- 1 Introduction
- 2 Measuring amnesia
- 3 Neurogenesis hypothesis
- 4 Measuring infantile amnesia with SuperMemo
- 5 Measurement example
- 6 Retrospective recall curve
- 7 Childhood amnesia: Mechanism
- 8 My first memories
- 9 Conclusions for parents
- 10 Summary: Childhood amnesia
Childhood amnesia is a period of 2-4 years when a child is unable to form memories that might last a lifetime. Interestingly, there is no actual amnesia in healthy childhood. Childhood amnesia is a misnomer. I will explain.
Most importantly, childhood amnesia has a monumental impact on learning and development strategies. A large amount of early learning and acceleration programs do more harm than good. A great deal of child development mythology has its roots in childhood amnesia. Understanding and measuring memory in childhood is essential for protecting childhood from misguided intervention that affects millions of kids around the world as we speak.
Childhood amnesia is known to almost everyone. Except for a few cases of false memory, we do not seem able to recall much or anything from before the age of 3-4. Researchers have for long been interested in the phenomenon hoping it could shed some light on how memory works and how it might benefit learning later in life. The old and tried method of collecting data on childhood amnesia is an interview.
The difficulty in establishing earliest memories by means of interview is compounded by the fact that the interview must be conducted early. Delaying the interview may shift the estimate of the first memory date to later years. The problem is that kids aged 4-5 cannot reliably answer the question "What is the earliest thing you remember?". Even the concept of "earliest" might be unclear. Dates are easily confused. Episodes get confused. Interview that does not involve adult witnesses is of little value. The only sensible approach to similar testimony is to assign each report a probability value. Only a large number of reports from early childhood with probabilities assigned to each event may provide a rough approximation of the likely date of the earliest memory.
I asked a 4 year old about memories from her baby times and toddler times. She said "I vomited egg when I was 1.5 years old". 1.5 seems way too early to form memories, but the girl was precocious and a little memory genius. However, the testimony of "1.5 years" is highly unreliable when taken from a kid. It might be a form of confused memory, implanted memory, misunderstanding, etc. Kids at 18 months have no concept of 1.5 years, hence the date itself must be an artifact. In this case, probability would have to be near zero and the main value of such a confession would be to warrant further investigation.
When the father of the girl confirmed that she indeed ate egg at 1.5 and vomited, the probability of that early memory could increase, but still be very low. It could be a case of implanted memory. For example, girl's grandmother might recall at feeding time: "when you were 1.5 years old, you vomited an egg". This would plant that "early" memory at later time.
When family insisted it was not the case, and added a fact that the girl refused to eat egg ever since, the probability would increase further. In this case, the aversion memory appears to be well documented and surviving since early childhood. However, in investigating childhood amnesia we are more interested in abstract declarative memories or episodic memories. Could the girl indeed form a memory of the episode and keep it for 3 long years despite a rapid growth of her brain? Memory reimplantation might simply occur on any day when a child refused to eat egg and someone mentioned the episode of vomiting. Considering the very early age of that memory and possible alternative scenarios, the veracity of the report would still have to be ranked relatively low.
It seems that for an individual child, only a close caregiver could note a sufficient number of episodes provided with sufficiently reliable documentation and sufficiently high probabilities to provide a good hint on the childhood amnesia window for a particular child. The error rate could still be very high.
These days, kids have many caregivers, spend a great deal of time in daycare, and reliable research would probably require a single parent living with a child in highly controlled conditions. Even then, it would be a result relevant to that child only.
Forming first memories
To find out when lifelong memories begin seems like an interesting question. It might shed some light on how the young brain works. However, if we look at details of the process in which earliest memories are formed, the precise timing becomes far less interesting. It is simply a matter of chance. As such, finding the exact number for an individual is interesting but largely inconsequential. Finding the number for a large number of kids would be far more valuable but very hard to achieve.
Here is a sequence of events that should lead to the formation of the earliest memory lasting a lifetime (for more details see Childhood amnesia: Mechanism):
- a traumatic, stressful, or highly exciting event would form a set of episodic memories
- those memories should survive for a while by involvement of sufficiently many connections that would resist forgetting and be restored by pattern completion
- complete forgetting would be prevented by a form of review, which may be a conversation, or more likely, a return to a memory by thinking about it or recalling it in similar contexts
- due to the involvement of a larger number of synapses, the memory would mutate, re-mold, transform, and get restored over and over again to take new shapes. All we need is the core meaning of the underlying event to survive
The above scenario shows that the formation of earliest memories will depend on random chance events such as a specific trauma or an exhilarating experience. It will then be contingent on further review which may depend on environmental cues that may or may not be helpful in retaining memories. Finally, earliest memories, once determined, may actually be forgotten in mere months or even years. In that sense, the estimate made at the age of 4 may look better (younger) than the same estimate made at the age of 7.
