Learning by drawing asks the learner to turn an idea into a representation they can inspect, explain and revise. The value is not artistic polish. It is the thinking required to decide what belongs, how the parts relate and whether the drawing agrees with the information. Research has found benefits from drawing in particular memory and educational tasks, but those findings do not mean that every sketch improves understanding. A useful drawing task needs a clear learning purpose, an accuracy check and a later test that goes beyond reproducing the same picture.
This article belongs to the How Learning Works series. Its focus is the learner’s act of constructing and revising a representation. Reading a supplied diagram, remembering a pictured object and explaining a model are related activities, but they are not the same task.
The revealing part is often the first decision
Give a learner this short description: “A team has twelve counters. One third of the counters are selected. Half of the selected counters are then placed in a box.” Ask for a drawing before asking for the final number. Some learners will begin with twelve objects. Others will draw three groups. Someone may divide the entire collection in half too early.
The drawings need not resemble one another. Several can be mathematically valid. The important question is what each mark represents. Is the learner taking half of the selected group or half of all twelve? The first grouping decision exposes a relationship that a final answer alone may conceal.
Now ask the learner to explain the drawing. If they say, “These four are one third of twelve, and these two are half of those four,” the representation and the language agree. If they point to six counters, the drawing has made the mistaken reference group visible. The next teaching move can address that specific relationship.
This is an original illustrative task, not a report of an eduKate experiment. Its answers follow from the stated arithmetic. Its instructional purpose is to show how a drawing can make a learner’s interpretation inspectable before the teacher supplies a complete solution.
Three different jobs for a drawing
A memory drawing is intended to help someone remember an item. A quick sketch associated with a word may serve that purpose even when it is not a detailed explanation. The relevant later question is whether the item can be recalled. It would be a mistake to treat successful recall as proof of conceptual understanding.
An explanatory drawing represents relationships. A partitioned bar, an event sequence or a labelled arrangement should allow another person to inspect what depends on what. Here, the key test is whether the relationships are accurate, not whether the drawing is vivid.
A working drawing helps solve the current problem. It may contain tentative marks, crossed-out alternatives, intermediate values and questions. It need not be a polished study resource. Its purpose is to hold the problem in an inspectable form while the learner reasons.
These jobs can overlap. Nevertheless, naming the job prevents a misleading evaluation. A memorable doodle is not automatically a sound explanatory model. An untidy working sketch may be extremely useful even when it would be unsuitable as a finished revision page. A beautiful copied diagram may reveal almost nothing about what the learner can generate independently.
Before beginning, ask what the drawing should help the learner do afterwards. Remember a term? Explain a relationship? Choose a method? Detect an error? The answer should shape both the drawing instruction and the follow-up task.
What the research actually tested
Wammes, Meade and Fernandes reported seven experiments comparing drawing with writing and other encoding conditions. Their study found a drawing advantage for later free recall of studied information, and they proposed an account involving the integration of semantic, visual and motor processing. This is evidence about the tested memory tasks, not a demonstration that sketching any school topic guarantees transfer or higher grades. Read the original paper’s abstract.
A different study by Schwamborn and colleagues involved 196 ninth-grade students reading a scientific explanation. Students instructed to generate drawings performed better than a reading-only comparison group on the study’s retention, transfer and drawing tests. Drawing accuracy was also associated with later outcomes. The distinction matters: drawing can be investigated both as an activity intended to support learning and as a product that provides evidence about understanding. The researchers’ abstract identifies the sample and outcomes.
Jalava and colleagues subsequently studied drawing and copying course definitions in an undergraduate Biology class. The current publisher abstract reports better recall of associated terms for drawn items after one and three weeks. The measured task was recalling a term from its definition, not independently explaining every biological mechanism. The article also has a linked correction notice, which should remain attached to its publication history. See the current article record and the correction record.
The practical conclusion is narrower and more useful than “drawing works”. Specify the drawing task, the comparison activity, the learner group and the later outcome. Then decide whether a proposed classroom use resembles the evidence closely enough to justify trying it. The workshop below is an educational proposal built around inspectable representations; its specific activities have not been validated as a packaged intervention.
