Von Restorff effect
An item that differs from the others in a list is recalled better than the same item sitting in a list where nothing stands out.
An item that differs from the others in a list is recalled better than the same item sitting in a list where nothing stands out.
The test is the isolation paradigm. One item in a list differs from the rest on some dimension, by material, colour, category or size, and memory for that isolate is compared with memory for the same item in a control list. Which control list is used decides what the comparison means. Against a homogeneous list, where every item is of one kind, the isolate wins by a wide margin.
Against an unrelated list, where every item differs from every other, the isolate still wins, and that second comparison is the one that carries the theory. Two readings of the result exist and they are testable apart. Either isolation lifts the odd item above what it would otherwise reach, or the similarity of the surrounding items drags them down, or both happen at once. Von Restorff's own data show both, and her explanation favoured the drag.
Replicated Hunt repeated von Restorff's own early-isolate procedure with 40 undergraduates and recalled the isolate at .70 and .80, against .40 and .30 for the same item in unrelated control lists.
Hunt ran the design between subjects rather than within, so each participant saw one list, and he fixed the isolate in the second serial position rather than the second or third. Twenty participants took the isolation lists and 20 the unrelated lists. Recall of the critical item was .70 when a syllable sat among nine digits and .80 when a digit sat among nine syllables.
The matched items in the unrelated lists came back at .40 and .30. Von Restorff's figures were .70 and .40. MacLeod, reviewing the literature in 2020, described the effect as clearly robust and readily replicable, and counted 236 citations of the 1933 paper in PsycInfo and 817 in Google Scholar. Robustness is not uniform across tests. Kamp and colleagues report that there is rarely an isolation effect for recognition, which they call inconsistent with intuition and with all known theories of memory.
Three of their experiments extended the work to a multilist procedure and found a free recall advantage for unexpected words, no recognition advantage unless recall was tested first, and an ambiguous result for source memory. Three further experiments took more observations per subject in the critical condition and controlled the order in which items were tested, and those revealed von Restorff effects for both recognition and source memory.
The one word in capitalsillustration
A shopping list is written in the same hand throughout except for one item in block capitals. That item is the one recalled at the till, and the ones around it fare slightly worse for its presence.
The slide with a different backgroundillustration
A deck of twenty slides in one template carries a single slide in another colour. A week later that slide is the one people can describe, whether or not it held the point that mattered.
The notes highlighted everywhereillustration
A student marks most of a chapter in yellow. Nothing now differs from anything else, and the advantage one highlighted sentence would have had disappears once highlighting is the norm.
Cases marked as illustrations describe nobody in particular. They are there to make the pattern recognisable, not to report that a particular thing happened.
MacLeod tabulates nine main experiments from the dissertation with samples between 4 and 84. The first showed five lists of eight paired associates built from five kinds of material: nonsense syllables, numbers, letters, keyboard characters and colours. Four pairs in a list came from one kind, and the remaining four were each the only pair of their kind.
Pairs appeared for 2 seconds, three times over, and recall followed a 6-minute conversation. Massed pairs came back at a little over 40%, isolated pairs at almost 80%. A control run over 22 subjects gave 189 hits from 440 for massed pairs against 328 from 440 for isolated ones. The experiment that carries the argument came later. Fifteen subjects saw three ten-item lists, one card every 1.5 seconds, then read a text for ten minutes and wrote down what they could recall.
Two lists held nine syllables and one number, or the reverse, with the odd item in the second or third position. The third list held ten unrelated things: a number, a syllable, a colour, a letter, a word, a photograph, a symbol, a button, a punctuation mark and the name of a chemical compound. Massed items were recalled at 22% and the isolate at 70%.
For the same critical items inside the unrelated list the translation prints no percentage, giving a mean of 6 against 6.5, or 43%, for the other items in that list. The .40 usually quoted for that condition is Hunt's figure, not hers. The translation's tables sit on the page as placeholders, so every figure above comes from its running text, and MacLeod's Table 2 gives .22 and .70 for the same experiment.
