How Intelligence Works | Functional Fixedness and Transfer

Did you know? A familiar object can hide a useful solution precisely because we know its usual job so well. A box looks like a container rather than a support, a ruler looks like a measuring tool rather than a straightedge, and a familiar spreadsheet looks like a finished report instead of transformable data. Functional fixedness is this task-dependent restriction in how people represent possible functions. It is not a permanent measurement of intelligence or creativity.

The practical repair is to separate the intended outcome from the tool normally used to obtain it. Write the requirement in functional language, list the properties of available resources, generate a few plausible alternative roles and then test those alternatives against the real constraints. A new function can be an insight without being a safe or correct solution; verification is the other half of intelligent flexibility.

This guide uses the historical candle-box task and primary developmental research to explain what is known, then shows how students, teams and software workers can practise reframing. It includes original worked arithmetic, ethical and safety boundaries, diagnosed failure patterns and a changed-context transfer exercise. Illustrative figures and scenarios are identified as invented rather than presented as observed school outcomes.

What a familiar function changes

People learn the usual purposes of objects because doing so makes life efficient. A mug holds a drink, a timetable allocates periods and a report summarizes observations. These associations let us act without reconsidering every possibility. However, a new problem may require a property that the usual object label fails to bring to mind. The mug has weight and shape; the timetable reveals overlaps; the report may also provide underlying evidence.

A conventional function is different from a capability. A cardboard box can contain items, separate objects or sometimes provide a small platform, but whether it can support a particular load depends on material and geometry. A database can store records, enforce constraints and preserve concurrent updates, while a CSV file can transport data; neither tool should be assumed adequate for every role merely because its name is familiar.

Functional fixedness concerns the representation of possibilities. It does not tell us that every unusual use is physically possible or ethically permitted. A computer account’s access restrictions remain valid even if bypassing them seems creative. A container unsuitable for a load remains unsuitable even when turning it upside down is an original suggestion. Good problem solving widens a search while preserving necessary constraints.

It is useful to distinguish missing domain knowledge from fixedness. A student may fail to use a magnet for separation because they do not know which materials are magnetic. Another may know the principle but never consider separation because the familiar classroom use was lifting paperclips. The appropriate teaching response is different in these two cases.

One diagnostic question is: what must the object actually do, and which properties could make it do that? Naming the capability before the object opens a route to alternatives. The reasoning becomes clearer when a learner can explain not only a new idea but why it is feasible.

The historical candle-box problem

Functional fixedness is associated with Karl Duncker’s problem-solving research. The classic candle-box task describes a candle, a box of tacks and matches, and a challenge to mount the candle so wax does not drip onto a table. The conceptually important step is seeing the box as a possible platform rather than merely as storage for tacks. Duncker’s 1945 monograph is available via On problem-solving.

Because flames and sharp tacks introduce real hazards, teachers and parents should explain the historic task through a diagram, an unlit paper model or a safe cardboard construction. The psychological principle does not require reenacting the hazardous physical conditions. A battery-powered drawing of a candle can preserve the geometry without requiring the learner to mount or light one.

When the tacks are already inside the box, the box’s container role can be strongly suggested. If the box is presented empty and its contents separately, the alternative support role can become easier to notice in some experimental conditions. The materials are similar but the presentation changes the representation the participant brings to the problem.

This is not evidence that all adults inevitably fail or that a single presentation change universally guarantees success. Task versions, materials, timing and participant experience matter. The careful lesson is that the ordinary role made salient by the environment can narrow what someone considers possible.

Historical research explains a mechanism for investigating problem solving. It should not be turned into a fixed recipe for raising test scores or into a personality label for a learner who does not immediately see the expected answer.

What the child-development evidence shows

In Immunity to functional fixedness in young children, Tim P. German and Margaret Anne Defeyter studied analogous object-use tasks. The authors reported that demonstrating a box’s usual containment function affected older child groups in their studied conditions differently from a younger group. Their focus included how developing knowledge of objects’ usual functions shapes problem solving.

The result is not a universal statement that younger children are smarter or more creative. A child who recognizes a tool’s usual function may handle many everyday tasks more efficiently, even if that experience occasionally makes an unusual use harder to imagine. Development involves accumulating useful knowledge as well as changing how it is represented.

Teachers can reasonably use varied presentations and ask for descriptions of material properties, but the original findings do not guarantee that every instruction will work in every classroom. The relevant evidence comes from the actual tasks, ages, comparisons and outcomes in the research paper.

