The research and applied-psychology terms I use in the courses and in client work. Where a concept is contested or routinely misapplied, the definition says so rather than flattening it into a tidy rule.
Looking for technical SEO, performance or accessibility terms? Technical SEO glossary
Comparing two or more versions of an element with real traffic to see which performs better on a metric defined in advance. It needs volume to reach statistical significance. Its limits are the most often forgotten: it tells you what works, not why; it only optimises within the current paradigm; and the statistical winner may not be the experience winner.
The bias by which a pleasing interface is perceived as easier to use than it is. It contaminates your research on two fronts: it acts on you as the auditor and on participants during a test. If the screen is easy on the eye, people forgive real problems and assume errors are their own fault, so they never even mention them to you.
The collaborative, visual version of thematic analysis: each observation goes on a sticky note and the team groups them by similarity, letting categories emerge from the wall. It distributes bias, because a lone analyst sees what they expect to see. What does not work is one person building the wall alone and then presenting it: that is a report with sticky notes.
The first piece of information we receive disproportionately influences later judgements. The initial price you show, even struck through, anchors the perceived value of the whole offer. As a researcher this forces you to watch the order of what you present: the first question in an interview anchors every answer that follows it.
A by-product of selective attention: we learned to ignore a page's peripheral areas, especially the right column, because historically that is where ads lived rather than content. It is learned behaviour and fairly predictive. The nuance usually lost is that position is not always the culprit: sometimes a useful element simply looks so ad-like that the brain discards it wherever it sits.
An information architecture technique where participants group content cards in whatever way makes sense to them, to uncover navigation structure and labelling. It can be open, when they create the categories; closed, when they sort into given ones; or hybrid. It tells you how people would group things, not whether the resulting structure works: tree testing answers that.
The idea that too many options paralyse rather than liberate. It is overstated and worth saying so: Iyengar and Lepper's famous jam study does not always replicate, and Scheibehenne and colleagues' meta-analysis found an average effect close to zero. It depends on the domain, on whether the person knows what they want, and on how the options are organised. Do not cut on dogma.
Merging several loose items into a single unit so they take up less room in working memory. A card number reads better in blocks of four than as sixteen digits in a row. Its limit: the chunk has to match what the user already knows. What is one unit to an expert is a hieroglyph to a novice.
The systematic error that results from applying a heuristic where it does not fit. Systematic means it repeats in the same direction, so it is predictable and you can design accounting for it. The other side matters just as much: biases operate on you while you research, not only on the user you are observing.
The mental effort you demand of a person to operate your interface. It rises when you force them to recall data from earlier steps, decode ambiguous groupings, or compare too many options. You do not lower it by reflexively stripping things out: you lower it by removing what does not serve the decision the person is making on that screen, right now.
The deceptive pattern that shames you for declining an offer: the decline button reads something like no, I would rather pay more. It is among the easiest to spot, which makes it a useful calibration case. When a pattern leaves you unsure, compare it against this one: a pre-ticked newsletter opt-in is far more arguable than obvious confirmshaming.
Observing the user where the action actually happens, not in a lab room. Beyer and Holtzblatt organise it into four principles, among them context and partnership under the master-apprentice model. It works because memory is fallible and people rationalise what they do: nobody mentions the password sticky notes on their monitor over a video call.
A design that uses what it knows about your head against you: it hides the exit, shames you for choosing well, charges you for something you did not ask for. Same psychological mechanism as a nudge, different beneficiary. Brignull now prefers the term deceptive patterns. The line has a real grey zone: mark it grey rather than forcing a verdict.
An on-screen reading pattern that Nielsen Norman Group identified in 2006: a horizontal sweep across the top, then a vertical run down the left side. It describes how people scan when text is poorly structured, not how they ought to read. Treat it as a symptom of flat content, not as a design template or a law of the gaze.
The time to reach a target depends on its size and how far away it is: bigger and closer means faster and with fewer errors. It was formulated for pointing tasks with a device, so on touch gestures the maths gets complicated. And beware the easy reading: a huge delete-account button is a Fitts triumph and an experience disaster.
