Worked examples and scaffolding are forms of guided instruction that help learners acquire a skill by making important steps, decisions, and underlying principles accessible while they are still unfamiliar. A worked example presents a problem together with a demonstrated solution, often including explanations of why each step is valid. Scaffolding is a broader process in which a teacher, peer, instructional resource, or digital system supplies temporary, responsive assistance and transfers responsibility to the learner as competence develops. These approaches are especially relevant when novices must coordinate multiple unfamiliar elements at once. Their educational value depends on the quality of the guidance, the learner's existing knowledge, the opportunities to think about the explanation, and the transition toward independently solving new problems Atkinson et al., 2000; van de Pol et al., 2010.
The central instructional principle is calibration: provide enough information for productive reasoning at the learner's present level, ask learners to make meaningful cognitive contributions, and adjust support when independent performance demonstrates growing competence. Experimental evidence for worked examples is well established in particular cognitive-skill domains, especially structured mathematics and related tasks. Evidence for scaffolding spans observational research, classroom interventions, and digital-learning studies, with greater variation in definitions, implementation, and outcomes. A lesson can generate successful guided performance without producing retention or transfer; both outcomes require assessment after assistance is withdrawn Renkl, 2014; Perry et al., 2021.
What Is a Worked Example?
A worked example is an instructional representation of a task or problem that shows a complete or substantially completed solution process. It identifies the initial conditions, the sequence of operations or judgments, and the result. A strong example also makes the rationale for the sequence inspectable: what goal each step serves, what principle authorizes it, and how learners can recognize when the same principle applies. In algebra, this may mean showing each legal transformation of an equation. In scientific explanation, it can mean demonstrating the movement from a claim to relevant evidence and a warranted inference. In writing, it may mean annotating decisions about a thesis, paragraph organization, and revision rather than merely displaying an exemplary final essay Atkinson et al., 2000; van Gog and Rummel, 2010.
The distinguishing feature is that an essential part of the solution process is made visible to the learner. A bare answer, an unannotated product, or a motivational story about someone succeeding does not provide equivalent information. Worked examples can be presented as text, diagrams, audio explanations, animations, video demonstrations, or expert think-alouds. The relevant design question is whether the representation reveals the important relationships without introducing unnecessary perceptual, linguistic, or organizational work van Gog and Rummel, 2010.
Worked examples, modeling, and demonstrations
A worked example usually represents an explicit solution to a well-specified problem. Modeling describes the demonstration of thought or behavior by a teacher, expert, or peer; it can include gestures, self-correction, uncertainty, and decisions that a static written solution omits. A demonstration may show a physical or intellectual procedure, while a worked example generally makes the problem-to-solution mapping its organizing structure. These formats can be combined: a teacher may solve a problem aloud, display the written steps, and identify the criteria used to select each step. Research on example-based learning connects the cognitive study of written solutions with social-cognitive research on learning from models, while noting that the two literatures investigate different tasks and mechanisms van Gog and Rummel, 2010.
Worked examples and completion problems
A completion problem presents a partly solved example and requires the learner to supply missing steps. It occupies the transition between studying an expert solution and solving a novel problem without help. A teacher might initially show a full equation solution, later omit the final transformation, and eventually provide only the problem statement. These stages can preserve the structure of a familiar task while raising the learner's responsibility in manageable increments. The particular sequence should be adjusted to demonstrated understanding rather than treated as a rigid universal script Renkl and Atkinson, 2003.
What Is Instructional Scaffolding?
Instructional scaffolding is contingent assistance that enables a learner to engage successfully with a task while developing the competence to handle similar tasks with less help. The concept entered the research literature through the study of adult tutoring of young children. Wood, Bruner, and Ross described forms of tutoring such as drawing a learner into the activity, reducing the number of decisions required, marking critical features, maintaining orientation to a goal, controlling frustration, and demonstrating useful actions Wood et al., 1976.
Contemporary analyses identify three linked features: contingency, fading, and transfer of responsibility. Contingency means that support responds to what the learner currently understands and can do. Fading means that support is reduced, transformed, or withdrawn as it becomes less necessary. Transfer of responsibility means that increasingly important decisions and actions become the learner's own. A visual organizer given to everyone throughout a course may be a useful instructional aid; it becomes scaffolding in the stronger research sense when its use is responsive to need and oriented toward independent performance van de Pol et al., 2010.
Scaffolding is a process of adjustment
Teachers sometimes use the word scaffolding for any helpful hint, worksheet, graphic organizer, sentence starter, or simplified activity. Those resources can serve as scaffolds, but the resource alone does not establish that scaffolding took place. A learner who repeatedly succeeds only because the teacher supplies the decisive step has not yet assumed responsibility for that step. Evidence of scaffolding requires observing the relation among learner need, support provided, adaptation, and subsequent performance. This operational distinction is important in research because studies that all claim to evaluate scaffolding may implement fundamentally different interventions van de Pol et al., 2010.
How worked examples fit within scaffolding
Worked examples can be one component of a scaffolded sequence. A complete solution reduces initial search demands; the teacher then asks learners to explain relationships, complete selected steps, compare cases, and solve fresh problems. Other scaffolds can include strategic questions, checklists, feedback, manipulatives, peer dialogue, vocabulary support, or temporary cues. The two terms therefore identify related but distinct levels of instructional design: an example is a specific representational tool, while scaffolding describes the adaptive organization of assistance across learning activity Renkl, 2014; van de Pol et al., 2010.
