How school turns off memory

Jump to navigation Jump to search

This text is part of: "I would never send my kids to school" by Piotr Wozniak (2017)

Serendipitous proof

This is a story of a single English word: "sausages". By an incredible coincidence of three events, "sausages" perfectly illustrate the futility of schooling as measured precisely in SuperMemo.

Graphs presented below show how the association with school can literally turn off long-term memory

My relative Thomas is 8 years old. He is a young user of SuperMemo. He makes his repetitions with his mom. His repetitions are not voluntary, but he chooses a great deal of his learning material on his own. One day, for a school test, he was to learn a dozen new words in English (his native language is Polish). One of the word pairs was "kielbaski = sausages". Thomas struggled to memorize the word, and failed it on the test. That was March 2019. His mom put the word to Thomas's SuperMemo collection in hope he would remember it for future tests.

Coincidence 1: duplicated item

The first coincidental event occurred on March 20, 2019. Thomas's mom uses a small laptop that makes typing difficult. Thomas's collection is full of empty items (literally dozens). This can be explained by mom's struggle with the small keyboard. On Mar 20, 2019, 68 seconds after creating a question about sausages, mom must have accidentally issued a command Duplicate (icon for duplications is a neighbor of the icon for clipboard operations). The question about "sausages" was duplicated, and created two separate items.

Coincidence 2: applied template

Thomas has different templates in his collection. He got a template for English vocabulary. He does not like this template because he associates it with difficult vocabulary drills for school. He also got a template for his own items about things he likes, e.g. animals (including English names of animals).

The second coincidental event occurred most likely on the same day, i.e. Mar 20, 2019. For unknown reasons, Thomas's mom applied his favorite colorful template to one of the duplicated question about sausages (in some context, auto-apply can also occur). The change of template, in effect, created two identical items with two different templates: the liked template and the disliked template.

Between Mar 20, and Jun 15, 2019, Thomas reviewed "sausages" with the disliked template 11 times. All repetitions ended in failure. In June, mom deprioritized the branch with school knowledge to make sure Thomas did not get to hate learning during his vacation.

In the meantime, a shocking thing happened with the duplicate of "sausages" that was marked with the liked template. The liked-template item about sausages was reviewed 10 times, and only the first review ended in failure. The schedule for the successful item was denser than it is usually determined by SuperMemo. This indicates a manual intervention in intervals, however, this has little bearing on the story. The next repetitions for the two items were scheduled on Sep 17 and Sep 20 (both affected by 11-16 postpones due to the low priority set at the start of the vacation). The review of both items in Spring overlapped in time, which leaves only one explanation for the strikingly different performance: the template color determined the outcome! The successful color was associated with the material the child chooses on its own. The failing color was associated with items learned for school. For those who believe in the magic of colors, the disliked template was white, the liked template was red with big orange fonts. However, I confirmed later, that it was the association of the disliked color with school that made it "toxic".

The pleasure of learning plays a role in recall even if it is not directly associated with the content of knowledge

Coincidence 3: real life surprise

On Sep 15, 2019, the third coincidental event happened. Without that coincidence, Thomas might have kept reviewing his duplicated items for years to come without ever noticing. At some point, he would probably be able to recall both items to perfection and they would disappear from view.

The third coincidental event occurred when Thomas's mom discussed the problem of English during a family dinner. She complained that Thomas cannot remember simple words. She asked dispassionately at the dinner table: "Thomas, how do you say "sausages" in English?". To her amazement, the kid replied correctly! However, he did not say "sausages", he answered "frankfurters"! This called for a major investigation! It was pretty obvious that Thomas learned the word from another person. Nobody around uses the word "frankfurters", Thomas never eats frankfurters. We hypothesized that the person who used the term was Thomas's grandfather. Indeed, using his rich record of pictures of his grandson, gramps has helped complete the picture. On August 23, 2019, gramps prepared hot dogs for Thomas, and later, to test Thomas's knowledge, he asked in English "Do you want some frankfurters?". The hungry kid exclaimed "Yes!". That one single event must have etched the word "frankfurters" in Thomas's memory. The interval of 23 days sounds huge in comparison to the progression of up to 11 days in spaced repetition of the "good item".

One memorable event may count for more than dozens of repetitions at school

Valuation hypothesis

Memory consolidation begins with memory valuation. The valuation is delivered by knowledge valuation network. Neural valuations can be overridden in coercion by means of metacognitive tricks. We can use our willpower to prompt the brain to remember things it might otherwise refuse to recall. We can memorize a boring phone number with some help from mnemonic techniques. It is easier to memorize a phone of one's girlfriend that it is to memorize a phone to a disliked co-worker. However, well-schooled adult brain can memorize phones easily. Before a test, a student can memorize almost anything by means of cramming.