The authors’ original paper allows interested readers to examine the experimental procedures directly. It is better to inspect those methods than to repeat a simplified story about children spontaneously thinking outside the box.

An educational application should therefore start with observation: what functions did the child consider, what information was available, and which instructions made a usual role salient? Those facts are more informative than an unsupported assumption about a child’s permanent ability.

How a problem becomes restricted

A problem often arrives wrapped in its conventional solution. ‘Hang a picture’ can become ‘find a nail’ before anyone has checked whether a hook, stand or removable fixing would satisfy the actual requirement. In Mathematics, ‘percentage’ can become ‘multiply by a percentage’ even when the question asks for the original amount. In a team, ‘report’ can mean ‘add more columns to the spreadsheet’ when the underlying task is an audit of source information.

Such restrictions can be useful shortcuts when conditions remain stable. They become harmful when the task differs from the one the shortcut was learned to solve. The first step is to make the objective explicit in a way that does not preselect a tool. ‘Keep the label visible for a lesson’ has a different search space from ‘find a clip’.

Next separate verified requirements from assumptions. Safety rules, data permissions and examination instructions may be real boundaries. A customary colour, file format or meeting routine may be optional. Treating the latter as mandatory can block useful ideas; treating the former as optional can cause real harm.

An observer can record the attempted solutions and ask what caused each rejection. Was the alternative physically impossible, unauthorized, irrelevant, or merely unfamiliar? The explanation reveals which mental rule needs to be examined. Failure alone is not diagnostic.

The most useful reframe is short: ‘We need something that can ___ under these conditions.’ That sentence forces attention onto the relation between capability and task, without promoting unusual methods that do not actually work.

A repeatable five-stage repair

First write the finished outcome with a measurable success condition. A student could aim to ‘explain a changed percentage question without help’ rather than ‘do more of the same worksheet’. A team could aim to ‘preserve verified updates without losing another writer’s changes’ rather than ‘save the CSV faster’. Measurable outcomes permit later tests.

Second inventory available properties. Physical properties include mass, flexibility, size and stability; information-system properties include versioning, latency, access control and reversibility. Names such as box, spreadsheet or timetable can be retained, but they must not exhaust the description.

Third generate several candidate uses independently of the favorite one. A container may serve as a separator; a document can become an evidence ledger; a timetable can reveal competition for resources. The goal is to expose possible functions, not to score a contest for the most surprising idea.

Fourth test each candidate against real requirements. A proposed support may fail load limits, a software shortcut may violate permissions, or an examination method may answer a different question. Record the reason for rejection so the search learns from reality instead of treating every idea as equally sound.

Fifth change a relevant condition and solve again. If a support becomes flexible, an earlier geometry may fail. If an original percentage changes, the student should reconstruct the equation. Transfer demonstrates that the reasoning has survived beyond one remembered trick.

A safe teaching activity

Use a cardboard tray, index cards and a paper label. Ask learners to keep the label upright on a desk using only permitted safe materials. A learner may first see the tray as storage and later consider placing it differently to create a support. No flame, sharp metal or heavy load is needed.

Define success before the exercise: the paper label should remain upright for a specified interval without damaging the desk and without anyone holding it. Students should be free to propose multiple valid constructions. The teacher can ask for an explanation of why the structure holds.

To compare different presentations fairly, give groups comparable materials, time and instructions. If one group receives extra hints, its success cannot be attributed simply to the way the box or tray was shown. A classroom demonstration need not be a research study, but it should avoid creating misleading claims.

Ask learners to describe alternatives they rejected. A rigid tray may work as a stand, while a soft one may bend. A proposed use can be inventive and still fail the actual condition. Recognizing that difference is an important part of sound reasoning.

Change the tray material or label width after the first solution. A learner should investigate whether the principle still applies rather than repeat ‘turn the tray over’ unconditionally. The practice is property-based re-evaluation, not imitation of an unusual answer.

Worked case: invention and capacity

Consider a hypothetical paper-model design. A student notices that a box normally used for storage might also function as a platform. The exercise assigns the item a mass of 120 grams. Each fastening point is assumed, purely for a classroom calculation, to support the equivalent of 60 grams under ideal independent load sharing. A factor of two is required by the invented design rubric.

The rubric therefore requires 120×2=240 grams of modeled capacity. Three fastening points supply 3×60=180 grams under the exercise assumptions, falling 60 grams short. The creative reframe has not made this particular design acceptable. It must be changed or rejected.