Historical Development and Theoretical Foundations
Studies of example-based instruction emerged from cognitive research on problem solving and knowledge acquisition. Sweller and Cooper compared algebra learning through conventional problem solving with learning arrangements involving worked solutions, arguing that novice problem solvers can devote substantial mental effort to searching for possible moves without adequately learning reusable structures. Their research established an important experimental foundation for the worked-example effect, although its original evidence arose from specific algebra tasks and populations Sweller and Cooper, 1985.
Instructional scaffolding developed through research on tutoring and learning within social interaction. Its intellectual background includes the importance of assistance from a more knowledgeable participant and the progressive internalization of practices. Wood, Bruner, and Ross's empirical tutoring analysis remains a direct historical source for the educational concept. Later scholarship broadened the meaning of scaffolding to classroom discourse, collaborative problem solving, and digital instruction, creating both new applications and measurement difficulties Wood et al., 1976; van de Pol et al., 2010.
Research on self-explanation supplied a further insight: learners differ substantially in how they process identical examples. Some students spontaneously justify steps, detect gaps, and connect operations to general principles; others copy surface features or depend on the displayed answer. This difference made the quality of example processing a central topic alongside the quality of example presentation Chi et al., 1989.
Cognitive Load and the Mechanism of Guided Learning
Cognitive load theory explains why complete solutions can be useful early in skill acquisition. Working memory has severe limits when processing unfamiliar interacting elements, while knowledge consolidated in long-term memory allows familiar patterns to be treated as integrated units. Novices may need to coordinate definitions, subgoals, operations, and constraints that skilled learners recognize as a single familiar pattern. Guidance can make the relationships easier to process by externalizing expert decisions and reducing fruitless search Sweller et al., 2019.
Search activity and schema construction
When solving a completely unfamiliar problem, learners may select actions mainly by judging whether each action appears to move them closer to the immediate goal. Such means-ends searching can produce an answer without strengthening the general schemas needed for later problems. A worked example reallocates attention toward the structure of a successful solution: the conditions that prompted a transformation, the relationship between successive steps, and the family of problems for which the method is appropriate. This account explains one route by which early guidance may improve the efficiency of initial learning; it does not imply that all productive problem solving is mentally wasteful Sweller and Cooper, 1985; Renkl, 2014.
Intrinsic complexity and avoidable processing demands
Intrinsic cognitive load relates to the interacting complexity of the content relative to a learner's existing knowledge. Extraneous load concerns processing demands produced by the way information and tasks are organized rather than by the learning objective itself. Dividing a complex demonstration into meaningful steps, integrating related labels with diagrams, and removing irrelevant decorative details can reduce avoidable demands. Yet a diagram accompanied by an overlong verbal explanation may introduce redundant processing, especially for knowledgeable students. Cognitive load categories are analytical constructs; researchers continue to debate how precisely they can be distinguished in actual measurements Sweller et al., 2019.
Why reduced cognitive load is not enough
Effort is not inherently undesirable. Learning often requires comparing alternatives, retrieving principles, detecting errors, and constructing explanations. A very easy guided exercise can generate fluency without durable knowledge. The instructional goal is to remove work that adds little learning value while preserving the reasoning necessary to build understanding. For that reason, example design must support meaningful processing, not passive exposure or the memorization of one finished solution Chi et al., 1989; Renkl, 2014.
The Worked-Example Effect: What the Evidence Establishes
The worked-example effect refers to the advantage of studying worked solutions over certain forms of unsupported or less efficiently structured problem solving during the initial acquisition of cognitive skills. A body of controlled studies demonstrates that novices can learn solution schemas efficiently from examples, particularly for mathematically structured problems. The major review by Atkinson and colleagues synthesizes principles for integrating solution components, highlighting the conceptual structure of problems, and supporting self-explanation. It is a research synthesis, not a single uniform effect size applicable to every classroom subject Atkinson et al., 2000.
Interpretation depends on the comparison condition. An example compared with unguided trial and error is not the same experimental contrast as an example compared with a well-supported tutor, a brief explanation followed by practice, or an inquiry sequence with feedback. A study investigating this concern reported that benefits for example study could be observed even against supported problem-solving conditions, including efficiency and conceptual-understanding outcomes. The result strengthens the case that examples are more than a remedy for poorly designed alternatives, while still leaving the size and generality of the advantage dependent on task and treatment characteristics Schwonke et al., 2009.
Performance efficiency and conceptual understanding
A student may obtain an equivalent test score with less instructional time, a higher score after equivalent time, or a higher ratio of performance to experienced effort. These are different results. Claims that worked examples improve learning should specify whether investigators measured acquisition speed, immediate accuracy, conceptual explanation, transfer, or delayed retention. Efficiency is educationally meaningful when the saved time can be used for further reasoning or appropriately challenging practice; it is not a substitute for verifying independent competence Schwonke et al., 2009; Renkl and Atkinson, 2003.
