# The narrow gate of memory

Someone reads you a phone number and you repeat it under your breath until you can dial it. If someone asks you the time in between, the number is gone. Almost everything you know, from the times tables to the way home, entered your mind through the same narrow gate: **about four things at a time**, held for **roughly twenty seconds** unless you keep repeating them. The best-tested rules of learning work with those two limits.

Full edition, with two memory tests and a review planner: https://mariuscomper.uk/poarta-ingusta/en/

## Two rooms and a gate

The mind works with two very different spaces: working memory, small and noisy, and long-term memory, vast and silent. Everything that reaches the second passed through the first.

- In 1956 George Miller published “The magical number seven, plus or minus two”; he later said seven had been a rhetorical joke. Nelson Cowan showed in 2001, gathering dozens of experiments without rehearsal or grouping, that the real limit is around four items in an adult.
- In 1959 Lloyd and Margaret Peterson read students groups of three consonants and had them count backwards in threes. After 3 seconds about 80% of the groups were recalled; after 6 seconds about 50%; after 18 seconds under 10%.
- Alan Baddeley and Graham Hitch described working memory in 1974 as a system with several parts. What matters is the consequence: a pattern kept active often enough consolidates and becomes long-term memory.

## The secret is the size of the pieces

The gate counts chunks, meaning anything the mind already treats as a single thing. “CIAB BCFB INAS A” is thirteen letters; “CIA BBC FBI NASA” is four chunks and passes through the gate in one go. Steve Faloon, studied by Ericsson and Chase in 1980, came after more than 230 hours of practice to repeat strings of 79 digits by turning them into running times; with consonants he fell back to six or seven. The first rule of learning: a new thing is taught after its parts have already become chunks.

## When the gate jams

John Sweller built cognitive load theory on this limit in 1988. The worked example helps the beginner; the same example hinders the advanced student (the expertise reversal effect, Kalyuga, 2003). Tracy and Ross Alloway followed 98 children in the United Kingdom from 5 to 11: working memory at 5 predicted reading and mathematics at 11 better than IQ, in a single small-sample study. Working-memory training improves tasks similar to the ones practised, but the gain does not transfer to reading, arithmetic or reasoning (Melby-Lervåg and Hulme, 2013, 23 studies; 2016, 145 comparisons).

## What is lost after passing through

Ebbinghaus measured in the 1880s the time saved when relearning. Murre and Dros repeated the experiment in 2015 with a single volunteer.

| Interval | Ebbinghaus, 1880 | Replication, 2015 |
| --- | ---: | ---: |
| 20 minutes | 58% | 56% |
| 1 hour | 44% | 47% |
| 9 hours | 33% | 34% |
| 1 day | 27% | 46% |
| 2 days | 23% | 43% |
| 31 days | 9% | 4% |

The replication’s authors themselves flag the 31-day value as unusually low.

## Three habits that contradict intuition

- **Practice testing.** Roediger and Karpicke (2006): after a week, 61% of the text reproduced by those with one reading session and three test sessions, 40% by those with four reading sessions. Karpicke and Roediger (2008): Swahili pairs still tested after being known, about 80% after a week; dropped from testing, 36% or 33%. Karpicke and Blunt (2011): free recall 81%, concept mapping 58%, rereading 57%, read once 50%.
- **Distributed practice.** Cepeda, Pashler and colleagues (2008), 1,354 volunteers, 32 facts: at the right pause, 64% more retained than after an immediate review. Best pauses among those tried, between learning and review: 1 day when the test comes 7 days after the review, 11 days for 35 days, 21 days for 70 and for 350 days.
- **Interleaved practice.** Rohrer and Taylor (2007): 63% on the test a week later after mixed practice, 20% after blocked practice. Kornell and Bjork (2008), direct replication in 2026: 38% after mixed paintings, 29% after paintings grouped by artist; participants believed grouping had helped them more.