The most important underlying process, which is the speed of forgetting, will be of consequence, but the timing of the earliest memory might change by several months or even years by sheer chance.
The old thinking about childhood amnesia is that it was either a storage problem or a retrieval problem. It is neither. Memories are stored ok, and can be retrieved ok until they can not.
Observing a child
To investigate infantile amnesia, the ideal setting would be a single parent with a single child living in some isolated mountain hut under close supervision of cameras. To make the research meaningful, we might need dozens of similar setups to investigate individual variability.
A single parent in isolation is ideal. Adding a second parent to the picture may make the documentation harder. It is impossible for one parent to perfectly communicate to the other parent about all interactions with a child. However, it might help if both parents were neuroscientists.
Sheena Josselyn and Paul Frankland are an unusual pair of neuroscientists. There are interested in how memories form and the role of neurogenesis in that process. They are also married and have a daughter Charlotte (see picture).
Jesselyn and Frankland do not live in an isolated hut, but they live in the second best place for research: science labs. Charlotte has spent her first five months at work with her parents.
Charlotte Frankland made history when she was mentioned in an important paper which hypothesizes on the role of neurogenesis in infantile amnesia.
That paper explains why brain growth and memories are incompatible, and why long-lasting childhood amnesia might actually be an indicator of long-lasting brain development, i.e. a good thing.
The neurogenesis hypothesis:
The young hippocampus is in constant flow. Memories stored in the cortex may lose their access wiring as a result of that fast growth. They can no longer be accessed, which means accelerated biochemical forgetting at the synaptic level after the actual forgetting at the neural level.
Nobel winning icon of memory research, Dr Eric Kandel agrees: "The hippocampus matures slowly and probably doesn’t reach any reasonable maturity until we’re 3 or 4. While 2- and 3-year-olds can remember things for a short time, the hippocampus is required for long-term storage of those memories".
Paul Frankland hopes to verify some of his hypotheses using research on kids suffering from brain cancer. In chemotherapy, drugs that slow down neurogenesis without actually damaging nerve cells may turn out to improve memory in childhood. If this research confirms Frankland's prediction, it will provide a harrowing proof that good memory isn't always a good thing in early childhood.
Kids have no long-term memory
Let's re-examine childhood amnesia as a retrieval failure in the context of the neurogenesis hypothesis. It is possible that due to neurogenesis, some cortical connection might become inaccessible. However, calling it retrieval failure would be a misnomer as it would suggest memories might somehow be retrieved given favorable circumstances. Those abandoned memories are like seeds of wheat on a desert: bereft and useless. Little kids virtually have no declarative long-term memory for semantic knowledge. Declarative recall in 6 months comes from review (e.g. names "milk" and "bottle" survive well with daily review). Episodic recall in 6 months comes also mostly from review, e.g. due to a traumatic nature of a memory, or in a specific repeated context (e.g. passing the same landmark on the way to daycare). Those kids that do have a genuine memory of a span of months might be precocious, but memory precocity may also imply shorter period for brain growth. If precocity is achieved by any form of acceleration or stress, this may spell the possibility of never growing up to one's true potential.
Measuring infantile amnesia with SuperMemo
SuperMemo is an excellent tool for studying memory. However, it cannot be used for studying childhood amnesia because amnesia refers to episodic memories that are not a typical subject of intentional review in SuperMemo. Moreover, SuperMemo would instantly evoke an observer effect in which duration of memory is artificially impacted by the measurement (review).
Here are the reasons while SuperMemo does not fit the bill:
- SuperMemo is not considered useful in retaining episodic memories, it is used for declarative memories that are useful but not too easy to remember
- by definition, review in SuperMemo contradicts the notion of earliest memories that should survive in child's mind as lasting memories without artificial support
- we have little SuperMemo data from children. Very few parents opt to use SuperMemo at young ages, and we actually discourage the practise as possibly harmful
On the other hand, the most interesting component of childhood amnesia is memory turnover and resulting forgetting. Forgetting in childhood is so fast that we can easily interpret it as a result of interference associated with the fast growth of the brain. In that sense, SuperMemo would be an excellent measure of the quality of young memory, except we would need to depart from the currently accepted notion of childhood amnesia.
