Before she was an engineering professor, Barbara Oakley was a linguist: Russian at the Defense Language Institute, a degree in Slavic languages, seasons translating on Soviet trawlers. For a language learner that biography is the argument. This piece sorts her methods by how well the evidence actually supports them.
There is a detail in Barbara Oakley’s biography that most introductions skip, and for a language learner it is the most important one.
She began as a linguist.
Enlisting in the US Army straight out of high school, Oakley spent a year studying Russian at the Defense Language Institute in Monterey, took a degree in Slavic languages and literature at the University of Washington, and then worked several fishing seasons as a Russian translator aboard Soviet trawlers in the Bering Sea. After four years as a signal officer in Germany she left the army and began studying engineering — in her twenties, having spent her entire education until then avoiding mathematics and science. She is now a professor of engineering at Oakland University.
Learning How to Learn, the online course she teaches with the neuroscientist Terrence Sejnowski, has enrolled more than 3.6 million learners.
One thing needs saying in order to read her properly: Oakley is not a neuroscientist, and has never claimed to be. She is an engineering professor doing the work of popularisation. Most of what she teaches is real cognitive psychology, reported reasonably faithfully. So the job here is not demolition but triage: what holds, what is a teaching metaphor, and what has changed since she wrote it.
Oakley’s most-repeated advice is also her best-supported: shut the book and try to recall, rather than read it again.
Roediger and Karpicke (2006) showed that self-testing consolidates long-term memory far better than restudying for the same amount of time. Karpicke and Blunt (2011) pitted retrieval against concept mapping — a method widely regarded as deep — and retrieval still won.
For vocabulary this settles how a deck should be built. A card that asks you to recognise a word is comfortable and nearly useless. A card that forces you to produce it is the one that makes the memory.
This may be Oakley’s most practically useful contribution: the feeling of understanding and the fact of remembering are different things, and they often move in opposite directions.
Karpicke, Butler and Roediger (2009) asked students how they studied on their own. Most reread. Very few tested themselves — the more effective method.
Bjork and Bjork (2011) named the phenomenon desirable difficulties: conditions that slow learning down and make it feel worse in the moment often produce more durable memory later. What feels fluent and easy tends to leave nothing behind.
For a language learner the illusion takes a specific form: you read a page of English, understand nearly all of it, and conclude that you know the words on that page. You do not. You recognise them. The gap between those two things is the gap between reading and speaking.
Cepeda et al. (2006) synthesised the distributed-practice literature: for the same total time, spreading sessions apart produces far better retention than massing them together.
Within language specifically, Nakata (2015) tested this for second-language vocabulary and found a clear benefit for spacing — along with something worth knowing: equal intervals worked about as well as expanding ones, though flashcard software usually markets the latter as an advantage.
Oakley recommends mixing problem types rather than finishing one type before starting the next. Rohrer and Taylor (2007) and then Rohrer, Dedrick and Stershic (2015) demonstrated this in mathematics: students who interleaved outperformed those who worked in blocks on later tests.
Scope it correctly. The evidence for interleaving is strongest in mathematics and motor skills; in other domains results are less uniform. And one thing is reliably true: interleaving feels worse while you are doing it. That is exactly why learners abandon it.
Oakley states plainly that the idea of individual learning styles — visual learners, auditory learners — has no basis. She is right.
Pashler et al. (2008) reviewed the literature and concluded that the crucial hypothesis had almost never been tested properly, and that where studies were designed rigorously enough to test it, the results ran against it.
Newton and Salvi (2020) pooled surveys and found that most teachers worldwide still believe it. That someone with 3.6 million learners is willing to contradict so popular a belief deserves acknowledgement.
This is Oakley’s most famous idea: the brain has two modes, a focused one for concentrated work and a diffuse one that operates when you relax and links disparate ideas. The trick is to alternate.
The underlying neuroscience is real. Raichle et al. (2001) described the default mode network — a set of regions that becomes more active when attention is not on an external task.
But this must be said clearly: diffuse mode as a switch you flip in order to learn is a metaphor, not an established mechanism. The default mode network is not a learning mode, and the brain does not have two settings.
What actually supports the practical advice lies in a different literature: incubation. Sio and Ormerod (2009) meta-analysed studies of stepping away from an unsolved problem and returning to it, and found a small but reliable positive effect — larger for open-ended problems, and larger when the break was filled with an undemanding task rather than complete rest.
In other words: the advice is sound; the picture should not be taken literally. When stuck, get up and walk — but not because you have just switched the brain into another mode.
Oakley teaches chunking: binding scattered pieces into a single unit that working memory can handle as one. The concept is solid, and its real foundation is Chase and Simon (1973) on chess players — strong players do not remember more pieces, they see positions.
What needs correcting is the number usually attached. The figure seven plus or minus two comes from Miller (1956) and still circulates everywhere, but Cowan (2001) reviewed the evidence and put the working capacity closer to four chunks. That does not undermine chunking — it makes it more important, because the compartment is smaller than we assume.
For anyone learning Chinese or Japanese, chunking is not abstract theory. A new character is not twelve separate strokes; it is two or three components you already know. Fluent readers do not remember more strokes — they see fewer pieces.
This is the most instructive part of the article, and it is not Oakley’s fault.