Evidence strength and boundaries
The clearest experimental base concerns novice acquisition of procedural and cognitive skills in reasonably structured domains. The same principles inform science, programming, technical training, and writing, but the transfer of specific effect estimates across subjects requires direct evidence. Large-scale classroom trials, sustained implementation studies, and assessments of remote transfer remain less complete than tightly controlled cognitive experiments. The Education Endowment Foundation's evidence review recognizes the promise of cognitive science approaches and emphasizes gaps in everyday classroom generalizability and coverage across age groups and subjects Perry et al., 2021.
Evidence for Scaffolding in Authentic Learning Settings
The literature on scaffolding is broad and heterogeneous. A foundational review of teacher-student interaction identified contingency, fading, and transfer of responsibility as core features and found a relatively small number of effectiveness studies meeting its conceptual criteria. The review's caution remains important: studies frequently label supports as scaffolding while measuring neither how assistance changes with learner need nor when responsibility passes to the student van de Pol et al., 2010.
A classroom experimental investigation by van de Pol and colleagues examined the relationship between contingent teacher assistance, the time students had for independent work, achievement, effort, and their appreciation of help. The findings underline that responsive support operates inside a classroom ecology: its educational effect depends partly on what learners can do after teachers step away. These results illuminate mechanisms of classroom assistance, but should not be inflated into a universal claim that any teacher prompt or any scaffold improves achievement by the same amount van de Pol et al., 2015.
A 2025 meta-analysis examined 68 experimental studies of prompts in digital learning environments. The authors reported an average positive effect on achievement of d = 0.394, reduced to d = 0.220 following publication-bias adjustment. Effects varied with features of prompts, learning domains, target populations, and regions. Those results concern digitally delivered prompts rather than all scaffolding practices or specifically static worked examples. The bias-adjusted estimate is especially useful because it shows how an attractive pooled result can change when limitations in the evidence are addressed Thomann and Deutscher, 2025.
A 2026 quasi-experimental study of 68 fourth-grade students compared worked examples with question prompts for scientific argumentation. The worked-example group showed stronger results on certain argumentation outcomes and lower measured extraneous cognitive load, particularly among children with lower prior knowledge. The small sample, particular science task, short time horizon, and nonrandomized group assignment limit causal certainty and broad extrapolation. Its contribution is evidence about an identifiable instructional contrast in a primary-school setting, rather than a verdict on all types of prompts or worked examples Effects of Question Prompts and Worked Examples, 2026.
Why Self-Explanation Matters
A student can read every step of a worked solution and still remain uncertain why those steps were taken. Self-explanation refers to generating explanations of the relationships among problem conditions, solution actions, and governing ideas. In a foundational think-aloud study of mechanics problems, more successful learners spontaneously supplied more useful explanations and monitored what they did and did not understand. Such observations identify a plausible learning process, although think-aloud differences among students alone do not demonstrate that any single self-explanation prompt causes improvement in every setting Chi et al., 1989.
The most useful prompts are targeted. Rather than asking learners to explain everything, a teacher can ask why an equation's sign changes, which piece of evidence supports a scientific claim, how a program's loop condition controls termination, or why a counterexample disproves an argument. These prompts bring the deep structure of the example into view. Their quality matters: learners may produce confident but incorrect explanations, and prompts that interrupt every line can fragment attention. Teachers should compare explanations with the disciplinary principle and provide corrective feedback when necessary Renkl, 2014; Atkinson et al., 2000.
Prompts that reveal understanding
Three diagnostic questions serve different purposes. “What is the goal of this step?” checks whether the learner recognizes its function. “Why is this operation justified?” checks the underlying rule. “What would change if a condition changed?” tests whether knowledge extends beyond the presented case. A teacher can use a short explanation, a diagram, or a peer comparison to respond to the answer. This sequence links example study with formative evidence of knowledge rather than relying on the learner's impression that the example looks familiar Chi et al., 1989; Rittle-Johnson and Star, 2007.
Designing Worked Examples for Durable Learning
Effective example design begins by identifying the transferable structure of the target skill. The learner should encounter a problem that is representative of the skill, a correct solution, and an account of the decisions that make the solution appropriate. The example should contain enough detail to make the logic visible, and its format should reduce the need to search between disconnected labels or infer missing steps Atkinson et al., 2000.
Make the problem and the objective explicit
The teacher should establish what the learner will eventually do independently. A procedure for solving a linear equation is a different objective from explaining why equivalent transformations preserve equality. A worked example may support both, but an assessment must separately test conceptual justification and procedural performance. Naming the goal helps determine which steps require explanation and which details can be omitted Rittle-Johnson and Star, 2007.
Segment the solution around meaningful decisions
Break the worked solution into units corresponding to decisions, subgoals, or conceptual relationships, rather than mechanically numbering every tiny movement. In a chemistry calculation, labels may identify selecting the equation, converting units, substituting values, and checking physical plausibility. In a history argument, the segments may distinguish interpreting a source, evaluating its reliability, and connecting it to the thesis. Segmentation is useful insofar as it serves the cognitive structure of the task Atkinson et al., 2000; Sweller et al., 2019.
