No. 045
This optional formatting bolds the leading part of each word to give your eye a focus point; some readers find it helps them stay locked in.
Don Norman, The Psychology of Everyday Things. First edition, Basic Books, 1988. The work was subsequently retitled The Design of Everyday Things, the title used by this library. This guide preserves the first edition's seven chapters and distinguishes them from Norman's substantially revised 2013 edition.
When a person repeatedly pushes a door that must be pulled, enters the wrong setting on a control panel, or cannot tell whether a device accepted a command, the failure is often blamed on carelessness. Norman redirects attention toward design. Objects should communicate what actions are possible, map controls to results, constrain invalid action, and provide timely feedback.
This shift matters beyond household products. Software, medical equipment, transportation, forms, workplaces, and public systems all distribute knowledge between people and environments. A design can support memory and understanding, or demand recall while hiding state. It can make an error reversible, or let one slip cause disproportionate harm.
The book belongs in a lifetime canon because it teaches a humane diagnostic habit: before blaming the user, ask what the system made visible, expected, and recoverable. It also gives a vocabulary for examining ordinary frustration. Its examples are sometimes technologically dated, but the underlying questions remain central.
Donald A. Norman is an American cognitive scientist and design researcher. His training and early work involved psychology, cognition, and human information processing. He held academic roles at institutions including the University of California, San Diego, and later worked in technology industry settings and design education.
Norman helped develop cognitive science and human-centered design as fields connecting psychology with artifacts. His industry experience later included Apple, where the term “user experience” gained prominence. That later career should not be silently read into the 1988 edition, but it demonstrates the continuing development of the book's ideas.
The 1988 text grew from cognitive research and Norman's attention to recurring failures in everyday objects. It originally appeared under The Psychology of Everyday Things, emphasizing how people understand action. The retitled work became widely known as The Design of Everyday Things. The revised 2013 edition updates examples and terminology, including a more precise distinction between affordances and signifiers. This guide names those updates as later clarification rather than attributing them verbatim to 1988.
Good design lets people discover appropriate actions, understand system state, predict consequences, recover from error, and rely on knowledge embedded in the world instead of carrying unnecessary complexity in memory.
The book moves from visible design failures to a psychology of action, then to memory, discoverability, error, complex design tradeoffs, and user-centered practice. Its seven chapters form a diagnostic sequence. First identify the problem. Then explain how action and interpretation work. Next decide where knowledge should reside. Then make action discoverable. Anticipate error. Understand why design becomes difficult. Finally, test design with real people.
Norman's central concepts include affordances, constraints, mappings, conceptual models, visibility, feedback, and the gulfs of execution and evaluation. An affordance is a possible relationship between an actor and an object, such as a surface that supports grasping. In the 1988 text, perceptibility is often discussed together with affordance. Later Norman uses “signifier” for the perceivable cue that communicates where or how to act.
Constraints narrow possible actions. Physical constraints prevent certain movements. Semantic constraints rely on meaning. Cultural constraints rely on conventions. Logical constraints use relationships among components and remaining possibilities. Good constraints guide without requiring extensive instruction.
Mapping is the relationship between controls and effects. A natural mapping uses spatial or conceptual correspondence, such as stove controls arranged like burners. A conceptual model is the user's explanation of how a system works. The designer has a model, the user develops one, and the system image, everything the product communicates, connects them.
Feedback reveals what action occurred and what state now exists. Without timely feedback, a user may repeat commands or construct a false explanation. Visibility allows relevant state and action possibilities to be found.
The gulf of execution is the distance between a person's goal and available actions. The gulf of evaluation is the distance between system state and the person's ability to perceive and interpret it. Design improves when both gulfs become easier to cross.
Norman opens with doors, telephones, appliances, and other objects that make capable people feel incompetent. A well-designed door communicates whether to push, pull, slide, or turn. A flat plate suggests pushing; a graspable handle suggests pulling. When hardware communicates the wrong action, a label may treat the symptom while leaving the contradiction.
The chapter introduces affordances, constraints, mapping, conceptual models, and feedback. It also presents the paradox of technology. New functions can make life easier, yet added controls and modes increase complexity. Designers may understand an object from drawings and specifications while users encounter only the system image.
Norman argues that people blame themselves because error is individualized. Yet repeated mistakes across users are evidence about design. The principle is not that users are never responsible. It is that predictable error should be treated as a system property.