Declarative semantic memories subject to spaced repetition are an excellent material to show the progression of long-term memory capacity with age. For example, longest inter-repetition interval achieved can be tracked over time. In SuperMemo, "best interval" is the longest interval that terminates with a pass grade.
In the following graph, we can see that until the age of 2, the child could hardly form memories reaching 2 months.
Naturally, in SuperMemo, best interval is also a function of the period in which the program had been in use (it takes years to produce intervals on the order of decades). For this reason, best interval graph is interesting, but it is not revealing.
A better measure of amnesia is to compare best interval ratios between two users: a child and an adult. In the graph, the starting point of 30% (for child's capacity) is an overestimate due the fact that the adult is initially limited not by his memory, but by the length of intervals achievable early in SuperMemo.
In addition, a 1 year old toddler may learn "obvious" things like "a car", which soon become "fixed" memory, which can be retained with or without SuperMemo.
The fact that a 3.5 year old can compete with an adult is already much better a reflection of memory properties. It needs to be emphasized strongly that adults also differ by a wide margin due to the differences in their ability to formulate well-structured questions in the learning process. A comparison with an excellent student might keep kids at a disadvantage for much longer. Even up to their teens years.
Every parent can attempt to contribute to the measurements of childhood amnesia. If you happen to follow the prescription below, please do not forget to mail your results to me.
Here is a simple algorithm:
- elect the minimum memory survival distance that qualifies for a data point. For example, 6-month period seems like a good yardstick. Events recalled with a 3-month delay might easily be classified as "candidates" due to the fact that amnesia does not end with the first remembered event, but with the first memorable event that can survive forgetting via incidental episodic review (e.g. through contextual reminiscence)
- document each data point as a triple: event date, recall date, and probability (e.g. 0.35 would mean there is a 65% chance that the recall could be hinted to or caused by factors other than actual surviving memory); naturally, each data point should get a detailed description of the episode because the data may need to be verified in the future
- keep sorting data point by the earliest recall date
- for each data point, starting with the earliest, compute the distance from earliest memory and the expected distance using the formula:
D- distance (i.e. recall date minus the earliest recorded remembered episode date)
P- probability (i.e. chances that the recall of the episode is not an artifact)
ED- expected distance (i.e. the best prediction of the distance to the earliest recalled episode in days)
CL- confidence level (i.e. cumulative chances that the recall is not an artifact)
In above formulas D(1)=0, ED(1)=0, and CL(1)=P(1). The end of amnesia is marked by the age computed from the date of the first recorded episode plus the expected distance (at a given confidence level).
An exemplary picture (actual data) based on 9 data points, and cut-off point of 8 months makes it possible to state with confidence of 90% that childhood amnesia lasted for the first 2.65 years of a child's life. The experiment could safely end at the age of 5 as confidence level on the end of amnesia was approaching 1.00, and the expected date of amnesia end was 2.84 (i.e. age of the first episode recorded plus the expected distance). Interestingly, at the end of childhood amnesia determined using a similar approach, outwardly, the child may still seem unable to remember beyond a week. This might explain why a high emotional content is essential for the memorable event to break through an average survival of episodic memories.
Obviously, measuring amnesia is a hazy science. Minimum memory survival is arbitrary. Confidence levels are wild approximations. Memories registered as "earliest" today may be forgotten in a year or in 10 years. New data points may show up at any time, even at teen age. New technologies make it easy to document events accurately even though verification of recall is always inexact science.
Dating amnesia at the age of 5 is interesting, revisiting the same data at the age of 10 can bring an entirely different result. Earliest memories can get lost. Testing recall of an older child should shift the actual boundary to a later age, but might also be more prone to false memories (e.g. implanted by a parent, recalled via pictures or videos, or resulting from conglomeration of memories).
Retrospective recall curve
Rubin and Schulkind in 1997 showed two phases of amnesia in the first decade of life using autobiographical memories. Those phases should easily be explained as a gradual process in which the capacity of long-term memory increase and interference related to brain growth is diminished (see: graph).
I was able to produce a similar graph using my own autobiographical memories. In an unrelated project, which I call "retrospective diary", I tried to collect all remembered memories from my childhood and adolescence. I started writing down those memories at the age of 13, and continue the process until this day. Most of events from the 1960s and the 1970s, I documented via recall that occurred in the early 1980s. I largely run out of new episodes to document by the 1990s. New data points are burdened with very low reliability. These might be planted memories or sheer fantasies and delusions. The greatest value of that effort comes from my meticulous effort to document all things I could recall. It was not a timed one-off effort. It was a continual lifelong effort in which I might interrupt my activities at any point in time only to jot down a recall of memories from primary school. The graph becomes meaningless after 1981 when my "retrospective diary" became serious and extended to all episodic memories from the times of college. Hence the explosion in the number of documented events. They were documented immediately after occurring.