Integrate explanations with the relevant information
Place brief labels close to the operation, diagram, or sentence they explain, so students can see relationships without repeatedly searching across materials. Avoid duplicating a fully readable written solution with an equally dense spoken script when the redundancy interferes with processing. Modality decisions depend on the learner, task, pacing, and access needs; the principle is coherent attention to the relevant structure, not a mandatory multimedia format Atkinson et al., 2000; Sweller et al., 2019.
Use variation without overwhelming novices
An initial example should clarify one stable pattern. Subsequent examples can change surface features while preserving the principle, and later contrast cases where a different strategy is needed. Students then learn which features are diagnostic and which are incidental. In an experiment with 70 seventh-grade students, comparing alternative equation-solving methods improved procedural knowledge and flexibility relative to examining the same methods separately, with comparable conceptual gains. That finding supports carefully structured comparison within its tested mathematical context, not an unrestricted rule that more examples always improve learning Rittle-Johnson and Star, 2007.
Integrate errors carefully
An incorrect solution can become instructive when learners can identify the precise misconception or invalid step and compare it with a correct explanation. Novices often require enough prior knowledge and feedback to recognize why an error is an error. Presenting several confusing incorrect alternatives without guidance may consume processing capacity or reinforce misconceptions. The choice between correct examples, contrasting examples, and error analysis should therefore follow the intended competency and the learner's readiness, rather than a blanket preference for one format Renkl, 2014.
Fading: How Instructional Support Changes Over Time
Fading is the planned reduction of information or assistance that the learner no longer needs to perform independently. A common sequence is complete worked example, partially completed example, substantially incomplete example, and independent problem. Renkl and Atkinson argued that the transition should be structured rather than abrupt because the demands of problem solving become more appropriate as schemas develop. Their work offers empirical support for example-to-problem fading while also emphasizing the question of how to adapt fading to prior knowledge Renkl and Atkinson, 2003.
Fading does not require a fixed timetable. A teacher can temporarily restore a demonstration if diagnostic responses reveal confusion about a central principle, and can remove familiar steps sooner when a learner shows mastery. The relevant unit of guidance may be one subskill rather than an entire lesson: a student who understands numerical substitution may still need a prompt to choose the appropriate equation. Support is therefore best adjusted by skill component and evidence of performance van de Pol et al., 2010; Renkl, 2014.
Fading by removing steps
In a backward-fading sequence, the learner first completes the final step, then the final two steps, and eventually takes over the entire solution. This arrangement maintains exposure to correct early structure while learners practice decisive closing actions. Other sequences remove early steps or selectively omit decisions with high instructional value. The optimal direction of fading can depend on the structure of the task and the type of knowledge needed Renkl and Atkinson, 2003.
Fading by changing the form of support
A teacher can replace a fully demonstrated operation with a strategic hint, a prompt to name a principle, a short checklist, or a request for independent checking. The learner's responsibility increases even when some contextual support remains. A checklist may persist as a legitimate tool for professional work, while instructional progress is demonstrated by using it strategically and adapting it when conditions change. The benchmark is competent agency with the tools appropriate to the domain van de Pol et al., 2010.
When is a learner ready for less guidance?
Useful evidence includes accurate explanation of essential steps, detection and correction of errors, success on a new but comparable problem, and retention after a short delay. Speed alone may indicate memorization of a visible pattern. A good diagnostic task changes a surface feature or removes the prompt to test whether the learner recognizes the principle. Teachers can begin with a small reduction in support and observe the response before removing more Renkl, 2014.
Expertise Reversal: Why Too Much Guidance Can Hinder Experienced Learners
The expertise-reversal effect describes the interaction between instructional support and learner knowledge: formats that help novices can become redundant, distracting, or less effective for more experienced learners. An advanced student who already knows how to solve a familiar problem may need a novel challenge or comparison of alternative strategies rather than another full demonstration. Reviews of the effect document this interaction across several instructional techniques and explain it through changing knowledge structures and processing demands Kalyuga et al., 2003.
A 2025 meta-analysis synthesized 176 effect sizes from 60 experimental studies involving 5,924 participants. Across the included comparisons, learners with low prior knowledge benefited from higher assistance (d = 0.505), while those with high prior knowledge performed better under lower assistance (d = −0.428). The pattern varied with learning domain, educational level, and how prior knowledge was assessed; evidence was less clear in some younger populations and language or humanities tasks. The asymmetry also matters: the educational case for providing assistance to novices was stronger than a sweeping argument for withholding help from experienced students. These standardized pooled contrasts require interpretation within the particular tasks and measurement designs rather than as universal grade improvements Tetzlaff et al., 2025.
Expertise is task-specific. Someone who is highly skilled at algebra may be a novice at interpreting statistical evidence. A student who has memorized one algorithm may still need substantial assistance with conceptual explanation or choosing between strategies. The appropriate response is to diagnose knowledge of the target task, not to assign all students a permanent status as beginners or experts. The effect also does not imply that expert learners benefit from the absence of all teaching; they may require sophisticated feedback, difficult cases, or access to expert reasoning at a different level Kalyuga et al., 2003; Renkl, 2014.
A controlled study with 98 adolescent mathematics learners reported better retention and lower cognitive load for worked-example instruction compared with problem solving, without a statistically demonstrated advantage on its transfer measure. This divergence illustrates why a learning method can be effective for acquisition while claims about generalized understanding require separate assessment Lee and Ayres, 2024.