Rules by which the brain imposes structure through grouping. The two most used in interfaces are proximity, where nearby objects are perceived as one unit, and similarity, where we group items sharing shape, colour or size. They sometimes compete with each other, and no universal rule says which one wins. Group elements with no functional relationship and you produce ambiguity, not neutrality.
The distance between what the system did and what the person manages to understand of it: did it work, what happened? You close it with good feedback. Like its execution counterpart, it is a frame for locating whether the problem sits in the action or in the interpretation, not a measure of how wide the gap is or a recipe for closing it.
The distance between what a person wants to do and what they know how to do in your system: I want to do this, how is it done here? You close it with good signifiers and good mapping between control and effect. Its limit: the frame locates the problem, not its size or its fix. It names the place; the size still needs researching.
A visualisation of where users click, how far down the page they scroll, or where they move the cursor. Warm colours mark more activity. It is a good detector of ignored zones, but it explains nothing: it shows aggregated behaviour stripped of the intent behind it. Use it to choose where to research, never as a report's conclusion.
A mental shortcut or rule of thumb people use to simplify decisions. They increase efficiency, which is why they exist, but they also produce systematic errors, which are cognitive biases. Do not confuse this sense with Nielsen's heuristics, which are interface evaluation criteria rather than shortcuts happening in the user's mind.
An expert review of an interface against a set of principles, usually Nielsen's ten heuristics, with a severity scale for prioritising the findings. It is fast and cheap, and that is its trap: it finds potential problems by expert judgement, not problems confirmed with users. It complements usability testing rather than replacing it.
The more simple, evenly matched alternatives you present, the longer a person takes to choose. Its limit is large and usually ignored: the law assumes equiprobable options, and a real interface has categories, hierarchy and familiarity that break that maths. Less is always more is a misreading. Sometimes people want to see everything, and cutting for its own sake frustrates as much as overwhelming.
A guided one-to-one conversation for understanding why: motivations, context, the decisions behind a behaviour. It is the method of why, not of how many: with small samples it generates hypotheses rather than validating them. Its known blind spot is that people report what they believe they did, not what they did, so pair it with observed behaviour.
People spend most of their time on other products, so they expect yours to work in much the same way. It is an inherited expectation with social roots: you learned it from collective behaviour, you did not deduce it alone. Its limit: following convention to the letter kills differentiation, and conventions change over time. Use it to pick your battles, not to copy.
A framework that shifts your focus from who the user is, the demographics, to what they want to accomplish. Levitt's formulation: people do not want a quarter-inch drill, they want a quarter-inch hole. Once you understand the real job the person needs done, you stop validating features and start validating solutions.
A representation of a person's experience across time and touchpoints, showing what they do, think and feel at each stage. It locates where the experience breaks between teams that only see their own segment. Build it from assumptions rather than research and you get a handsome diagram of your own org chart.
The representation a person builds of how your system works, assembled from what they learned in other products. Problems appear when the system's model does not match theirs: that is where the gulfs of execution and evaluation open up. You cannot decree it in documentation; you only uncover it by researching how people explain what they think happened.
Deliberately combining qualitative and quantitative data, not stacking both just in case. Three patterns are useful: explore then validate, quantify then go deep, or run both in parallel and compare. When the two sources contradict each other, that contradiction is usually the most valuable finding you will get in the whole study.
A push that arranges the decision the person wanted to make anyway: it puts the healthy option at eye level, but the cake stays available and in view. It differs from a deceptive pattern by four questions: who it plays for, whether it preserves freedom of choice, whether it survives being explained out loud, and whether it is reversible without penalty.
You do not remember an experience's average: you remember its most intense moment and how it ended. A hostile ending colours months of good use. Its limit is that it describes memory, which can differ from how the experience was lived moment to moment, and it comes mostly from studies on pain and affect. Its transfer to UX is reasonable, not a certainty.