## Why it feels exactly backwards

Ease tells you how open the gate is right now. Rereading and immediate repetition keep the pattern lit in working memory, so everything feels familiar; that very familiarity is the sign that long-term memory has not been put to work. Robert Bjork called the habits that help, effort included, “desirable difficulties”.

The Dunlosky and colleagues review (2013), ten techniques: practice testing and distributed practice, high utility; interleaved practice, elaborative interrogation and self-explanation, moderate; rereading, highlighting, summarisation, keyword mnemonic and imagery for text, low.

## Four rules derived from two limits

1. Teach after the parts have become chunks.
2. Cut everything into pieces that fit.
3. Pull out rather than push in.
4. Let yourself forget a little, then return.

## The edges of these figures

Four items is an adult laboratory average, with individual variation from about three to about five. The 18 seconds come from one experiment with 24 students and nonsense syllables. The Alloway study has 98 children. The 2015 curve comes from a single volunteer. The mixing effect with painters came out more modest in reruns. Nothing here is medical or educational advice for a particular case.

## Sources

- Miller, “The magical number seven, plus or minus two”, Psychological Review, 1956. https://psycnet.apa.org/doi/10.1037/h0043158
- Cowan, “The magical number 4 in short-term memory”, Behavioral and Brain Sciences, 2001. https://pubmed.ncbi.nlm.nih.gov/11515286/
- Peterson and Peterson, “Short-term retention of individual verbal items”, Journal of Experimental Psychology, 1959. https://pubmed.ncbi.nlm.nih.gov/14432252/
- Baddeley and Hitch, “Working memory”, 1974. https://doi.org/10.1016/S0079-7421(08)60452-1
- Ericsson, Chase and Faloon, “Acquisition of a memory skill”, Science, 1980. https://doi.org/10.1126/science.7375930
- Sweller, “Cognitive load during problem solving”, Cognitive Science, 1988. https://doi.org/10.1207/s15516709cog1202_4
- Kalyuga, Ayres, Chandler and Sweller, “The expertise reversal effect”, Educational Psychologist, 2003. https://doi.org/10.1207/S15326985EP3801_4
- Alloway and Alloway, “Investigating the predictive roles of working memory and IQ in academic attainment”, Journal of Experimental Child Psychology, 2010. https://pubmed.ncbi.nlm.nih.gov/20018296/
- Gathercole, “Understanding how working memory problems impair classroom learning”, MRC CBU Cambridge. https://www.mrc-cbu.cam.ac.uk/our-research/gathercole/classroom-learning/
- Melby-Lervåg, Redick and Hulme, Perspectives on Psychological Science, 2016. https://pubmed.ncbi.nlm.nih.gov/27474138/
- Murre and Dros, “Replication and analysis of Ebbinghaus’ forgetting curve”, PLOS One, 2015. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0120644
- Roediger and Karpicke, “Test-enhanced learning”, Psychological Science, 2006. https://pubmed.ncbi.nlm.nih.gov/16507066/
- Karpicke and Roediger, “The critical importance of retrieval for learning”, Science, 2008. https://doi.org/10.1126/science.1152408
- Karpicke and Blunt, Science, 2011. https://pubmed.ncbi.nlm.nih.gov/21252317/
- Cepeda, Vul, Rohrer, Wixted and Pashler, “Spacing effects in learning”, Psychological Science, 2008. https://doi.org/10.1111/j.1467-9280.2008.02209.x
- Rohrer and Taylor, “The shuffling of mathematics problems improves learning”, Instructional Science, 2007. https://doi.org/10.1007/s11251-007-9015-8
- Kornell and Bjork, Psychological Science, 2008; direct replication, 2026. https://pubmed.ncbi.nlm.nih.gov/42329512/
- Dunlosky, Rawson, Marsh, Nathan and Willingham, Psychological Science in the Public Interest, 2013. https://doi.org/10.1177/1529100612453266

Romanian edition: https://mariuscomper.uk/poarta-ingusta/