The story of two sausage items can best be explained by the effect of the item color on knowledge valuation. A child's brain may develop a conditioned response that links the "white color" with the "low valuation". In such conditioning, anything associated with school will be hard to remember or even unmemorizable. The brain will automatically cut off the consolidation signal, and the recall signal.

A bad memory for the school material might result from the conditioned depreciation of knowledge that cuts off the recall and/or consolidation signal

Semantic brain

A child's brain employs semantic learning. This is highly beneficial for intelligence. However, this stands in the way of academic instruction, which is largely asemantic (see: Asemantic curriculum). In free learning, a child's brain is an excellent discriminator of asemantic content due to the natural guidance of the learn drive. At school, that discrimination is quickly conditioned out. Well-schooled kids become excellent crammers, however, they lose their semantic learning skills. This affects a child's intelligence.

For a little kid, SuperMemo does not work too well. If the item is associated with unpleasant context of schooling, it can become a toxic memory. If the item carries good associations and clicks semantically, it can be reasonably successful. However, nothing can compete with "real life" in the efficiency of encoding semantic memories. These memories can then efficiently be perpetuated with spaced repetition.

It takes years of practice to develop metacognitive skills that make it possible to abstract the essence of knowledge from its inconsequential context for the purpose of learning with SuperMemo. Having used the program for three decades, I am entirely insensitive to the color of templates, which I use as domain-specific context indicator. This is why my love for SuperMemo is untarnished. However, this ability to abstract the essence of knowledge is a metacognitive skill. Little kids cannot do it effectively. Many adults cannot do it. Schooling favors literal memorization of knowledge. Striving at building abstract knowledge in free learning does the opposite.

Self-directed free learning exhibits unmatched efficiency at younger ages. Intervention in that process is usually harmful

Follow up: hypothesis

If minor variations in the context have a major impact on recall, the interesting question arises: what answer would Thomas provide to his two differently colored items in SuperMemo today? The words were not reviewed during the summer vacation. At the moment of writing these words I was not sure. I would guess that (1) verbal question would result in "frankfurters", (2) question called by the liked item in SuperMemo might result in "frankfurters" or "sausages" (due to the old context), or failure (due to a large interval), and (3) the disliked item might result in recall failure, or perhaps the same answer as the liked item (due to memory transfer, or due to the actual use of the knowledge). The experiment to test my ideas would be very hard to design for easy interpretation. In good design, the questions should be separated in time (to avoid interference), but not too far (to avoid new learning).

Follow up: test

Due to time constraints, I designed a procedure to test all memories on the same day (Sep 16, 2019). Instead of spreading items in time, I spread them in the learning queue. For a metacognitively weak child, interleaving items has a similar effect as longer intervals. I buried the test questions in a larger number of items. I took them from the disliked English branch of Thomas's collection to minimize the outlier effect. I got my answers in just 7 minutes. The test was successful in that it provided unambiguous answers. The results were congruent with the original hypothesis:

  • first I asked about template preferences. As expected, the kid explained "I like the red template because the tests are easy". The items Thomas chooses and likes are easy so he likes them even more
  • I started from a verbal question about sausages, the answer was "frankfurters" as expected
  • when the disliked item about sausages arrived, the answer was "I do not know". This means that the disliked template makes the kid fail a question that he answered correctly 1-2 minutes before! This is a classic illustration of toxic memory. As if the brain was paralyzed by the association with school, coercion, and displeasure. There are no apparent signs of making an effort to retrieve a word from memory!
  • when the liked item arrived, the answer was a surprise! It was not "sausages", it was not "frankfurters", but it was "meat". I presume this would qualify as a correct answer, and it is possible the kid kept answering "meat" in the past (which may explain mom's anxious work on shortening intervals). We may never know. Mom has no recall
  • for goodbye, I decided to ask again the answer with the disliked question and, this time, the kid confabulated. He mangled the Polish word "kielbaski" to make it sound like English: "kelbosky". When kids do not know the answer, they do not like admitting it. School teaches them to game the system with a plug word "meat" or a trick "kelbosky"

Incidentally, a bit of confusion was added by the fact that, in the same test, Thomas answered to 2-3 others items with the word "meat". This was his plug word for "I do not know". Perhaps he did not remember "sausages" in the first place either? However, context dependence and the impact of liking are strikingly visible:

Minor changes in the context, may result in major differences in recall

Memory in coercion

The graphs of the two identical items associated with different contexts tells the shocking story of the negative power of coercion:

