Word association test with no scoring key

There is not one, and that is a design decision rather than an unfinished feature. A score needs a key, and free association has no key: whatever word arrives first is the correct answer to the only question being asked. The two figures at the end — how quickly words arrived and whether repeated cues brought back the same word — are measurements of the run rather than judgments of it, and neither is combined into anything. If a site hands you a score for this task, ask what it compared your words against; there are only two possible answers, a licensed dataset of what other people said, or nothing.

It uses the timing idea and none of the interpretation. Jung's version, a century ago, read a list aloud, recorded response times with a stopwatch, and treated a long pause, a repeated word or a failure to answer as a marker of an emotionally loaded area — a complex — to be explored afterwards. The timing survived into ordinary experimental practice, because how long a word takes to arrive is a real and reproducible quantity. The interpretation did not survive: the inference from a slow response to a hidden conflict never established the kind of reliability that would let anybody act on it, and this page makes no inference of that kind at any point.

Because everything between the first letter and Enter is typing, not association. A person who thinks of feather and types it in 900 ms and a person who thinks of feather and types it in 2,400 ms have had the same association, and timing to Enter would put a full second and a half between them for reasons that belong to their keyboards. Stopping at the first printable keystroke also makes the measurement fair to a phone, where an on-screen keyboard slows every word by a constant that has nothing to do with the cue. The cost is that a word thought of and then reconsidered is timed at the first thought, which is the right end of that trade.

Because a cue list chosen for emotional weight is a different instrument with a different purpose, and it is the one this site stays out of. Words like mother, failure or blood produce responses that people then want interpreted, and interpretation of a single response is precisely the step that has no evidence behind it. Concrete, common, unloaded nouns give the same measurable quantities — a latency and a word — with nothing in them that invites a reading. There is a second, duller reason: a flat cue list produces latencies that vary with the cue's frequency and imageability rather than with whatever the reader brought to the page, which makes the median a more stable number.

Only with a dataset this page does not have, and the honest thing is to say so rather than to approximate it. There is no norm to score a person against here, because free association has no correct answer. What exists instead is a different kind of reference: datasets recording how often each response was given to each cue. The best known of those collections is Nelson, McEvoy & Schreiber (2004), The University of South Florida free association, rhyme, and word fragment norms, built from several thousand respondents across several thousand cues. Carrying one would change this page from a record into a comparison, and it would need the licensing and the disk space to do it. Inventing a typicality figure without one would be the worst option of the three, because it is the number a visitor would believe.

Most often because the first showing changed what was available. Producing a word makes its neighbors easier to reach for a while and the cue itself slightly less able to reach anything else, and the eight or more cues in between have each left their own residue — so the second showing happens in a different mental context from the first, by construction. A different word is therefore not instability and not an error; the count exists to show how strongly a cue dominates everything else that has happened since, which is a real property and varies enormously between cues. Two matches out of four is an ordinary result.

Nothing this page can responsibly claim, and nothing it will guess at. The word a cue produces depends heavily on what you read this morning, what you did yesterday and which sense of the cue you happened to reach first — all of which change week to week — and the associations that are stable across a population are stable because of the language rather than because of the people speaking it. That leaves very little room for a reading of an individual list, which is why this page returns the list instead of a reading. What the list is genuinely good for is looking at yourself: it is a snapshot of what twenty ordinary words pulled up on one particular afternoon.