The refrigerator temperature-control example shows a false conceptual model. Two controls may appear to regulate separate compartments while actually interacting through a shared cooling mechanism. Labels do not reveal the causal structure. The user adjusts one control, observes a delayed result, and forms an inaccurate theory.
The condensed principle is to make the right action perceptible, the relation between action and result understandable, and the result visible.
Norman describes action as a cycle. A person forms a goal, forms an intention, specifies an action sequence, executes it, perceives the system state, interprets that state, and evaluates the result against the goal. The stages are analytical, not a claim that conscious thought always follows seven explicit steps.
The gulf of execution appears when a person knows the desired outcome but cannot determine what to do. The gulf of evaluation appears when an action has occurred but the person cannot tell what changed or whether the goal was reached.
Design can support each stage. Visible controls help specify action. Natural mapping helps selection. Physical ease supports execution. Feedback supports perception. Clear state supports interpretation. A conceptual model supports evaluation.
The chapter also examines explanations and blame. People build causal stories from temporal coincidence, incomplete evidence, and prior expectations. Learned helplessness can arise when repeated difficulty teaches a person that effort does not matter. Taught helplessness occurs when systems repeatedly blame users for design failures.
Norman's action cycle provides an evaluation tool. When a task fails, identify the stage: Was the goal unclear? Was the action unavailable? Did execution fail? Was feedback missing? Did the user misinterpret state? Fixing the wrong stage can add complexity.
People often perform accurately without memorizing complete instructions. Knowledge can reside in memory, physical arrangement, labels, conventions, constraints, or the partially completed task. The environment acts as an external memory.
Norman distinguishes knowledge of and knowledge how. Declarative knowledge can be stated, while procedural skill is often difficult to verbalize. He also discusses arbitrary, meaningful, and explanatory mappings. Arbitrary associations demand memorization. Meaningful relationships are easier to learn. Explanatory models support prediction beyond practiced cases.
Memory for precise detail is frequently weaker than confidence suggests. People recognize coins, keyboards, and familiar symbols without being able to reproduce every feature. Exact memory is unnecessary when the world supplies the detail at the moment of use.
Reminders combine a signal that something must be done with information about what the task is. A reminder that rings without identifying the obligation may create anxiety rather than action. Good prospective-memory support appears at the right place and time.
The tradeoff is not always “put everything in the world.” External labels can clutter, disappear, expose private information, or slow expert performance. Knowledge in the head enables speed and action when cues are unavailable. Design should place stable, arbitrary, or infrequently used knowledge in the world while supporting learned skill for frequent tasks.
This chapter explains discoverability through constraints and mapping. Physical constraints make impossible actions literally impossible. A plug that fits only one orientation is a physical constraint, though poor constraint design can create frustration or force.
Semantic constraints use knowledge of the situation. A rider belongs facing forward on a motorcycle model. Cultural constraints rely on learned conventions, such as reading direction or standard interface symbols. Logical constraints arise when components and available locations imply a unique relation.
Natural mapping reduces translation. Controls placed in the same spatial pattern as controlled objects communicate correspondence. Arbitrary rows of stove knobs require labels and memory, and can produce selection errors.
Visibility must be selective. Exposing every internal function can overwhelm. The goal is to make relevant actions and state visible at the appropriate time. Progressive disclosure can reveal complexity as needed, though the term is more common in later interface design.
Standardization can rescue a domain when natural mapping is impossible. Conventions reduce learning across products, but an arbitrary standard must be taught and consistently applied. Once established, changing it carries costs. The principle is to prefer natural relationships, then consistent standards, and avoid unnecessary arbitrary mappings.
Sound can provide feedback when visual attention is elsewhere, but alarms can annoy, startle, or become indistinguishable. A useful sound communicates urgency and source while allowing action. Modern accessibility requires redundant channels so information is not available only through sight, hearing, color, or fine motor action.
Norman distinguishes slips from mistakes. A slip occurs when the goal is appropriate but execution goes wrong. A mistake occurs when the goal, plan, or interpretation is wrong. Different causes need different remedies.
Common slips include capture errors, where a familiar sequence takes over; description errors, where an action is performed on a similar target; data-driven errors, where incoming information intrudes; associative activation, where one thought triggers an unintended action; loss-of-activation errors, where the goal is forgotten; and mode errors, where the same control behaves differently in an unnoticed state.