Freedom facilitates memory consolidation

When the question is adorned with enjoyable colors that bring no negative association, memory performance is undisturbed. The following graph is typical of review of difficult material in spaced repetition:

Figure: Changes in the ability to remember the word "sausages" in spaced repetition in the course of 6 months. 10 peaks in the top panel (retrievability) correspond with one failure (in red) and nine successive recall successes (in blue). Blue area in the bottom panel illustrates the buildup of memory stability. The presented graph is typical for a dense schedule in an 8-year-old child. For comparison, the exactly same question, at the same age, in the same child, may result in a drastically different performance if learning is associated with coercion. See: Impact of coercion on recall

Coercion disables memory recall

Associating a white background color with coercion results in the cut off of the knowledge valuation signal. This in turn makes it impossible to retrieve a memory that has been consolidated successfully with another item that was free of toxic associations.

Figure: The mere association of learning with coercion can turn off knowledge valuation, knowledge recall, and consequently, knowledge consolidation in spaced repetition. The same piece of knowledge, at the same time, in the same person, may be shown to reside in memory having been correctly consolidated. Compare: Consolidation in free learning

Harmful schooling

In conclusion, this example shows the effect of schooling and coercion. Instead of learning English, the kid learned to dislike English. The good teacher, the loving mom, and the reputable school did nothing wrong except for demanding learning for a test. The formulation of the item was not flawed! It performed perfectly with a change in the context. The only identifiable culprit in forming a toxic association is coercion. SuperMemo can magnify the hate of learning by drilling disliked items. However, it can also magnify the pleasure of learning by reviewing well-formulated items of one's own choice. In that sense, SuperMemo is just a bystander and an amplifier.

At school, instead of learning, kids often learn to dislike learning

Cheating the brain

In the weeks after the serendipitous discovery of the magic of a template, I discussed the subject with teachers and psychologists in my cycle of friends. I knew the impact of associations on pleasure, incl. the color of templates. I know it well from my own collection too, esp. in reference to older knowledge from the days of lesser metacognitive "SuperMemo skills". However, the episode with sausages provides the clearest and most powerful example that coercion in learning can literally turn off memory.

Several times in my conversations I heard an impromptu idea from experienced educators: "to improve kid's English, swap the templates, so that his favorite template was used to present the knowledge of English". This is actually a horrible idea! The memory problem does not stem from the color, but from the valuation of knowledge. The happy association with a liked color would be extinguished fast.

This idea of cheating the brain is actually pervasive in the entire process of schooling. Kids keep receiving fake rewards and fake signals at school. The purpose of all that cheating is to drag the kid through one more stage of the curriculum. However, this also leads to a signaling conflict.

Swapping the colors of liked and disliked items would send contradictory signals to the brain. This is the basis of war of the networks. This is how schools and parents lose authority. Associations become meaningless because they stop serving an important purpose: distinguishing meaningful knowledge from meaningless knowledge. Very often, the division goes along semantic vs. asemantic lines. However, the ultimate answer is in the knowledge valuation network. Brain saves a lot of time and effort by detecting the disliked color and by giving up on the effort of understanding or encoding dislikes knowledge. If we keep tricking the brain and overriding its associative achievements, we quickly arrive at learned helplessness (as in this figure). The brain starts ignoring all knowledge delivered at school, and this may generalize to learning overall. It may lead to claims such as "I am not good at physics" or "I have no head for languages". Those claims are hard to uproot and many people die with false convictions of their own intellectual shortcomings.

Very often, continual cheating of the brain may result in the whole idea of learning become associated with displeasure. Kids lose the pleasure of learning, which is their best weapon on the way to high achievement.

Teachers have excellent intuitions and skills for cheating a child's brain. Sadly, they are not aware of the extent of the damage inflicted by signal conflicts to a child's intelligence

Further reading


  • little children employ semantic learning and their metacognitive skills are poor
  • schooling conditions tolerance for asemantic learning, which leads to fast forgetting, weak generalization capacity, and the hate of schooling
  • minor changes in context can result in major changes in recall
  • pleasure of learning plays a role in recall even if it is not directly associated with the content of knowledge
  • even a weak association with school can lead to turning off the consolidation signal (the learning does not occur)
  • at school, instead of learning, kids often learn to dislike learning
  • early in life, self-directed free learning is the most efficient form of learning
  • early academic instruction may cause harm by forming toxic memories
  • one memorable event may count for more than dozens of repetitions for school

For more texts on memory, learning, sleep, creativity, and problem solving, see Super Memory Guru