Context Effect: Why the Same Thing Reads Two Ways

Recognition is a guess made from two sources at once: the marks on the page and everything that makes some readings likely. Alongside letters, a hurried upright with two curves is almost certainly a B, because the alphabet is what that row contains. Between 12 and 14 the same marks are a 13. Neither reading is a mistake; both fit the company the figure keeps.

That is why a context effect is usually a feature rather than a fault. Handwriting is ambiguous, speech in a noisy room is worse, and objects arrive half hidden. Stephen Palmer showed in 1975 that a briefly presented object was identified more accurately when a scene shown beforehand fit it, which is the same machinery paying off. The cost arrives only when the surroundings point the wrong way: a dose misread on a chart because of the line above it, or a price read as a bargain because of what sat next to it.

Memory works on the same principle, with the surroundings acting as a retrieval cue. Material learned in one place comes back more readily in that place, because the room is encoded along with the words. The classic demonstration is Duncan Godden and Alan Baddeley's 1975 diving study, and the honest summary of the literature since is that the effect is real but modest: Steven Smith and Edward Vela's 2001 review and meta-analysis put the effect at about a quarter of a standard deviation whatever the test: d = 0.29 for free recall, 0.27 for recognition and 0.25 for cued recall, with no reliable difference between them. It shrank where the test itself supplied strong cues, and the belief that recognition is immune comes from a few tightly controlled word-list experiments, not from the pooled data.

One caution belongs here. When Jaap Murre ran a direct replication of the diving study in 2021 with 16 divers, the same-environment advantage did not come out as Godden and Baddeley's had. A 2022 addendum corrected the analysis to a repeated-measures test with an order covariate; the interaction then reached significance, p = 0.017, but not in the pattern a replication needs. His divers used a heated indoor pool and did all four conditions in about 90 minutes, against 7 meters of open water over four days in the original, so the gap may be the size of the context change. Either way, that study is less secure than its fame suggests.

The phrase has a second life in marketing, where context effects are the ways the rest of a choice set moves a decision; those are covered here as the decoy effect and the contrast effect.

Jerome Bruner and Leigh Minturn published "Perceptual identification and perceptual organization" in The Journal of General Psychology in 1955. Their stimulus was a single ambiguous figure that could be read as the letter B or the number 13, drawn with a small gap between the upright and the curves.

Different groups of participants saw it in different company. One group was shown a series of letters, the other a series of numbers, and the figure was then flashed briefly before they reported what they had seen and drew it. The reports followed the series: the letter group identified it as a B and drew a closed B, the number group identified it as a 13 and drew the two parts separately. The drawings mattered more than the labels: they showed context shaping what was recorded, not only what was said. The demo above puts the same manipulation into words, with a described mark instead of a flashed figure.

The memory half of this page has its own classic. Godden and Baddeley, in the British Journal of Psychology in 1975, had divers learn spoken lists of words either on dry land or underwater at about 7 meters, then recall them four minutes later either in the same environment or in the other one. Every diver did all four combinations, spread over four days of open-water diving near Oban in Scotland. Recall was better when the environment at recall matched the environment at learning, and a second experiment ruled out the simple explanation that moving between environments was itself disruptive.

That result became the standard illustration of context-dependent memory and stood for 46 years with no published replication attempt; Murre's 2021 attempt, even after its 2022 reanalysis, did not reproduce the pattern. Smith and Vela's pooled evidence still supports a small effect on recall and recognition alike.

Context works silently, which is why it is noticed mostly when it misfires. The rows below are illustrations, not findings from the studies cited on this page.

The aim is not to switch context off, which would make recognition much worse, but to notice when the surroundings are deciding.

• When identification matters, get a second context: read the ambiguous number next to a different line, or ask the writer. A figure that changes reading between settings was never legible.
• When learning something you will be tested on elsewhere, vary where you study, so the cue spreads across several settings rather than one.
• When a thought goes missing, reinstate the context: go back to the room, picture it, or retrace the sentence you were reading.
• When judging a price, a candidate or an offer, write down what the comparison set is. Change the set and the same option reads differently: the consumer-choice version of the phenomenon.
• When proofreading, break the context: a different font, a printout, reading aloud, or starting from the bottom.
• Keep the size in proportion. The environmental memory version is modest, between d = 0.25 and 0.29 depending on the test in Smith and Vela's meta-analysis, and the one direct replication of the diving study did not reproduce it, so matching a room is worth little next to practicing retrieval.

Several effects involve the surroundings of a judgment; they differ in what the surroundings do.

• It is not the contrast effect, where a judgment is pushed away from whatever was judged just before it. Contrast moves a rating along a scale; a context effect changes what the thing is taken to be.
• It is not the decoy effect, where adding an unattractive third option changes the choice between the other two. That is the consumer-research sense of context, and it works on a choice set rather than on perception.
• It is not cue-dependent forgetting, which is the failure to retrieve something because the cue is missing. Context-dependent memory is the mirror image: the surroundings are one of the cues, so matching them helps a little and losing them costs a little.
• It is not expectation bias, where a researcher's or observer's hope for a result shapes what gets recorded. Context effects need no stake in the outcome; letters and house numbers have no agenda.