Five such idealized points would give 5×60=300 grams, exceeding the target by 60 grams. That calculation is not a real structural rating. Real loads are not necessarily shared equally, materials may fail, and a working design needs domain-appropriate safety validation. The figures deliberately serve arithmetic rather than instructions for hanging objects.

Now change the load to 180 grams. With the same model and factor, the required capacity becomes 180×2=360 grams. The previously adequate paper calculation of 300 grams now falls short by 60 grams. A learner relying on the old solution without checking the changed constraint would fail the transfer test.

The lesson is two-sided: a familiar function should not be the only candidate, and a newly imagined function must not be accepted without evidence. Intelligence includes both flexible possibility generation and disciplined error detection.

Worked case: reversing a discount

A student may have learned that percentage questions are solved by multiplying the amount by 1.20, then apply this pattern indiscriminately. Suppose an item costs $72 after a 20% discount. The unfamiliar part of the task is that the original amount is unknown; it would be a mistake to treat the reduced price as the original base.

The reduced price equals 80% of the original, so the equation is $72=0.80×original. Dividing gives original=$72/0.80=$90. Multiplying $72 by 1.20 gives $86.40, which is not the correct inverse of a 20% reduction. The reference whole has changed.

The reframing question is ‘Which quantity represents 100%?’ rather than ‘Which memorized multiplier usually appears with 20%?’ It links the method to the underlying proportion. The answer can be checked by subtracting 20% of $90, which is $18, leaving $72.

Change the discount to 25% and the final price to $90. The final is now 75% of original, so original=$90/0.75=$120. A student who understands the reference quantity can adapt while a student fixed on adding 25% to $90 will find the wrong result.

Not every percentage error is functional fixedness. Some arise from decimal place value, arithmetic skill or reading comprehension. A tutor should diagnose the earliest unstable link before choosing an intervention.

Measuring improvement cautiously

Imagine, solely as a teaching example, two groups of twenty students completing a safe alternative-use puzzle. Eight of twenty in one group and fifteen of twenty in the other solve the problem under the same declared scoring rule. The first observed fraction is 8/20=40%; the second is 15/20=75%. The difference is 35 percentage points.

The ratio of observed rates is 0.75/0.40=1.875. Neither figure proves that the changed presentation caused the observed difference. The groups may differ in preparation, instruction, hints or simple sampling variation. A proper causal study needs a suitable design, not just attractive arithmetic.

Completion rate and time-to-completion are different outcomes. A learner might eventually solve a problem while taking longer to reach the insight; another might generate many novel attempts that fail the conditions. Which outcome matters depends on the declared learning goal.

Repeated puzzles also introduce learning effects. If students see one box-as-support solution, they may repeat it later without showing general capacity to reframe other objects. Changed-context questions are more informative than rehearsing a newly supplied trick.

The published German and Defeyter findings are research evidence under their specific experimental conditions. The 8-versus-15 example above is invented to demonstrate denominators, percentage-point comparisons and the limits of inference. Keeping that boundary visible is part of intellectual honesty.

When experience is a strength and a constraint

Expertise should not be treated as the enemy of flexibility. A trained electrician recognizes unsafe wiring faster than a novice, and a skilled teacher can detect missing prerequisite knowledge. Familiar functional associations often embody valuable evidence. A good intervention asks whether the familiar method still fits, not whether familiarity itself is bad.

A problem becomes interesting when its constraints differ from earlier work. An old method may cease to apply because the load is different, permission has changed, the input representation is new or a student is solving for an unknown original rather than a forward result. Recognizing the change is a sign of understanding.

Experienced teams can deliberately conduct a brief alternative-use review before expensive commitments. Ask one person to state the required outcome, another to list current tools and their properties, and a third to identify assumptions that might not actually be required. Evidence should decide the final choice.

A person who can explain why the conventional method remains best after considering alternatives may be demonstrating excellent judgement. Flexibility does not require choosing the strange option; it requires keeping the space of legitimate possibilities open until evaluation.

Organisations, titles and roles

Organizations also develop fixed representations of people’s capabilities. A librarian may be consulted only about books despite expertise in source evaluation. A tester may be invited only after software is written although their knowledge could improve acceptance criteria before coding. A scheduler may be treated as a calendar operator rather than someone who understands competing resource constraints.

Expanding a role should be done with consent and realistic workload boundaries. Reframing functions does not justify assigning unpaid or unsafe responsibilities to colleagues. An alternative is meaningful only if the person has appropriate authority, information and support.

