You change a password, and for the next week your fingers keep typing the old one. Studying does the same thing: learn one drug, then its close cousin, and the first one keeps answering for the second. Psychologists call this proactive interference, the harm earlier learning does to later learning. Corbett and Duarte (2020) use a version of the password case to introduce it.
In retroactive interference, the reverse, later learning harms memory for what came earlier (Abel & Roediger, 2017). This post covers the forward direction. Unlike most topics in this series, it is a constraint rather than a strategy: you can shrink it, but you cannot opt out of it.
Similar Things Compete
Interference depends on resemblance. Proactive interference grows as items become more similar and shrinks as the gap between trials lengthens (Hamilton et al., 2022). A classic demonstration shows both the problem and a way out (Wickens et al., 1963, as cited in Hamilton et al., 2022). People saw a short list of numbers or letters, counted backward for a few seconds to stop rehearsal, and then recalled the list. Recall got worse with each new list. When the type of item changed, recall recovered. Researchers call this recovery release from proactive interference, and it shows the decline was not simply fatigue.
Arithmetic gives a clean example from outside the lab. Learning 9 × 3 = 27 is harder than learning 5 × 5 = 25, because many more multiplication facts share the digits 9, 3, 2 and 7 (De Visscher & Noël, 2014, as cited in Hamilton et al., 2022). One way to read this: each of those digits is a cue shared by so many facts that it struggles to pick out any one of them.
Medicine is full of material built this way. Drug classes share name endings and mechanisms. Murmurs, rashes and fracture classifications come in sets designed to be compared. Like the digits in 9 × 3, a shared suffix or mechanism is a cue that points to every member of the family at once, so it cannot pick out the one you need.
Most of these findings come from short laboratory tasks with lists, measured over seconds and minutes. How far they carry to weeks of study has not been tested directly.
What Sets the Limit on Working Memory
Working memory is the small system that holds whatever you are thinking about right now. Its capacity is small. Oberauer et al. (2016) weighed three explanations for that limit and judged interference among the items held to be the best of them, while leaving room for a small general resource. That verdict concerns interference in general. Hamilton et al. (2022) go further and count proactive interference among the main limits on working-memory capacity.
Endress and Potter (2014) tested this directly. They flashed pictures of familiar objects at four per second, too fast to rehearse, and then showed one test picture: had it been in the sequence? For one group every picture was new. For the other, pictures came back from a small set, so earlier trials could interfere with later ones. With recycled pictures, capacity estimates stayed in the usual range of about three to five items. With all-new pictures, estimates grew with sequence length, reaching about 30 out of 100 pictures in one experiment.
Endress and Potter warn that this does not show people have a real capacity of around 30 items. They attribute the large numbers to a temporary visual store, not to working memory as usually defined, and they do not expect the result to hold for meaningless items such as colors or shapes. On a surprise test about half an hour later, people did only slightly better than chance. Still, much of the familiar three-to-five limit in this task came from old pictures competing with new ones, and that competition is what you set up when you study look-alikes back to back.
Holding Interference Off Takes Attention
Kane and Engle (2000) asked why some people resist interference better than others. They screened undergraduates for working-memory span, a measure of how many words a person can remember while solving arithmetic problems. High and low scorers then recalled successive lists of words from one category, so interference built up from list to list. Some did this while tapping out a demanding finger sequence during study, recall or both.
Without the demanding task, low scorers suffered more interference than high scorers. With it, the two groups suffered equally, because the extra load raised interference only for the high scorers. The authors concluded that high scorers use controlled attention, deliberate and effortful focus, to hold interference off both while learning the words (encoding) and while recalling them (retrieval). They describe the process as “perhaps inhibitory in nature” (p. 336). They also suggest that high scorers may give each list its own distinctive encoding context, a reading Oberauer et al. (2016) share.
The high scorers’ edge depended on attention being available, and a second task took it away. Studying with a phone buzzing beside the textbook is a rough everyday version of that second task. Studying late, when you are tired, may work the same way, but these studies did not test tiredness.
Who Is Most Vulnerable
Children are more susceptible than adults. Kail (2002) combined 26 studies of the list-recall task described above and found that interference declined between ages 4 and 13, although most of the children studied were 4 to 9. In a separate experiment, a statistical model linked the decline to faster processing. Hamilton et al. (2022) propose a different mechanism, the slow maturing of frontal and parietal brain networks that resolve interference, and note that the evidence for it in children is mostly indirect. Neither account has been tested against the other.
Older adults are a more mixed case. Corbett and Duarte (2020) report that earlier studies generally found more interference in older adults, while results for associative memory, remembering which things went together, were mixed. In their own brain-scanning study, 25 younger and 25 older adults linked objects with faces or scenes by forming interactive mental images, and the two groups were similarly susceptible. The authors suggest the imagery may have helped, and that older volunteers for brain scans may be unusually healthy.
What to Do About It
Because most of the evidence comes from short tasks, each move below is an inference from it. None has been tested as a study plan over weeks.