Workshop one: draw the changing reference group
Return to the twelve-counter problem. The required relationships are simple enough to state precisely. There are twelve counters in the original whole. Four are selected because one third of twelve is four. Two are placed in the box because half of four is two. An accurate drawing must preserve this nesting of groups.
A student could draw twelve dots arranged in three groups of four, circle one group and box two dots inside that group. Another could use a bar divided into three equal sections and split one selected section in half. Both can represent the same calculation. Neither is automatically better because it is more pictorial.
Ask the learner to point to the original whole, the selected third and the final half. Require a sentence for each. This makes it difficult to hide an ambiguous picture behind a correct number. A line drawn through the whole bar may mean a partition, a boundary or a crossing out; the learner should explain which.
Next, change only the second instruction: “Half of all the counters are placed in a box.” The correct amount is now six. Ask for the smallest necessary revision to the drawing. A student who keeps boxing two counters may be carrying the previous template rather than rereading the reference group.
Now change the original quantity to eighteen while restoring the original instruction. One third is six and half of that selected third is three. The visual organisation can remain similar while the counts change. This gives the learner a chance to distinguish stable structure from changing quantities.
Finally, ask for a verbal solution without drawing. Then give a new problem and allow the student to choose whether drawing helps. The objective is not to make the learner draw forever. It is to make the reference relationships available for accurate reasoning. In some later tasks, an equation or a short written line may be the more efficient representation.
Workshop two: draw a sequence without inventing a cause
Use this invented passage: “Mira checked the cupboard before leaving home. At the library, she discovered that the notebook was missing. Later, she found it inside a folder in her bag.” Ask the learner to draw a three-panel account of the events. The task is not to produce a comic with artistic detail. It is to preserve the sequence and the information available at each point.
The first panel can show the cupboard check. The second can show the discovery that the notebook was not where Mira expected it. The third can show its location inside the bag. The drawing should not add a thief, a forgotten desk or a deliberate deception. None is stated.
Now ask two different questions. “Where was the notebook found?” is answered directly. “Why did Mira believe it was missing?” may require an interpretation of what she had checked and what she had not. The student should distinguish a detail supplied by the passage from a plausible explanation generated by the reader.
A useful annotation is “stated” beside an explicit detail and “possible inference” beside a proposed interpretation. The drawing becomes an evidence map rather than a substitute story. An arrow between two panels can indicate that one event occurred after another without claiming that the earlier event caused the later one.
For revision, change the final sentence: “Later, a friend returned the notebook.” The student must revise the final panel and any earlier inference that depended on the notebook being in Mira’s bag. This exercise makes revision concrete: a change in the source should change the representation where relevant.
End with a written answer supported by the original wording. Drawing is useful here only insofar as it helps the learner keep sequence, evidence and inference separate. A vivid picture that encourages an unsupported story would be working against the reading task.
Workshop three: represent a process and label what is unknown
Consider a hypothetical sorting process. A box receives ten tokens. A rule sends tokens marked with a triangle to Tray A and all other tokens to Tray B. The description says that four tokens have triangles. Ask the learner to draw the process and show the counts. The system is specified by the task; no real-world experiment is being reported.
A defensible representation shows ten entering, a decision based on the triangle mark, four going to A and six to B. The drawing should identify the decision criterion. Simply connecting the box to two trays does not explain why the tokens separate as they do.
Now remove the information that four tokens have triangles. The route can still be drawn, but the counts cannot be determined. A sound revision replaces the numbers with an unknown quantity or a statement that the count is not given. It does not fill the trays with equal numbers merely to make the picture look balanced.
Next, introduce an incomplete description: “Some tokens are set aside before sorting.” The learner must now mark a boundary on what can be concluded. Ten may enter the initial box, but the number reaching the sorting rule is no longer specified. The visual should preserve that uncertainty instead of silently assuming no tokens were removed.
This kind of drawing task trains a useful discipline: represent what the description permits and mark what it leaves unknown. The finished page can be less visually satisfying than a fully numbered diagram, but it is more faithful to the information. Accuracy includes knowing when not to complete the picture.