Mistakes arise from incomplete or inaccurate conceptual models, overgeneralized rules, and interpretation of ambiguous evidence. Training may help a knowledge gap, but design should still reveal state, warn about consequences, and support recovery.
Forcing functions are constraints that prevent moving forward until a requirement is met. Interlocks prevent incompatible states. Lock-ins keep an operation active until completion, while lockouts prevent entry into a dangerous state. These mechanisms can protect users but must account for emergency escape and failure modes.
Error messages often arrive too late and describe the system's complaint rather than the user's recovery path. Better design treats error as normal, preserves work, makes reversal easy, confirms destructive action in proportion to risk, and explains how to proceed.
Mode errors deserve special attention. If a control's meaning changes, the active mode must be visible and distinguishable. Eliminating unnecessary modes is better than adding another warning.
The principle is to design for human variability, interruption, habit, and imperfect attention. Error tolerance is more humane and often more reliable than exhortation.
Design is difficult because products serve multiple people and goals. Manufacturers consider cost, reliability, production, repair, sales, appearance, regulation, and schedules. Users vary in experience, ability, culture, context, and frequency of use. No single optimization resolves every tradeoff.
Feature accumulation creates the “creeping featurism” problem. Each added function may be individually defensible while the total interface becomes incoherent. A separate control for every function creates clutter; modes reduce controls but hide state. Designers must organize complexity rather than merely add capability.
The designer's conceptual model must reach the user through the system image. Marketing, manuals, labels, controls, behavior, and feedback all contribute. A good internal architecture cannot compensate for an incomprehensible surface.
Evolutionary design can improve products through repeated small changes, but local modifications can preserve a bad underlying concept. Radical redesign can improve the model but introduces unfamiliarity and risk. Designers need both iteration and willingness to reconsider structure.
Designers are not typical users. Familiarity makes labels seem obvious, memory demands seem easy, and edge cases seem rare. Economic buyers may also differ from operators, as in workplace software or medical devices. Research must include the people who perform the task and bear the consequences.
The condensed principle is that complexity belongs somewhere. A humane design asks the system and designer to carry as much as practical rather than transferring it invisibly to the user.
The final chapter turns concepts into method. Begin with users' tasks, goals, and environments. Make possible actions discoverable. Use natural mappings and constraints. Provide feedback. Design for error. Standardize when necessary. Test with actual users and revise.
Observation matters because people may not accurately report automatic behavior. Watch where they hesitate, form false models, recover, ask for help, or invent workarounds. A workaround is evidence about unmet needs, though it may also introduce risk.
Instructions should not compensate for avoidable confusion. Manuals remain necessary for complex systems, but frequently needed information should be available at use. Labels should clarify rather than contradict physical cues.
The chapter advocates simplification while recognizing that tasks can be inherently complex. Simplification may mean reorganizing structure, automating a reliable subtask, making state visible, or distributing knowledge across person and world. Automation must provide understandable handoff and recovery when it reaches its limits.
Testing should include first use, repeated use, interruption, error recovery, and atypical conditions. Success is not that a participant eventually completes a task after coaching. It is that representative users can discover, execute, understand, and recover within defined criteria.
The first idea is the system image. Users do not receive the designer's intention directly. They infer operation from what the product shows and does. Documentation, interface, physical form, and feedback must communicate one coherent model.
The second is the action cycle. Design supports the movement from goal to action and from result to understanding. Execution and evaluation are separate gulfs, so a visible control without feedback is incomplete.
The third is distributed knowledge. Memory is not the only place cognition occurs. Arrangement, labels, history, constraints, and visible state can make action reliable.
The fourth is error as design information. Slips and mistakes have recognizable forms. Recovery, reversibility, state visibility, and constraints should be planned before failure.
The fifth is mapping. Relationships between controls and outcomes should use spatial, semantic, or cultural logic. Where arbitrary standards are unavoidable, consistency becomes essential.
The sixth is iteration with real users. Expertise in the product can blind designers to novice experience. Observation exposes the gap.
The concepts should be used together. Suppose a person changes the wrong stove burner. The affordance of turning a knob is clear, but the mapping between knob and burner is poor. Feedback may reveal heat only after delay. Similar controls invite a description slip. A label adds knowledge in the world but still requires visual search. A spatial control arrangement may solve the error more directly.