Separate look-alikes in time. In the lab, interference shrinks as the gap between trials grows, so two similar drug classes learned in different sessions may compete less than the same two learned back to back.
If two similar topics have to share one evening, separate them with something different. Release from interference came from switching item type. If beta-blockers and calcium-channel blockers both have to happen tonight, an unrelated block such as anatomy cards could sit between them. Kane and Engle (2000) add a caution: they noted, anecdotally and without data, that interference was not always fully released even after a change of category and a short break.
The setting may matter too. In a modest test by Anderson (1983), 80 people learned three lists of word pairs, one a day, in one of two very different settings: a computer in a windowless cubicle or a person reading the list aloud in a windowed seminar room. Lists learned after a change of setting showed little build-up of interference. Lists learned in an unchanged setting showed more. A different library or time of day is a much smaller change.
Shared cues are the problem, so give each look-alike a cue it does not share: the one feature that sets it apart, or the clinical situation where it alone fits. It also helps if the first version you learn is an accurate one. This follows from the distinctive-context reading above, and it may help explain why interleaving, mixing similar categories so that you must tell them apart, seems suited to confusable material (covered earlier in this series).
Test yourself, though the evidence here is indirect. In one experiment retrieval practice protected word pairs from retroactive interference, the reverse direction, and the interference was small (Abel & Roediger, 2017). It concerns retention of the practiced pairs, not transfer to new problems.
For anyone who teaches or designs a course, the same reasoning applies to sequence. The multiplication result suggests that how much look-alike facts overlap, and the order in which they are taught, bears on how much they interfere. A lecture that presents five similar agents in a row creates the conditions the lab studies use to build interference. Spreading them out, or teaching the contrasts between them directly, is the obvious alternative, though no one has tried it at course scale.
What This Looked Like for Me
In my first two years of medical school, the subject that humbled me most was pharmacology, and interference was most of the reason. The drugs came in families of deliberately similar agents, with shared name roots and overlapping mechanisms. I would learn one beta-blocker cold, then learn the next two days later, and suddenly I was no longer sure which selective property belonged to which, or which of two similar agents was contraindicated in the case I was picturing. On a clean recall test I was fine. On questions that made me choose between near-identical options, the older drug crowded out the newer one.
What helped was arranging my studying around the problem. I stopped cramming similar agents into one sitting and spaced them apart. I made each one distinctive on purpose, anchoring it to the specific patient I would use it for and the single feature that set it apart. I also noticed that the confusions multiplied late at night, when I was tired and half-distracted.
I cannot say how much of the improvement came from managing interference. I changed several things at once, including the spacing, retrieval practice and elaboration I adopted around the same time, and I ran no experiment on myself. What did change was how I read the blurring: as a predictable result of learning similar things close together, something to plan around.
The Core Lesson
When similar material blurs, interference is often the cause. Earlier learning competes with later learning, more so when the two resemble each other and arrive close together. Changing the type of material or the setting eases it in laboratory tasks, and holding it off draws on attention. How well these short-task findings carry to weeks of study is still untested, so treat the practical moves as reasonable bets.
The next post turns to mnemonics, devices built to give a specific fact its own retrieval route.
This post is part of a series on the science of learning. Each post covers one evidence-based principle and how to apply it to your own studying. Follow us on Instagram @acemeded to keep up with future blog posts and related content
References
Abel, M., & Roediger, H. L., III. (2017). Comparing the testing effect under blocked and mixed practice: The mnemonic benefits of retrieval practice are not affected by practice format. Memory & Cognition, 45(1), 81–92. https://doi.org/10.3758/s13421-016-0641-8
Anderson, J. R. (1983). A spreading activation theory of memory. Journal of Verbal Learning and Verbal Behavior, 22(3), 261–295. https://doi.org/10.1016/S0022-5371(83)90201-3
Corbett, B., & Duarte, A. (2020). How proactive interference during new associative learning impacts general and specific memory in young and old. Journal of Cognitive Neuroscience, 32(9), 1607–1623. https://doi.org/10.1162/jocn_a_01582
Endress, A. D., & Potter, M. C. (2014). Large capacity temporary visual memory. Journal of Experimental Psychology: General, 143(2), 548–565. https://doi.org/10.1037/a0033934
Hamilton, M., Ross, A., Blaser, E., & Kaldy, Z. (2022). Proactive interference and the development of working memory. WIREs Cognitive Science, 13(3), Article e1593. https://doi.org/10.1002/wcs.1593
Kail, R. (2002). Developmental change in proactive interference. Child Development, 73(6), 1703–1714. https://doi.org/10.1111/1467-8624.00500
Kane, M. J., & Engle, R. W. (2000). Working-memory capacity, proactive interference, and divided attention: Limits on long-term memory retrieval. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26(2), 336–358. https://doi.org/10.1037/0278-7393.26.2.336
Oberauer, K., Farrell, S., Jarrold, C., & Lewandowsky, S. (2016). What limits working memory capacity? Psychological Bulletin, 142(7), 758–799. https://doi.org/10.1037/bul0000046