Interestingly, on multiple occasions I documented forgetting of episodes recalled earlier. A primary school memory recalled at 20, might be forgotten at 30, or even at 50. Exhaustive recall documentation will then bring different results at different ages. If I was to begin this project at later ages, the result would differ. The number of documented episodes would drop significantly and the shape of the curve would be different. In other words, the recall curve in adulthood keep being reshaped by forgetting.
Earliest memories carry strong persistence, which probably comes from the badge of being the earliest memory.
Childhood amnesia: Mechanism
There is no actual amnesia in healthy childhood. Childhood amnesia is a misnomer.
The child begins forming memories as a fetus. Some of those are structural memories based on physical connections between brain cells, some affect entire pathways, and some will last for life. The child will never "forget" how to control its gastric functions or its heart rate using the central nervous system.
A good visual pattern recognition that makes it possible to recognize a bottle of milk, or mom's face is also based on memory. This type of memory will rather not fade. Recognition of sounds of one's native language is another unfading or low-fading example. Those networks form in a critical period, and neglect or denial will have a lasting effect.
The child also starts forming declarative memories early. Naming that bottle of milk would definitely be a specific declarative memory based on specific connections that should be retained in a longer term.
Those early declarative memories will rely on molecular changes in existing synapses. Some of those memories may also survive for life, however, we have no good way of knowing it. The problem is that there is a tremendous turnover of declarative memories in a developing brain. As a result, we have noticed that baby memories have a very short duration. That observable lifespan will increase gradually in proportion to a slowing turmoil of rewiring in the young brain.
There is little chance for using SuperMemo for those early declarative memories. They are volatile and keep reshuffling. Good score in SuperMemo might equally well be a result of reviewing the term in real life a day before. The algorithm makes little sense when memories get destroyed by fast growth.
A two year old toddler might find it difficult to remember things beyond a month. This span will be hugely dependent on temperament, health, upbringing, environment, etc. It can be annoyingly short in some toddlers, it can be "genius-level" in others. Paradoxically, long memories in a young child may be a sign of brain growth slowing down. The longer the period of growth and rewiring, the bigger and better the effect of child's "bad memory".
There is no actual amnesia. The term refers to a window of time in which a recall of childhood memories is unlikely. For most kids that window spans from 2 to 4 years. There is nothing amnestic about that window. The window is only a matter of competition between six primary processes: (1) forming memories, (2) forgetting memories, (3) memory interference, (4) brain re-wiring, (5) exposure to memorable events, and (6) post exposure review.
To form the first memory that can survive the lifetime, the following scenario must occur (memory process classes in bold):
- the child encounters an event that is particularly memorable or traumatic, e.g. trip to Disneyland (exposure to memorable events and forming memories)
- that memory must survive the turmoil of brain growth. The younger the child, the faster the growth, the less likely the survival of the memory (brain re-wiring)
- the event must be memorable enough to come back to child's mind in some circumstances, e.g. mention of Donald Duck on TV (post exposure review)
- the frequency of re-exposure to that prompted recall must be high enough to meet the criteria set by the 3-component model of long-term memory derived from spaced repetition experiments. In short, review must be frequent enough to prevent forgetting via decay (forgetting memories)
- brain growth will interfere with the core assumption of the 3-component model of memory: the neural substrate of memory is constant. The model is based entirely on forgetting via the decay of memory traces. The interference from brain growth will wane over time (brain re-wiring)
- high volume learning interferes with old memories. In a child, the volume is high due to a high novelty exposure and a high forgetting rate in a growing brain (memory interference). Newly formed neural patterns are subject to interference. Interference is massive. The disruptive brain growth component of interference will wane, the novelty component will wane too, however, knowledge interference is a factor that affects children and students for life. There is an effective remedy to interference. This remedy depends on knowledge formulation, and can be reduced with automated learning techniques such as incremental reading. Last but not least, lifelong self-directed learning is the best form of training in avoiding interference. Lifelong learning is the best formula for lifelong memories.
The above scenario explains why with each passing day, the chances of forming the first lifelong memory increases. The string of memorable events is unceasing and might increase with brain maturity (baby pram is not a good place for memorable encounters). Brain growth is slowing, which favors memory survival. Novelty level and the resulting interference is decreasing. At some point, a memorable event will stick and this moment in time closes the window termed childhood amnesia.