Guidance, Productive Struggle, and Independent Problem Solving
Guided instruction and meaningful intellectual challenge belong in the same instructional sequence. Learners need opportunities to make decisions, test understanding, and confront novel conditions. The timing and structure of struggle matter. A student attempting a complicated unfamiliar procedure with minimal information may focus on trial and error; the same student, after studying a principled model, can practice the procedure with enough knowledge to profit from errors. Reviews emphasizing guidance for novices should be interpreted in relation to particular knowledge demands and instructional comparisons Kirschner et al., 2006; Renkl and Atkinson, 2003.
There are research programs investigating problems before explicit instruction, including productive-failure approaches. Their rationale is that an initial attempt can activate prior knowledge, reveal gaps, and make later explanation more meaningful. This work raises important questions about when problem-first instruction can succeed, the types of support present in the task, and how outcomes are measured. It does not erase evidence for worked examples in initial skill acquisition, nor does the worked-example effect settle every choice of instructional sequence. The comparison must specify learners, domain, assessment timing, and the information available in each condition Loibl et al., 2017.
A practical sequence allows learners to examine a correct model, explain its key decisions, complete a related partial problem, solve independently, and then encounter a problem that requires selecting or adapting the method. In some settings, a short initial question can reveal what learners already know before modeling begins. The principle is purposeful calibration of information and challenge, supported by evidence rather than allegiance to a named teaching style Renkl, 2014; Perry et al., 2021.
Worked Examples in Mathematics: A Complete Teaching Sequence
Consider the equation 3(x + 2) = 18. The learning goal is to solve a one-variable linear equation while understanding equivalence-preserving transformations. A teacher first presents a solved model: divide both sides by 3 to obtain x + 2 = 6; subtract 2 from both sides to obtain x = 4; substitute 4 into the original equation to check that 3(4 + 2) = 18. Each action is linked to the principle that equivalent operations on both sides preserve equality. This representation is an illustrative instructional example, not a reported experimental treatment.
A second example, 5(y − 1) = 20, is partly worked: the student chooses how to isolate the expression, completes the subtraction or addition step, and checks the result. A third problem, 4(z + 3) = 28, is presented independently. The teacher then varies the surface arrangement with 2a + 7 = 19 and asks whether the original sequence of transformations must be copied exactly. The learner explains why the method changes while the equivalence principle remains. Gradual completion and variation help the teacher observe whether the student can select operations without the earlier model Renkl and Atkinson, 2003; Rittle-Johnson and Star, 2007.
The final assessment should include an unfamiliar equation and a request to explain a deliberately incorrect transformation. This tests independent procedural performance and conceptual understanding separately. The teacher can revisit a specific step if the learner loses the equivalence concept, while allowing independent work on already mastered steps. The quality of transfer depends on what is asked and measured, not simply whether students completed the guided worksheet successfully Renkl, 2014.
Worked Examples in Science, Writing, and Programming
In science, a worked example can reveal how a problem solver chooses a model, interprets units, evaluates evidence, and checks whether an answer is plausible. For example, a scientific argumentation model can annotate where a claim is supported by evidence and why the reasoning connects the two. A 2026 small-scale classroom study offers preliminary evidence that such examples may outperform question prompts on particular argumentation measures among fourth-graders; the result should be applied within its contextual limitations Effects of Question Prompts and Worked Examples, 2026.
In academic writing, a useful example explains decisions: why a topic sentence organizes the paragraph, how one source supports a precise claim, and how a revision resolves ambiguity. Students then analyze a second paragraph, complete a partial revision, and revise an original passage. This is an instructional application of example-based principles. The claim that it will achieve the same standardized effects as algebra examples would require writing-specific experimental evidence, not analogy alone Atkinson et al., 2000; Perry et al., 2021.
In programming, the worked solution can demonstrate a short function together with its input-output behavior, variable states, and termination conditions. After the learner explains why the code works, a completion exercise removes a line, and an independent task changes the requirements. One major risk is copying syntactic patterns without understanding the underlying program state. Requiring explanation, debugging, and modification can help reveal whether learners have acquired reusable knowledge rather than a solution-specific template Renkl, 2014.
Assessing Retention, Transfer, and Independence
Studies of guided instruction should distinguish at least four outcomes. Immediate performance is the accuracy or quality achieved during or shortly after instruction. Retention is the ability to reproduce relevant knowledge after time has passed. Near transfer is the successful application of a principle to a new problem that shares important structural features. Far transfer requires applying ideas across more distant contexts and is substantially harder to establish. The support that improves performance during a lesson may not produce equal benefits on an unassisted delayed test Renkl, 2014; Perry et al., 2021.
An informative evaluation uses unassisted tests, task variants, and separate assessments of procedural and conceptual knowledge. Researchers should report what comparison group received, whether teaching time was equivalent, whether the intervention was implemented as planned, and whether assessors knew which students received which treatment. Studies should also describe age, prior knowledge, subject, language demands, follow-up interval, and attrition. A positive result from one test administered immediately after exposure supports a narrower claim than a durable effect across independent transfer tasks Rittle-Johnson and Star, 2007; Perry et al., 2021.