A user archetype built from research data rather than imagination, summarising a segment's needs, goals and frustrations. Its value lies entirely in the evidence behind it: a persona invented in a meeting room turns assumptions into an official document, and that makes them harder to question later on.
One of Nielsen's ten heuristics: any data the user entered earlier and needs now should be visible, in a summary or a prefilled field. Recognising is cheaper than recalling. Its limit is that showing everything saturates too. Show what is relevant to the current decision, not the entire history of the flow.
The deceptive pattern where you get in easily and getting out is nearly impossible: one-tap sign-up, cancellation buried in a hidden menu, three retention screens and a narrow time window. In Brignull's current taxonomy it was renamed hard to cancel, a more literal label and one easier to defend before a regulator. You will meet both names in the literature.
A classification of UX research methods along three dimensions: data type, qualitative answering why versus quantitative answering how many; approach, what people say versus what they do; and context of use. Use it to pick a method from your objective, not from habit or from whichever tool you already pay for.
The point where new participants stop contributing new codes and only confirm the ones you already have. It is the real criterion behind the five-to-eight-users-per-segment habit: not a magic number, but where the learning curve flattens. If you are still finding new themes at participant eight, your segment is more heterogeneous than you assumed.
The filter the brain uses to avoid collapse: it actively discards whatever it judges irrelevant to its goal. On screen this creates zones the user will not look at, even if you put something important there. Its limit is that the problem is not only about position. A useful element designed with too much colour or visual noise also gets classified as an ad and filtered out.
Narrow your eyes until the screen blurs. Whatever still stands out is what wins the first glance; whatever dissolves does not exist. Defocusing stops you processing fine detail and leaves only masses of contrast and dominant groupings. Its limit: it tells you what stands out, not whether what stands out is right for the user's task.
The tendency to stick with the default option rather than change it. It is neutral in itself: it can serve the user, when the default protects them, or work against them, when it sustains a subscription they no longer want. So the relevant question is not whether the default influences behaviour, but who it favours and whether it can be reversed without penalty.
The two processing modes Kahneman describes. System 1 is automatic, intuitive and emotional, and it dominates most digital interactions: people scan for familiar patterns. System 2 is deliberate and analytical, and it costs effort. When your interface forces System 2 on with no need for it, you produce frustration and abandonment.
The most repeated and worst understood rule in the field: menus should not exceed seven options. Miller's paper is literally titled the magical number seven, plus or minus two, and that margin already shows how far it was stretched afterwards. Cowan revisited it and proposed four. Question the inherited number instead of applying it from memory.
The base method for qualitative data in UX: familiarise yourself with the material, code fragments, group codes into themes, check them against the data, and name them. Coding means assigning descriptive labels, not programming or transcribing. The rule that prevents fragile conclusions: if a theme rests on one quote from one participant, it is not a theme, it is a quote.
Evaluating a navigation structure presented with no visual design: you ask people to find specific items and measure whether they succeed and by what path. It runs after card sorting, to validate the proposed structure before design work starts. Stripping the design isolates architecture problems from interface problems, which is precisely its value.
Watching real people attempt concrete tasks in your product. It can be moderated or unmoderated, in person or remote. The method rule most often forgotten: do not ask whether it felt easy. Measure error rate and time on task, because perceived ease and actual performance do not always agree.
How visual weight is distributed across a screen: which element grabs attention first and which gets pushed to the bottom. Perception assigns that weight before logic does, in milliseconds and without asking you. The practical consequence is that an element standing out does not mean it is the right one. Visual hierarchy should match the user's task hierarchy, and that match is your decision.
What you use to hold information while doing something with it. Think of it as a small table with room for three or four objects in view, not a wide desk. If your screen forces someone to recall data from three steps back, the table saturates and something falls off. That is not user distraction: the table is small for everyone.
A glossary explains the terms; deciding which one applies to your case is a different conversation. If you want to think it through with someone, get in touch.