Consider a software upload that appears frozen. The action is available, but feedback and system state are absent, widening the gulf of evaluation. The user clicks again and creates duplicates. A progress indicator, disabled repeated action, visible queue, and reversible cancellation address different parts of the cycle.
Consider a hospital infusion pump with modes. Training cannot be the only defense because interruption and fatigue are predictable. Mode state, dosage units, confirmation, constraints, independent checks, and recovery all matter. High-stakes systems require current human-factors standards and professional validation beyond this book.
The book's greatest strength is its refusal to treat confusion as a character flaw. It provides concepts that lead directly to inspection and redesign. Its door and stove examples remain useful because the causal relationship between cue, model, action, and feedback is visible.
The 1988 affordance terminology sometimes merges possible action with perceived information about action. Later Norman distinguishes affordances from signifiers more sharply. James Gibson's ecological psychology also gives affordance a specific relational meaning. Readers should identify which usage is active rather than treating terminology as timeless.
The book's examples privilege physical controls and relatively bounded individual tasks. Modern services are distributed across devices, organizations, policies, and people. A confusing benefits application or content-moderation appeal may have no single object to redesign. Service design, organizational incentives, accessibility, and governance extend the analysis.
User-centered design can become too narrow if “user” means the purchaser or frequent customer while excluding workers, nonusers, bystanders, communities, and people subject to environmental or privacy effects. A product can be easy for its direct user and harmful to others.
Ease is not the only value. Security may require friction. Learning may require productive effort. Consent may require interruption. Safety may require confirmation. The correct goal is appropriate effort aligned with user purpose and risk, not the removal of every obstacle.
Observation does not automatically reveal need. Researchers interpret behavior through assumptions and power. Participatory design, accessibility expertise, privacy safeguards, and representative recruitment reduce but do not eliminate this limitation.
Finally, familiar design can normalize bad conventions. Consistency helps present action but can slow adoption of a safer model. Migration design must account for both immediate learning and long-term benefit.
Cialdini's Influence and Norman's book both study cues. A signifier should truthfully communicate action; a persuasion cue should truthfully communicate value. A fake button, fake review, or resetting timer exploits learned interpretation.
Kahneman helps explain why visible defaults and familiar mappings matter under cognitive load. Gawande extends error-tolerant design into team communication and critical checks. Grove's production system adds constraints, indicators, and quality placement, while Norman emphasizes the operator's conceptual model.
Marquet connects through authority near information. A worker who sees a design hazard needs both a channel and standing to stop action. Design is therefore partly organizational.
Run a door test. Select five doors in ordinary use. Before observing others, record the apparent action, signifier, physical affordance, constraint, mapping, and feedback. Observe ten uses without identifying individuals. Count hesitation, wrong attempts, labels consulted, and recovery. Propose the smallest physical change that addresses the observed error.
Use the seven-stage action cycle on one failed task. Write goal, intention, action specification, execution, perceived state, interpretation, and evaluation. Mark the first stage where information or capability failed. Change that stage and retest with a representative user.
Conduct a knowledge-location audit. List every fact a person must remember. Move stable and arbitrary facts into visible, timely cues where privacy and clutter permit. Preserve expert shortcuts. Measure completion time, errors, assistance requests, and confidence before and after.
Perform an error analysis. Classify five recent failures as slips, mistakes, or uncertain. For slips, examine targets, modes, interruption, and feedback. For mistakes, examine the conceptual model and rules. Make destructive actions reversible where feasible. Do not use this exercise to assign clinical or legal blame.
Test accessibility with qualified standards and affected users. Check keyboard operation, screen-reader names, contrast, text size, captions, touch targets, language, motion, and error recovery. Automated checks are useful but insufficient.
For a high-stakes product, involve domain professionals and current standards. Norman's concepts support questions but do not certify medical, aviation, industrial, or safety-critical equipment.
Close the guide and draw the seven-stage action cycle. Label the gulf of execution on the action side and the gulf of evaluation on the interpretation side. Then define affordance, signifier, constraint, mapping, conceptual model, system image, feedback, slip, and mistake.
Active-retrieval questions: Why do Norman doors fail? What is the system image? Where can knowledge reside? What four constraint types appear? What makes mapping natural? How do slips differ from mistakes? What is a mode error? Why can a forcing function help or harm? What creates creeping featurism? Why must designers observe users?
Application questions: Which recurring error in your environment is predictable? What state is hidden? Which memory demand can move into the world? Where is friction protective rather than harmful?