How to interpret reported effect sizes
An effect size such as Cohen's d expresses a standardized difference between groups; it is not a percentage of learners helped and does not directly state practical value in a particular school. Effects should be considered alongside the outcome measure, instructional cost, comparison activity, precision of the estimate, heterogeneity, and potential publication bias. The 2025 digital-prompt meta-analysis illustrates this need: its reported pooled d of 0.394 declined to 0.220 with adjustment for publication bias. The appropriate interpretation is an average positive association under the included experimental comparisons, moderated by design and context, rather than a guaranteed benefit from every digital hint Thomann and Deutscher, 2025.
Prior Knowledge, Age, Language, and Inclusion
Students of the same age or grade can enter a lesson with different relevant knowledge. Expertise-reversal research implies that differentiation of guidance should track what a learner knows about the actual target content. A short preassessment can identify learners who require complete demonstrations and those who can move directly to partial problems or novel applications. Temporary flexible grouping can make this feasible, but its value depends on regular reassessment and opportunities for movement between levels of support Kalyuga et al., 2003.
Language also affects access to examples. A learner may understand a mathematical relationship but struggle to parse unfamiliar verbal instructions. Carefully placed terminology, visual representations, explicit statement of goals, and opportunities to explain in accessible language can help reveal the underlying skill. Such accommodations should preserve the intellectual objective while reducing incidental linguistic barriers. Their specific effects, however, require evaluation for the target population and subject; research on mathematics examples does not automatically establish best practices for every multilingual classroom Perry et al., 2021.
For learners with disabilities or neurodevelopmental differences, accessible presentation can include readable formatting, captioning, alternative representations, reduced visual clutter, and additional processing time. Support should be determined by assessed educational need and the required task, rather than a simplistic learning-style classification or diagnostic stereotype. A worked example can be part of an individualized instructional plan; it is not itself a diagnostic procedure or treatment for a developmental condition. Evaluation should include both equitable access and the ability to act independently with appropriate tools Sweller et al., 2019; Perry et al., 2021.
Digital Scaffolding, Adaptive Systems, and Generative AI
Digital environments can deliver examples, completion tasks, response-sensitive hints, and immediate feedback. An adaptive system may observe an incorrect step and offer a contextual explanation rather than revealing the entire answer. The educational quality of this assistance depends on whether the system correctly infers the underlying difficulty, whether the prompt is relevant, and whether support is reduced as the learner improves. Meta-analytic results for digital prompts suggest positive average effects with meaningful heterogeneity and a lower bias-adjusted pooled estimate Thomann and Deutscher, 2025.
Generative AI can produce worked solutions quickly, but fluent text is not proof that an explanation is correct. A generated example may conceal an invalid step, cite nonexistent evidence, or supply the final answer without developing the student's knowledge. An educational deployment should verify solutions against trusted disciplinary resources, expose the reasoning or evidence needed to evaluate them, and test what the student can do after access to the tool is removed. Automatically completing homework is a form of performance assistance whose learning consequences depend on subsequent independent reasoning and assessment. Evidence from research on rule-based instructional prompts cannot be transferred directly to open-ended generative AI tutoring systems Thomann and Deutscher, 2025; Sweller et al., 2019.
Systems that track responses and attention may collect sensitive student data. Schools should specify what data are needed for the instructional purpose, who can access them, how long they are retained, and how erroneous inferences can be corrected. Adaptive guidance should have a clear educational objective and a route toward student autonomy. A recommendation to use AI for this purpose requires independent evidence on accuracy, efficacy, costs, equity, and privacy in the actual educational setting; the existence of adaptive technology alone establishes none of those outcomes.
A Research-Informed Implementation Framework
A coherent implementation begins with a defined independent-learning objective and a short diagnosis of prerequisites. The teacher prepares an accurate worked example that displays the relevant structure, presents it in a clear format, and asks targeted questions about the decisions rather than checking only whether students can repeat the answer. The teacher then uses a partially completed example or a problem with fewer hints and monitors the learner's response Atkinson et al., 2000; Renkl and Atkinson, 2003.
If the student succeeds and can explain the rationale, the next task removes further guidance or introduces a meaningful variation. If the student fails because a prerequisite is missing, the teacher demonstrates that component and reassesses. Independent problems and delayed checks are built into the sequence, since guided completion does not by itself establish lasting understanding. For larger classrooms, brief written explanations, peer discussions with explicit criteria, and low-stakes independent items can provide actionable evidence for adjusting support van de Pol et al., 2015; Rittle-Johnson and Star, 2007.
Instructional quality also requires subject expertise. A formally neat example can still communicate an inaccurate concept, a misleading shortcut, or a rule that works only in the displayed case. Teachers should verify the domain logic, identify likely misconceptions, and revise examples after examining learner explanations and errors. A scalable implementation needs usable preparation time, accessibly designed materials, and evaluation of genuine learning rather than only completion rates Renkl, 2014; Perry et al., 2021.
Frequent Instructional Failures and Their Remedies
A common failure is providing a correct example without explanation of why the steps follow. Learners then reproduce superficial features rather than acquiring a flexible rule. Brief labels and carefully chosen self-explanation questions can make the deep structure visible. Another failure is providing many nearly identical examples while never requiring students to act independently. Completion problems and transfer tasks address this by shifting responsibility at observable points Chi et al., 1989; Renkl and Atkinson, 2003.