The memory that we will label as "earliest memory" is not a memory that naturally lasts for 6-8 decades, but the one that is memorable enough to be reviewed often enough.
Sadly, the earliest recalled event isn't usually the trip to Disneyland. Traumatic or unpleasant events are more likely to survive. This is why so many adults bring up daycare events as their first memory. The unpleasant memory may also come from unexpected direction. See how birthday candles contributed to the novel hypothetical neural mechanism explaining childhood amnesia.
All parents of boys who are late to speak, or late to ride a bike are told that "boys are slower than girls". This is true. This primarily boils down to testosterone and its impact on the brain. Slow development in boys is actually good news. Their brain growth window might be more turbulent. In addition to the proverbial glass ceiling facing all girls, there is still some advantage to slow-growing male brains. It will be males who will largely populate math and physics departments. They will rein in chess or even in Nobel Prize awards. Naturally, Judith Polgar or Marie Curie prove that all ceilings can be shattered with self-discipline and persistence.
My first memories
My second earliest memory is about daycare itself. In compulsory nap periods, I was not sleepy. Perhaps it was a wrong time, or I was a type of kid that grew out of napping at an early age. The kids were marched to beds after lunch and the supervisor would make sure they all had eyes closed. As I was not sleepy, I pretended to close my eyes and watched the room around. The supervisor barked: "Wozniak! I can see your eyes are not closed!". This seems like a monumental waste of time, wasted opportunity for learning and exploration and the first taste of the need to comply with authorities. This is the anti-thesis of healthy unconstrained development.I have lots of other memories from that period, but I suspect they might have been planted in my head later by reviewing family pictures. Those less pleasant memories, e.g. of bullying, seem to have survived best. This is exactly what theory says: early stress accelerates development of memory. This leads to many short-sighted strategies employed in child development. Early memory may often mean shorted time for development.
Conclusions for parents
Childhood amnesia has wide implications for parenting strategies. These are top conclusions I suggest be taken home from this text:
- all forms of early instruction must be approached with caution. Declarative memories do not survive long in the period corresponding with childhood amnesia. Memorizing capitals of the world at the age of 2 is likely to be wasted time. Until the age of 3-5, for most kids, best learning is learning via play. Child's brain knows best what serves it well
- stress accelerates learning but this is not the type of acceleration that is welcome in development
- stress slows down neurogenesis and favores memory survival. Parents should never celebrate acceleration of development coming from daycare. This is the type of "acceleration" that actually undercuts child's long-term development potential
- stress accelerates synaptic pruning. Parents should never celebrate acceleration coming from daycare or early tutoring. If it comes from stressful settings, it will fix the child's brain at the earlier developmental stage. It will prejudice child's position in life, her reactions, and her preferences
- stress of learning can lead to toxic memories (at any age)
- boys are slow to learn and develop. Boys are unruly and hard to discipline. Boys need extra patience and should not be brought down to heel. Freedom is essential for development. Paradoxically, bad behavior today may translate to high creativity decades later
- all "techniques" aimed at improving memory, learning, and concentration in early childhood are largely aimed at stifling natural behaviors that may favor long-term development and learning. Memorization and discipline can wait. Parents should focus on quality sleep, nutrition, exercise, love, and plenty of playtime
All the above points can be met by ceding some authority and entrusting kid with more decision-making. Child's brain has a great deal of "magic" powers in determining what's best for its development.
Summary: Childhood amnesia
- children rarely recall memories from before the age of 2-3
- retrospective recall shows exponential decline back in time and approaches zero around the age of 3
- children form episodic and declarative memories from birth. The term "childhood amnesia" refers to rapid memory turnover in childhood
- neurogenesis, i.e. birth of new neurons in the brain, is one of the reasons for childhood amnesia
- lifetime survival of early memories depends on (1) memory formation, (2) forgetting, (3) interference, (4) re-wiring, (5) exposure to memorable events, and (6) post exposure review
- earliest memories are often unpleasant (e.g. related to daycare). Stress tends to improve learning, and undermine brain growth
- with declarative memory span from 1 to 6 months, toddlers and young children have virtually no long-term declarative memory
- interviewing children for memories marking the end of childhood amnesia is hardly ever possible. Only a close caregiver can provide highly unreliable record that can be used in research
- children memory may start approaching adult performance slowly between ages 5-15. Exact numbers are hard to obtain. SuperMemo can be used in such comparisons
- poor declarative memories in children imply that formal instruction should best be delayed until the age of 7-9
- children learn best via self-directed play