An organization might also equate a weekly meeting with collaboration. The intended outcome could instead be decisions recorded, risks surfaced or dependencies resolved. A shorter review, written decision log or targeted escalation may satisfy some goals more efficiently under the current team structure.

Standardization can still be valuable. A reliable routine may reduce accidents, onboarding effort and miscommunication. The repair is not a demand for perpetual novelty but a periodic check that the customary function matches the present job.

Software tools and hidden assumptions

Consider a service that began by storing customer observations in CSV files. It was sensible for a small single-writer project. Two automated workers are now asked to edit records concurrently, and a developer proposes that both simply read, change and rewrite the same file. The familiar representation of the CSV as the project’s database obscures the need for concurrency control.

If both workers read version five, worker A may add twelve records while worker B corrects three fields. If B writes its old snapshot after A, B may erase the twelve records even though both edits were valid. Unique row identifiers do not prevent the lost-update problem when an entire stale file is replaced.

The goal is an auditable and consistent history of updates. A transactional datastore or version-guarded write protocol may fit that requirement better, while CSV remains useful for export. One familiar object can retain a suitable role without being forced to perform every role.

Now suppose an external reviewer accepts only CSV. The system can export a verified snapshot after a durable commit. That changed constraint does not require abandoning transactional storage. It reveals how different tools can take complementary functions when the original objective is separated from the customary implementation.

Attention to genuine constraints

The word constraint must not become an excuse for inflexible thinking, but neither may it become an obstacle to ignore. Physical stability, data permissions and accurate arithmetic are real limits. The proposed alternative must satisfy those limits or explain how the design changes to comply.

For a physical problem, examine weight, geometry, friction, material compatibility and safe handling. For a software problem, examine ownership, transaction isolation, sensitive data and replay risk. For a school task, inspect what the question actually asks and whether the proposed reasoning proves the answer.

A learner can mark each statement as confirmed, assumed or unknown. Confirmed restrictions should remain until new evidence legitimately changes them. Assumed restrictions are candidates for reframing. Unknown restrictions require investigation before the alternative can be declared valid.

That classification prevents two opposite errors: rigidly dismissing every unfamiliar idea and recklessly accepting any exciting workaround. Disciplined creativity is an investigation, not a permission bypass.

A specific diagnostic for teachers

Begin with one ordinary example the student can explain correctly. Then change a relevant condition without telling the learner which operation to use. Ask what the original method depended on, what has changed and which relationships still hold. This reveals whether the issue is memorization, reading or missing conceptual knowledge.

When a student is stuck, ask them to describe the unknown and the evidence already provided. A fraction problem may need a different denominator; a composition task may need a different narrator; a science observation may require distinguishing correlation from mechanism. Avoid giving the familiar step so quickly that the student never examines its assumptions.

Collect the student’s reasoning, not only the final mark. A valid unexpected route should be recognized, while a creative-sounding answer with a wrong denominator should be corrected. This makes flexibility compatible with accuracy.

On a later day, test a new context instead of repeating the same diagram. The goal is independent transfer. Immediate success after a tutor’s hint is useful evidence of guided learning but not yet evidence that the learner can reconstruct the method alone.

Diagnosing false creativity

Some interventions produce many ideas without improving solutions. A group may suggest using a paper box as a bridge or a document as a password without checking feasibility or security. The number of suggestions is not a reliable measure of successful problem solving.

Another failure is substituting an unusual word for the ordinary one while keeping the same incorrect mental model. Describing a spreadsheet as a ‘data universe’ does not provide transaction guarantees or resolve ownership. New language helps only when it reveals new properties or better matches the actual requirement.

An instructor can ask for one predicted consequence of each candidate and a way to check it. If the prediction cannot be tested, the idea may require more specification. If it fails a known safety rule, reject or redesign it rather than rewarding its originality.

The most useful outcome is a small number of alternatives with clear mechanisms, risks and evidence. Inventiveness and discipline should reinforce each other.

A miniature independent comparison

Suppose an invented tutor runs a paper-only task with two otherwise similar versions. One version names the object by its usual role; the other describes its shape and properties. The tutor first writes down the success rule, time allowance and hint policy. Without this preparation, an apparent difference in success may reflect inconsistent instructions.

Imagine 40 participants in total, 20 in each version. If ten finish in the first version and fourteen in the second, the observed rates are 10/20=50% and 14/20=70%. The observed difference is 20 percentage points. These invented counts illustrate analysis; they are not results from a conducted study.