How to respond to an inaccurate drawing
Begin by asking the learner what they intended. A line that looks incorrect to the teacher may be a poorly explained convention rather than a conceptual error. Conversely, a plausible-looking arrow may conceal a mistaken relationship. Interpretation should come before correction.
Separate the possible difficulties. A content error puts the wrong relationship into the model. A convention error uses a symbol inconsistently. An omission leaves out necessary information. An unsupported addition includes a claim not supplied or justified. A production difficulty makes a correct intention hard to communicate.
The repair should match the difficulty. Explain the missing concept for a content error. Provide a clear symbol key for a convention error. Ask a targeted question for an omission. Request the source for an unsupported addition. Offer a simpler response format when drawing mechanics are obscuring otherwise sound reasoning.
Do not redraw the whole page immediately. That replaces the learner’s representation with the teacher’s and can erase the evidence of where the misunderstanding began. Ask the learner to revise the relevant part and explain the change. Preserve enough of the earlier version to compare what changed in the model.
A correction is not complete because the new picture looks right. Give a fresh case that requires the same repaired relationship. In the counters task, change the original quantity and the selected fraction. In the narrative task, change the final evidence. In the sorting task, change the rule. The next attempt tests whether the learner can make the decision again.
Assess the relationships, not the artistic performance
The following criteria are a practical discussion guide, not a validated scoring scale. Use them to ask better questions about the learner’s representation. Do not convert them into a diagnosis or claim that a particular total measures general understanding.
| Criterion | Question to ask | Evidence to inspect |
|---|---|---|
| Selection | Are the necessary parts included? | The learner identifies what the task requires and what can be omitted. |
| Relationship | What does each connection mean? | Arrows, groups and positions agree with the explanation. |
| Conditions | What assumptions make this representation valid? | Wholes, units, counts, sequence and decision rules are stated. |
| Boundary | What does the drawing not establish? | Unknowns and unsupported conclusions remain visibly separate. |
| Revision | What changes when the source changes? | The learner modifies the relevant relationship, not merely the decoration. |
Two valid drawings may organise the same information differently. Ask whether each preserves the required relationships. The teacher’s preferred layout should not become the only acceptable answer unless a particular convention is itself the learning objective.
Drawing accuracy can provide useful evidence, as the Schwamborn study illustrates, but a relationship between drawing quality and later performance is not a licence to infer everything about a student from one sketch. Combine the drawing with the learner’s explanation and a fresh task. The original study distinguishes the activity from the predictive value of its product.
Support the drawing without doing the reasoning
A blank page is not always the right starting point. A learner unfamiliar with flow diagrams may need the meaning of a decision point taught first. A learner handling a complex text may benefit from a small set of labelled components. Support can reduce the mechanics of production while leaving the relationships for the learner to decide.
For the sorting task, provide a box and two trays but omit the arrows and labels. For the counter problem, provide twelve dots but no grouping. For the narrative, provide three empty panels but no event labels. Each scaffold leaves a different intellectual decision open.
Inspect what the support is carrying. If the completed diagram is already visible, the learner may mainly be copying. If arrows are supplied, the sequence may already be decided. If labels contain the explanation, filling in a shape may contribute little. A scaffold is useful when it removes an irrelevant obstacle without removing the learning decision.
Fade support selectively. Remove a component label, then ask the learner to choose the grouping, then ask for a fresh representation. Do not remove essential task information or an access accommodation simply to make the exercise look independent. Independence concerns the targeted knowledge and decisions, not doing every physical action without assistance.
There is no requirement to use increasingly elaborate pictures. As the learner understands more, a compact symbolic representation may be sufficient. The goal is a better model, not a permanent commitment to the most visually detailed format.
Keep recall and understanding as separate tests
A student may remember that they drew a bicycle beside a definition while still being unable to explain the definition. That does not make the memory task worthless. It means the result answers a narrower question. A method intended to improve recall should be evaluated with a recall task; a method intended to improve explanation needs an explanation task.
The Wammes experiments are useful precisely because their outcome is specified as free recall. The undergraduate classroom study similarly specifies a later term-recall task. Neither outcome should be casually relabelled as universal conceptual mastery. Wammes and colleagues; Jalava and colleagues.