Excessive verbalization can also be counterproductive. If every familiar step is explained repeatedly to a proficient learner, the extra guidance may distract from solving the more demanding problem. Conversely, removing help suddenly can overload a novice who has not yet learned the governing concepts. Ongoing diagnosis, timely fading, and selective restoration of support respond to both difficulties Kalyuga et al., 2003; van de Pol et al., 2010.
Finally, a polished group product can mask uneven individual understanding. One student may perform the hard reasoning while others copy. Short individual explanations and independent transfer problems should accompany collaborative work. They distinguish socially supported participation from individual knowledge that can be mobilized without the original group. This distinction is essential when evaluating both classroom scaffolding and digital assistance van de Pol et al., 2015.
What the Research Can and Cannot Yet Conclude
Established evidence supports well-designed worked examples as an efficient route to initial acquisition of many structured cognitive skills among novices, with benefits that depend on comparisons and outcomes. Evidence also supports the educational logic of gradually increasing student responsibility and warns that assistance can become redundant with expertise. However, theories of working-memory constraints do not deliver a single precise prescription for every classroom, and not all examples, fading schedules, or teacher interactions have equally strong experimental validation Sweller and Cooper, 1985; Kalyuga et al., 2003; Renkl, 2014.
The evidence for scaffolding is conceptually and methodologically more diverse. There is support for contingent help in authentic instruction and experimental evidence concerning digital prompts, but findings vary by intervention, population, domain, and setting. Future claims about specific practice must rest on direct comparisons using clear definitions of support, appropriately delayed independent outcomes, and transparent reporting of negative as well as positive findings. The strength of this research lies in explaining when guidance serves learning and how independence can be measured van de Pol et al., 2010; Thomann and Deutscher, 2025; Perry et al., 2021.
Frequently Asked Questions
Are worked examples better than practice problems?
For novices acquiring a complex unfamiliar skill, studying appropriately designed solutions can be more efficient than immediately engaging in poorly supported problem solving. As knowledge grows, independent practice becomes central and examples can be faded. The result depends on what students already know, which tasks they face, and whether the outcome is immediate performance, retention, or transfer Sweller and Cooper, 1985; Renkl and Atkinson, 2003.
What is the difference between a worked example and scaffolding?
A worked example is a demonstrated problem-solution process. Scaffolding is the larger process of providing contingent assistance, fading it, and transferring responsibility. Worked examples may function as one scaffold within an instructional sequence; a fixed example that never responds to student development does not capture the complete scaffolding process van de Pol et al., 2010.
How many examples should students study before solving independently?
Research supports matching example study and practice to knowledge and task complexity, rather than a universal number. A teacher should consider whether students can explain a crucial step, complete a partially solved case, and solve a structurally similar new problem. Additional examples are useful when they clarify a missing principle or introduce a meaningful variation; repetitive exposure without independent tests offers weaker evidence of understanding Renkl, 2014.
Does scaffolding mean making learning easy?
Scaffolding makes a demanding task accessible while preserving the work that builds competence. Appropriate support can remove unproductive search or irrelevant processing while requiring students to justify decisions, compare strategies, and take responsibility for progressively larger parts of the task Sweller et al., 2019; van de Pol et al., 2010.
Can guided instruction reduce creativity or independent thinking?
Rigid repetition can restrict opportunities to select and adapt strategies. A learning sequence that moves from modeling toward variation, explanation, and novel applications uses guidance to prepare independent judgment. Whether a particular method supports creative transfer must be evaluated using relevant tasks and outcomes; performance on one familiar procedure is insufficient evidence Rittle-Johnson and Star, 2007; Renkl, 2014.
Are AI-generated worked solutions reliable educational scaffolds?
They may become useful learning materials after disciplinary verification and when learners are required to examine, explain, and ultimately reproduce the relevant reasoning independently. An AI-generated answer can be incorrect or can substitute for the learning activity, and prompt-based digital-learning evidence is not a direct effectiveness estimate for generative tutoring. Human oversight, transparent limitations, appropriate privacy safeguards, and independent outcome measures remain necessary Thomann and Deutscher, 2025.
Conclusion
Worked examples and scaffolding provide a scientifically grounded approach to teaching unfamiliar skills when guidance is accurate, relevant, and adjusted to learner competence. The essential instructional movement is from observing a reasoned solution, through explaining and completing its key decisions, toward independently selecting and adapting principles in new situations. Strong instruction evaluates that movement through independent reasoning, delayed retention, and transfer rather than through guided task completion alone. The same principle organizes classroom tutoring, written demonstrations, completion problems, and carefully designed digital supports Renkl, 2014; van de Pol et al., 2010.
Related Articles
Cluster Learning and Educational Psychology — the disciplinary overview of learning mechanisms, instruction, motivation, and education research.
Retrieval Practice: Why Testing Yourself Improves Memory — evidence for retrieval as an active learning process after guided acquisition.
Spaced Repetition: How the Spacing Effect Supports Long-Term Memory — scheduling practice for longer-term retention of acquired knowledge.
Education in the Age of AI: Learning, Motivation, Assessment, and Cognitive Development — interdisciplinary treatment of educational technology and the Artificial Era.