Von Restorff placed the isolate early on purpose, writing that at the second and third positions the subjects 'could not know anything about the contents of the whole list', so the item 'was not perceived as unusual and was not particularly salient'.
8 primary research2 commentary and critique1 background and reference
The paper is her Berlin dissertation, printed as pages 299 to 342 of volume 18. Those pages carry the joint heading 'Analyse von Vorgängen im Spurenfeld' with Köhler named alongside her, and Crossref holds a later instalment under that same joint title at volume 21, pages 56 to 112, in 1937. MacLeod records that the work was done with Köhler and that she took the doctorate magna cum laude on her birthday at 27.
No English translation has ever been published, and the German original was not obtained for this entry. Everything attributed to her here is read off an unpublished translation by Andrea Dorsch, which MacLeod cites as Köhler and von Restorff (1995).
The claim, and the replication this entry reads its figures from.
No meta-analysis of the isolation effect was located. Europe PMC, Crossref including its preprint records, OSF Preprints, Semantic Scholar and general web search were tried on 17 August 2026, and what came back was reviews and single studies. Those same searches found no preregistered replication of the isolation paradigm. Wallace's 1965 review sits behind the APA paywall, and ProQuest, Unpaywall, Semantic Scholar, Europe PMC and PsycNet were tried on the same day without reaching the text, so its content is reported as MacLeod describes it: three groupings of study design, a conclusion that isolation facilitates learning of the isolated item, and an admission that at the theoretical level the phenomenon remained controversial.
The early-isolate result, which is the load-bearing part of Hunt's case, does not hold for every kind of contrast. Schmidt and Schmidt tested 87 students in their first experiment and analysed 84, and they found the usual advantage for a midlist isolate, .55 against .43 for the same item in a homogeneous list and .27 in an unrelated list, but no advantage for an early isolate defined by semantic category.
Early conceptual isolates came back at .44 against .63 for the same items in homogeneous lists, an impairment rather than a gain, and their second experiment reproduced that impairment without the distracting task. An early advantage survived in the same paper when the contrast was a number among words or a word among numbers. Applied claims rest on thin evidence.
Pick, Sweeney and Clay tested 60 undergraduates, 22 men and 38 women of mean age 19.25, on lists of nine advertising slogans under incidental learning, and found the effect for a distinctive high-quality slogan but not for a distinctive substandard one. The same venues on the same day turned up no comparable experiment in interface design.
Each of these is named in this entry's own prose as something it gets taken for. Where the two came out differently, that is worth knowing before you quote either.
The textbook picture, one odd item in the middle of a long list, appears once in her paper. Hunt counts a single experiment where she isolated one item midlist, and it was a recognition test, adopted because her earlier recognition lists with several isolates gave differences smaller than her recall lists had. Everything else used paired associates with four isolates in eight pairs, or a single isolate near the start.
Her explanation also runs the other way from the popular one. MacLeod credits Hunt with the point that she treated the advantage as a cost to the massed items rather than a bonus for the isolate, and her own numbers carry both: 70% for the isolate, 22% for the massed items, and, on Hunt's figure for her unrelated control list, .40 sitting between them.
Perceptual salience is not the mechanism she found, and it is not required for the effect. Hunt cites Pillsbury and Raush's 1943 report of a substantial isolation effect with the isolate in the first serial position, where no context yet exists to stand out from. MacLeod reports the same result from Kelley and Nairne in 2001. She did not name the effect.
MacLeod records that Jenkins and Postman still called it the isolation effect in 1948 and that Green first attached her name to it in 1956. She was not a psychiatrist or a paediatrician, though Wikipedia's article on the effect still described her as both on 17 August 2026. MacLeod, working from Humboldt University archives, has her dismissed from the Psychological Institute in 1935, entering medical school that year at the Pharmacology Institute in Berlin, licensed to practise in 1939, publishing a second dissertation in medicine in 1940 and working as a family doctor in Freiburg until her death on 6 July 1962.
Every claim on this page was held against these sources on 2026-08-17. Nothing here is written from memory. Written and checked by Krishna Chagti; how entries are written and corrected. Found a mistake? Report it, with the sentence and a source.