Comparison questions: How do Norman and Cialdini differ on cues? How would Gawande handle a critical action that cannot be designed away? How would Grove measure the output of a redesign?
Review after one day, three days, one week, two weeks, one month, three months, and six months. Retrieve concepts before rereading. At one week perform an action-cycle diagnosis. At two weeks observe a real task. At one month retest a change. At three and six months review whether the improvement persisted across users and contexts.
Teaching exercise: give another person a confusing object or interface. Ask them to think aloud while attempting a task. Explain the failure using only visible evidence, then propose one change and one metric. Do not coach during the initial attempt.
Thesis: Design should bridge the gap between what people want and what systems allow by making action, state, consequence, and recovery understandable.
Five ideas: users infer models from the system image, execution and evaluation are separate gulfs, knowledge can reside in the world, natural mapping reduces translation, and predictable error should be designed for.
Three applications: diagnose a task with the action cycle, move appropriate memory demands into the environment, and observe representative users before and after a change.
Strongest limitation: the original artifact-centered examples require extension to services, institutions, accessibility, privacy, and harms affecting people other than the direct user.
Ten final questions: What was the first title? What is an affordance? What is a signifier? What is a conceptual model? What are the seven action stages? What constraint types exist? How does a slip differ from a mistake? What is a forcing function? Why does feature accumulation hurt? What makes design user-centered?
The book's humane achievement is to make frustration investigable. Once an error is understood as an interaction among goals, cues, mappings, state, and recovery, blame can give way to design.
The narration identifies the 1988 title once, then uses the library title. “Affordance” is pronounced “uh-FORD-əns.” “Don Norman” is spoken naturally. Citations, raw URLs, production instructions, edition apparatus, and Source Notes are excluded from narration.
For revised technical classics, create an edition firewall. Preserve original chapter structure and period examples while labeling later terminology, research, and standards. Do not silently modernize quotations or attribute a revised concept to the first edition.
Every design application should specify user, task, environment, error class, intervention, result measure, and balance measure. Include nonusers and affected parties when design creates external effects.
For safety-critical and accessibility work, the guide must point to current professional standards and testing. A historical design book supplies analytical concepts, not certification or a complete compliance method.
Retest after realistic interruption, fatigue, time pressure, and partial failure. A design that succeeds only during a calm demonstration has not yet shown dependable everyday usability.
The base text is Donald A. Norman, The Psychology of Everyday Things, first edition, Basic Books, 1988. Title history, publication data, and the seven-chapter structure were checked against Basic Books, Library of Congress, and WorldCat bibliographic records. The library filename uses the later title The Design of Everyday Things.
Later terminology was checked against Don Norman, The Design of Everyday Things: Revised and Expanded Edition, Basic Books, 2013, and Norman's published clarification “The Way I See It: Signifiers, Not Affordances,” Interactions, 2008. James J. Gibson's The Ecological Approach to Visual Perception, 1979, provides the relevant earlier affordance context.
Biographical and field context was checked against Norman's University of California, San Diego profile, ACM professional biography, and Norman's published history of user experience. Current accessibility cautions were checked against the Web Content Accessibility Guidelines issued by the World Wide Web Consortium. Safety-critical cautions reflect the need for current domain-specific human-factors standards rather than reliance on the 1988 text alone.
Paste any of these into an AI assistant to keep exploring this book.
Explain Don Norman's gulf of execution and gulf of evaluation, along with the idea of a system image, using two or three concrete modern examples from apps, appliances, or public kiosks where people get blamed for a confusion the design actually caused.
Norman argues that predictable user error should be treated as a design property rather than a personal failing. Steelman the objection that this framing can go too far and remove healthy friction that protects against real harm, like a confirmation step before an irreversible action, and tell me where the line between helpful simplicity and dangerous ease should sit.
Help me pick five doors, switches, or apps I use often, predict what each one's signifiers suggest I should do before I test it, then design a small observation of a few real people using them and compare their actions to my prediction.
Connect The Design of Everyday Things to Robert Cialdini's Influence, and explain the difference between a signifier that truthfully shows what an object does and a persuasion cue that manipulates what someone believes about value.
Norman separates a slip, where the goal was right but execution failed, from a mistake, where the underlying model was wrong. Help me walk through one recent error I made and diagnose honestly which of the two it actually was, and what specific change would prevent it next time.