Cognitive Load Theory: Working Memory and Instructional Design — Explore intrinsic and extraneous cognitive load, working-memory limits, instructional design, and research debates.
Transfer of Learning: Why Knowledge Does or Does Not Generalize — Investigate near and far transfer, abstraction, analogy, practice design, and limits of generalization.
How Learning Works: Knowledge Acquisition, Understanding, and Practice — Explain how prior knowledge, attention, understanding, practice, and feedback interact to form durable learning.
Prior Knowledge and Misconceptions: How They Shape Learning — Examine how existing knowledge and misconceptions shape understanding, conceptual change, and instruction.
References
Atkinson, R. K., Derry, S. J., Renkl, A., and Wortham, D. (2000). Learning from examples: Instructional principles from the worked examples research. Review of Educational Research, 70(2), 181–214. https://doi.org/10.3102/00346543070002181.
Chi, M. T. H., Bassok, M., Lewis, M. W., Reimann, P., and Glaser, R. (1989). Self-explanations: How students study and use examples in learning to solve problems. Cognitive Science, 13(2), 145–182. https://doi.org/10.1207/s15516709cog1302_1.
Kalyuga, S., Ayres, P., Chandler, P., and Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23–31. https://doi.org/10.1207/S15326985EP3801_4.
Kirschner, P. A., Sweller, J., and Clark, R. E. (2006). Why minimal guidance during instruction does not work. Educational Psychologist, 41(2), 75–86. https://doi.org/10.1207/s15326985ep4102_1.
Lee, H. M., and Ayres, P. (2024). The worked-example effect and a mastery approach goal orientation. Education Sciences, 14(6), Article 597. https://doi.org/10.3390/educsci14060597.
Loibl, K., Roll, I., and Rummel, N. (2017). Towards a theory of when and how problem solving followed by instruction supports learning. Educational Psychology Review, 29, 693–715. https://doi.org/10.1007/s10648-016-9379-x.
Perry, T., Lea, R., Rübner Jørgensen, C., Cordingley, P., Shapiro, K., and Youdell, D. (2021). Cognitive science in the classroom: Evidence and practice review. Education Endowment Foundation. https://educationendowmentfoundation.org.uk/education-evidence/evidence-reviews/cognitive-science-approaches-in-the-classroom/.
Renkl, A. (2014). Toward an instructionally oriented theory of example-based learning. Cognitive Science, 38(1), 1–37. https://doi.org/10.1111/cogs.12086.
Renkl, A., and Atkinson, R. K. (2003). Structuring the transition from example study to problem solving in cognitive skill acquisition: A cognitive load perspective. Educational Psychologist, 38(1), 15–22. https://doi.org/10.1207/S15326985EP3801_3.
Rittle-Johnson, B., and Star, J. R. (2007). Does comparing solution methods facilitate conceptual and procedural knowledge? An experimental study on learning to solve equations. Journal of Educational Psychology, 99(3), 561–574. https://doi.org/10.1037/0022-0663.99.3.561.
Schwonke, R., Renkl, A., Krieg, C., Wittwer, J., Aleven, V., and Salden, R. (2009). The worked-example effect: Not an artefact of lousy control conditions. Computers in Human Behavior, 25(2), 258–266. https://doi.org/10.1016/j.chb.2008.12.011.
Sweller, J., and Cooper, G. A. (1985). The use of worked examples as a substitute for problem solving in learning algebra. Cognition and Instruction, 2(1), 59–89. https://doi.org/10.1207/s1532690xci0201_3.
Sweller, J., van Merriënboer, J. J. G., and Paas, F. (2019). Cognitive architecture and instructional design: 20 years later. Educational Psychology Review, 31, 261–292. https://doi.org/10.1007/s10648-019-09465-5.
Tetzlaff, L., Simonsmeier, B. A., Peters, T., and Brod, G. (2025). A cornerstone of adaptivity: A meta-analysis of the expertise reversal effect. Learning and Instruction, 98, Article 102142. https://doi.org/10.1016/j.learninstruc.2025.102142.
Thomann, H., and Deutscher, V. (2025). Scaffolding through prompts in digital learning: A systematic review and meta-analysis of effectiveness on learning achievement. Educational Research Review, 47, Article 100686. https://doi.org/10.1016/j.edurev.2025.100686.
van de Pol, J., Volman, M., and Beishuizen, J. (2010). Scaffolding in teacher–student interaction: A decade of research. Educational Psychology Review, 22, 271–296. https://doi.org/10.1007/s10648-010-9127-6.
van de Pol, J., Volman, M., Oort, F., and Beishuizen, J. (2015). The effects of scaffolding in the classroom: Support contingency and student independent working time in relation to student achievement, task effort and appreciation of support. Instructional Science, 43, 615–641. https://doi.org/10.1007/s11251-015-9351-z.
van Gog, T., and Rummel, N. (2010). Example-based learning: Integrating cognitive and social-cognitive research perspectives. Educational Psychology Review, 22, 155–174. https://doi.org/10.1007/s10648-010-9134-7.
Wood, D., Bruner, J. S., and Ross, G. (1976). The role of tutoring in problem solving. Journal of Child Psychology and Psychiatry, 17(2), 89–100. https://doi.org/10.1111/j.1469-7610.1976.tb00381.x.