, or an entire section of the procedure. The learner assumes increasing responsibility while retaining enough structure to avoid reverting to unproductive search.
Completion problems are especially useful because they require production without discarding the context provided by a model solution. They also make misunderstandings more visible. A learner who can recognise a correct step may still be unable to generate it. Requiring selected parts of the solution to be completed exposes this difference before the learner reaches fully independent work.
Fading guidance as expertise develops
Instruction should change as the learner changes. A common progression begins with a complete worked example, moves to examples with increasingly substantial omissions, and ends with independent problems. This gradual fading of guidance aligns instructional support with developing expertise.
Fading should be based on evidence of learning rather than on the simple passage of time. Accurate completion, sound explanations, efficient method selection, and successful transfer to varied problems provide stronger signals than the number of examples viewed. Learners may need support removed at different rates for different parts of the same task.
The sequence of fading also matters. In a tightly dependent procedure, leaving later steps for the learner to complete may preserve a coherent model of how the process begins. In other tasks, asking the learner to plan the opening steps may be more important because method selection is the principal difficulty. The appropriate pattern depends on where the most consequential decisions occur and which capabilities the learner is ready to assume.
Guidance can be faded in more than one dimension. Steps may be removed, explanations shortened, prompts made less specific, visual cues withdrawn, or task variation increased. Eventually the learner should be expected to select the method, execute it, monitor progress, diagnose errors, and evaluate the result without instructional assistance.
Mathematics: seeing the structure behind the calculation
Consider a learner encountering multi-step algebraic equations. An unsupported exercise may prompt experimentation with operations that appear to move the expression closer to an answer. A worked example can instead show the invariant being maintained: the same valid operation is applied to both sides, like terms are consolidated, and the result is checked against the original equation.
A useful sequence might begin with a fully worked solution whose annotations explain why each transformation preserves equality. A second example could ask the learner to predict the next transformation. A completion problem might provide the setup and first operation but require the remaining simplification and verification. Independent problems would then vary the arrangement of terms so that success depends on recognising the structure rather than copying a fixed sequence.
Variation is important. If every example has the same surface form, learners may acquire a narrow routine. Comparing examples that require the same principle in different forms helps establish which features are relevant and supports transfer to unfamiliar problems.
Programming: making invisible reasoning visible
Programming solutions often conceal the reasoning that produced them. A novice can read finished code line by line without understanding how the developer decomposed the problem, selected a data structure, anticipated edge cases, or decided where validation belonged.
A worked programming example should therefore expose decisions as well as syntax. It might begin with a statement of the required behaviour, identify the inputs and constraints, decompose the task into functions, trace a representative case, and then show the completed implementation. Questions can ask learners to predict variable values, explain why a loop terminates, identify the purpose of a condition, or compare two implementations with different performance or maintainability characteristics.
Completion tasks can remove a conditional branch, a function body, a test case, or the error-handling logic. Later, learners can receive only an interface contract and a test suite. The endpoint remains independent construction, but the path to it gives novices a usable model of planning, implementation, testing, and revision.
Science: connecting representations and principles
Scientific problems frequently require coordination among a verbal description, a diagram, a conceptual principle, a mathematical representation, and an interpretation of the result. Novices may focus on a prominent number or familiar formula without first determining whether it applies.
In a mechanics example, the solution can show how the system boundary is selected, how forces are represented, why a particular conservation law or equation is appropriate, and how signs and units are checked. A self-explanation prompt might ask which aspect of the situation justifies the chosen model or what assumption would invalidate it.
Later completion problems could provide the diagram but omit the governing equation, or supply the model while requiring the learner to interpret an unexpected result. Independent tasks should include cases with irrelevant data, different surface contexts, and occasions when a previously used method is inappropriate. This ensures that learners practise deciding, not merely calculating.
Procedural training: from demonstration to controlled execution
In procedural training, a worked example may take the form of a narrated demonstration, annotated sequence, simulated case, or completed record. The critical content includes decision criteria, safety constraints, exception handling, and verification—not just the visible order of actions.
For example, training on equipment calibration might first present a complete procedure that explains the purpose of each check, the tolerances that determine acceptance, and the response to an out-of-range reading. A completion exercise could ask the learner to select the next action at designated decision points. A later simulation could remove prompts while retaining feedback on unsafe or irreversible actions. The final stage would require an independently planned and executed procedure under realistic conditions.
This progression is particularly important when unguided trial and error would be costly, unsafe, or capable of producing misconceptions. Errors can still play a useful role, but they should occur in conditions where they can be diagnosed and corrected without unacceptable consequences.
Why more guidance eventually becomes less useful
The same explanation that helps a novice can become redundant for an experienced learner. Once a schema is established, repeatedly processing steps that can already be generated imposes additional mental work without adding useful knowledge. The learner must reconcile instructional detail with an internal procedure that may be faster, more compact, or better adapted to the situation.
This expertise reversal means that instructional support cannot be classified as effective independently of the learner’s existing knowledge. Complete examples, detailed prompts, and highly segmented demonstrations are often valuable early in learning. For more experienced learners, complex problems, broader goals, reduced prompting, and opportunities to compare or optimise solutions are usually more productive.
Excessive guidance can also restrict strategy selection. Experienced learners may need to integrate several procedures, adapt a method to unusual constraints, or discover a more efficient approach. A rigid demonstration can direct them towards a solution they no longer need and prevent them from exercising the judgement that distinguishes competent performance from procedural compliance.
Designing a progression towards independence
An effective sequence does not treat examples and problems as competing formats. It assigns each a different instructional function. Examples establish or refine a model of successful performance. Explanation prompts encourage learners to process the model. Completion problems require partial production. Independent tasks consolidate performance and reveal whether the learning transfers.
In practice, a progression should:
- Begin with a clear account of the problem type, the goal, and the conditions under which the demonstrated method applies.
- Show decisions and checks, not only the visible sequence of actions.
- Integrate explanations with the steps, diagrams, code, controls, or calculations they describe.
- Use targeted self-explanation prompts at points where principles, choices, or misconceptions are most important.
- Include contrasting examples so learners can distinguish structural features from superficial similarities.
- Introduce completion problems that require increasingly substantial parts of the solution to be generated.
- Fade steps, prompts, cues, and feedback in response to demonstrated competence.
- Use independent practice to build fluency, method selection, monitoring, error recovery, and transfer.
- Reduce or remove explanatory support when it becomes redundant for more experienced learners.
Assessment should match the stage of the progression. During example study, questions can test interpretation and explanation. During completion tasks, assessment can focus on the missing decisions or procedures. At the independent stage, learners should encounter sufficiently varied tasks to demonstrate that they can determine what to do without being told which method the problem is intended to exercise.
Guidance is a route to practice, not an alternative to it
Worked examples are most valuable when they shorten the distance between unfamiliarity and productive practice. They allow novices to invest their limited cognitive resources in understanding a solution before they are required to generate every part of it. Through self-explanation and completion problems, observation becomes increasingly active performance.
Independent problem solving remains indispensable. Learners must eventually retrieve knowledge, select methods, coordinate steps, manage uncertainty, and recover from mistakes without assistance. But independence is an instructional destination, not necessarily the most efficient starting condition.
The strongest designs provide enough guidance to make important structures visible, then withdraw that guidance as those structures become part of the learner’s own expertise. The result is neither passive example study nor premature discovery, but a deliberate movement from supported understanding to capable, adaptable, and independent performance.