For understanding, ask the learner to explain why the model has that structure, interpret a new version or repair a deliberate error. For transfer, change the context or representation while retaining the relationship. For durability, return after a delay. These checks should be chosen because they match the learning goal, not because every session needs every possible assessment.
Also distinguish remembering the appearance from reconstructing the meaning. A student may reproduce the teacher’s layout accurately but fail when the same relationship is arranged differently. Ask them to create a second valid representation. The new layout removes some of the support provided by visual imitation.
When drawing is not the best next move
Do not require a drawing merely because the topic is difficult. The obstacle might be an unknown word, a missing fact, an unclear instruction or a procedure that first needs demonstration. A drawing cannot supply knowledge that the learner has not acquired, and an unconstrained picture may hide the gap rather than repair it.
Some tasks need exact quantities or formal relationships better handled in a table or an equation. Others need a carefully worded distinction that a picture cannot express on its own. A learner should be allowed to explain why another representation is more useful. That choice can itself reveal understanding of the task.
Drawing can also consume too much time when the learner treats it as finished artwork. Set a clear stopping condition: the required parts and relationships must be inspectable. Do not demand shading, decoration or repeated copying unless those features serve a stated purpose.
When fine-motor, visual or other access demands interfere, offer alternatives such as arranging labelled objects, describing positions verbally or selecting and connecting prepared components. The response format should not conceal the learner’s knowledge behind an unrelated production barrier.
A manageable routine for home study
Choose one short explanation or problem rather than a whole chapter. State what the drawing should show. Make a first version without chasing neatness. Explain every important mark. Then compare the drawing with the source, noting any omission, contradiction or unsupported addition.
Revise one meaningful error at a time. Write a brief explanation of the change: “I took half of the whole instead of half of the selected part,” or “I drew an exact time even though the passage only gave an order.” This record makes the revision about knowledge rather than cosmetic editing.
Finish with a fresh question that uses the same relationship. The student may draw again, use symbols or explain verbally. The point is to see whether the revised relationship can be used, not to require the same page twice.
For parents, one calm question often reveals more than a long inspection: “Tell me what this connection means.” Follow the answer. If the child can explain it accurately, ask what would change under a different condition. If the explanation is unclear, help locate the missing information without taking over the entire task.
For teachers, retain an occasional early and revised drawing with the learner’s explanation. Compare the changed relationship, not just the final appearance. Such a pair is a record of the particular correction. It is not, by itself, proof that the same understanding will survive every later task.
The drawing should remain answerable to the idea
The strength of learner-generated drawing is that it makes choices visible. The learner chooses the parts, the boundaries, the order and the connections. Those choices can be discussed and corrected. But visibility is not automatically validity. A clear diagram can clearly express a misconception.
Keep the process connected to a source, definition, calculation or other appropriate standard. Ask what justifies the representation. Mark uncertainty when the information is incomplete. Avoid adding a satisfying arrow merely because the page looks unfinished without it.
Most importantly, do not outsource the central decisions and then credit the learner with making them. A supplied model may be an excellent teaching aid. A copied model may be useful practice in notation. They should simply be described honestly. The learner-generated task begins where the learner must decide what the model means and how to represent it.
A successful drawing is not necessarily the neatest one. It is the one the learner can justify, repair and use to reason about another case. That is the point at which the drawing becomes more than a picture on a page.
Research record and next reading
Wammes, Meade and Fernandes (2016), The Drawing Effect examines drawing in free-recall experiments. Schwamborn and colleagues (2010), Drawing as a Generative Activity and Drawing as a Prognostic Activity examines a scientific-text task with ninth-grade students. These are different evidence settings and should not be collapsed into one universal result.
Jalava and colleagues (2023), Drawing Your Way to an A reports delayed term recall in an undergraduate course. Despite the title, the result should not be translated into a promise of an A grade. The publisher links a correction notice; this guide relies on the current abstract for its limited description, not on an independent reanalysis of the data.
Continue to Generation for the value and limits of producing before seeing, Self-Explanation for making relationships explicit, and Conceptual Change when an inaccurate model must be revised. The How Learning Works hub remains the starting point for the wider mechanism library.