Even this comparison does not establish causality by arithmetic alone. Assignment, background experience, the facilitator and the number of prior puzzles may differ. A serious experiment needs a design that handles such alternatives and an uncertainty analysis suited to the actual sample.

A teacher need not run a formal experiment to help learners, but should avoid claiming a universal effect from a small class activity. The observational notes can instead guide the next lesson: which object function was assumed, what alternative was missed and what evidence changed the learner’s mind.

Transfer exercise: reusing an information artifact

Imagine a project whose primary deliverable has always been a slide presentation. A team now needs readers to verify where a science conclusion came from. The members assume the answer is more slides, because slides are what presentations are for. The real outcome, however, is transparent evidence.

A separate data appendix with observations, units and calculations may better support auditability, while slides communicate the main result. If measurements are 12, 15 and 18 units, the mean is (12+15+18)/3=45/3=15 units. A static slide showing ’15’ does not allow a reader to inspect the denominator unless the underlying values are also preserved.

Change the dataset by adding 27 units. The revised mean is (12+15+18+27)/4=72/4=18 units. A learner who manually changes a headline without updating the source table may introduce an inconsistency. The design should make the update traceable.

Change the audience: some people cannot open spreadsheet files. An accessible table in a document might better preserve the evidence for them. The useful function is verifiable communication; no single application should be treated as essential in every context.

Transfer exercise: flexible use of rules

A rule can become a familiar object in thought. For instance, a student may memorize that dividing by a fraction means multiplying by its reciprocal, but may not understand what quantity the calculation represents. When a word problem changes, the operation is applied without interpreting the units.

Suppose three quarters of a litre of juice are divided into servings of one eighth of a litre each. The number of servings is (3/4)÷(1/8)=(3/4)×8=6. The units support the interpretation: litres divided by litres per serving gives servings.

Now change the portion to one sixth of a litre. The same three quarters of a litre give (3/4)÷(1/6)=(3/4)×6=4.5 servings. If only complete portions count, four full portions can be made with some juice left. The arithmetic result and the practical decision are related but not identical.

This example is not a laboratory measure of fixedness. It shows why learners should retain flexible representations of the quantity and the question instead of treating the first memorized procedure as the whole problem.

Failure patterns and targeted repair

Failure one is saying ‘think outside the box’ without identifying a specific unnecessary assumption. Repair it by naming the outcome and resource properties. Failure two is celebrating novelty without testing safety. Repair it with a declared feasibility check. Failure three is assuming every learner who is stuck lacks creativity. Diagnose reading, prerequisites and representation separately.

Failure four is showing the novel answer and then treating immediate imitation as independent insight. Use changed-context practice after a delay. Failure five is encouraging use of resources outside the user’s actual permissions. Preserve ethical and technical boundaries. Failure six is assuming customary practice must be wrong because an alternative exists. Compare results instead.

Failure seven is claiming developmental superiority from one experimental age comparison. Respect the studied conditions and sampling limitations. Failure eight is keeping only successful unusual ideas in a report, hiding those that failed validation. Keep enough negative evidence to understand the search.

The consistent repair is to distinguish generating a possibility, checking the possibility and verifying transfer. Each stage needs an explicit question and honest evidence.

Limitations, FAQs and sources

Is functional fixedness a personality defect? No. It describes a particular restriction in considered object uses under particular circumstances. Does experience always make it worse? No. Expertise can provide essential domain knowledge and may also make some conventional roles more salient. The outcome depends on task and context.

Can a student practise escaping it? Yes, through safe alternate-use tasks, property descriptions, multiple valid methods and independent changed-context tests. But the activity does not guarantee any specific examination result. Is this the same as divergent thinking? Not exactly: producing many ideas differs from identifying when a familiar function inappropriately narrows the search.

Primary source links: Duncker, On problem-solving (1945), German and Defeyter, Immunity to functional fixedness in young children (2000), and the author-hosted research article. These establish a mechanism and particular experimental findings, not a universal classroom success rate.

For ongoing learning, begin with How Intelligence Works and education that transfers into life. If the barrier is an optimistic schedule rather than object use, the planning fallacy is a different, more appropriate mechanism.

All paper exercises, participant counts, software incidents and numerical examples here are illustrations, not independently observed experimental outcomes. An intelligent alternative is one that meets the real requirement and survives verification when conditions change.

How Intelligence Works | Functional Fixedness — Seeing New Uses, Escaping Familiar Tools and Testing Solutions

Contents