Explaining why sleep matters for learning, Oakley cites Xie et al. (2013) in Science: a study in mice reporting that the spaces between brain cells expand during sleep, letting cerebrospinal fluid flush metabolic waste away more quickly. The image of sleep washing the brain clean spread from there.
In 2024, Miao et al. published a contrary result in Nature Neuroscience: by their measurements, clearance of solutes from brain tissue slowed during sleep and under anaesthesia. They argued that earlier work had misread deeper dye penetration as evidence of faster clearance.
The original group and other specialists have objected on methodological grounds, and there have been objections to the objections. The dispute is unresolved. The honest position today is that we do not know enough to say whether sleep plays an important role in clearing waste from the brain.
But here is what matters, and it rescues the advice: sleep’s role in memory consolidation is a separate and much better-established body of evidence, and it never depended on the clearance story at all.
Rasch and Born (2013) review it: during sleep, memory traces are reactivated and reorganised, and depriving someone of sleep after learning impairs the process.
So the advice stands — sleep properly after a vocabulary session — but the correct reason is not to rinse the brain. It is that the session is not over when you close the book; it is over after you sleep.
That is a lesson in method worth more than any study tip: when good advice is attached to an appealing mechanism, check whether the advice survives if the mechanism collapses. This time it does.
Oakley uses the Pomodoro technique — set a timer for twenty-five minutes, work, break. Let us be direct: Pomodoro itself has almost no direct research literature. It is a productivity device from the 1980s, not an experimental finding.
The principle underneath does have support. Sirois and Pychyl (2013) argue that procrastination is largely short-term mood repair: we avoid a task not from laziness but because it evokes something unpleasant, and avoiding it removes the feeling immediately.
That explains why a timer works. It does not increase willpower; it shortens the commitment to a length at which the unpleasant feeling never wins.
For exams Oakley recommends a counterintuitive approach she calls the hard start: begin with the hardest question, work at it for a few minutes, then leave it and move to easier ones, returning later.
The rationale is precisely the incubation effect Sio and Ormerod measured: the hard problem keeps being processed while you work on something else. The critical part is the leaving — sitting on the hard question turns the technique into a disaster.
For JLPT, TOPIK or IELTS the application is concrete: read the longest comprehension passage first, note a few points, move to another section, then come back.
Oakley warns about the Einstellung effect, first described by Luchins (1942): a solution you already know blocks a better one. Bilalić, McLeod and Gobet (2008) tracked chess players’ eye movements and revealed the mechanism precisely — once the familiar option was seen, their eyes kept returning to that region of the board even while they believed they were searching elsewhere.
In language this takes a form everyone meets: the first meaning you learn for a word blocks the rest. Learn 気 as spirit and it takes a long time before 気をつける reads as take care. Learn get as obtain and get it keeps sounding odd.
The remedy is not to try to forget, but to meet the word in enough different contexts that no single sense holds a monopoly — exactly as Nation (2013) and Webb and Nation (2017) describe knowing a word as a process of repeated encounters rather than a single act of memorisation.
Oakley teaches the method of loci — attaching things to be remembered to points in a familiar space. It has genuine evidence behind it: Dresler et al. (2017) trained ordinary people in the technique for six weeks and found both their memory performance and their brain network connectivity shifting toward that of memory champions.
But it is laborious, and impractical across thousands of words.
Learners of Chinese, Japanese and Korean have a far cheaper version built in: the components of a character are already memory hooks. 忙, busy, is the heart radical 忄 beside 亡, to lose — a heart mislaid. You do not need to construct a palace; the structure sits inside the character, and the Sino-Vietnamese layer gives you one more thread to tie it to.
The science in Oakley’s books is knowledge she is relaying from others, and like all popularisation some of it has aged.
But there is one claim she is better placed to make than almost anyone else writing on the subject, and it comes from no laboratory at all.
She was a language specialist, translating Russian on fishing boats in the Bering Sea, who started mathematics again from the beginning in her twenties and became a professor of engineering.
For a thirty-five-year-old wondering whether it is too late to start Japanese, that is a more concrete answer than any citation.
The same article, told in plain words — for younger readers, or for anyone who wants the point quickly.
There is a detail in Barbara Oakley's biography that most introductions skip, and for a language learner it is the most important one.
She began as a linguist.
Enlisting in the US Army straight out of high school, Oakley spent a year studying Russian at the Defense Language Institute, took a degree in Slavic languages, and then worked several fishing seasons as a Russian translator aboard Soviet trawlers in the Bering Sea. After four years as a signal officer she left the army and began studying engineering — in her twenties, having spent her entire education until then avoiding mathematics and science. She is now a professor of engineering.
Learning How to Learn, the online course she teaches with the neuroscientist Terrence Sejnowski, has enrolled more than 3.6 million learners.
One thing needs saying in order to read her properly: Oakley is not a neuroscientist, and has never claimed to be. She is an engineering professor doing the work of popularisation. Most of what she teaches is real cognitive psychology, reported reasonably faithfully. So the job here is not demolition but triage: what holds, what is a teaching metaphor, and what has changed since she wrote it.
Retrieval instead of rereading. Oakley's most-repeated advice is also her best-supported: shut the book and try to recall, rather than read it again. Roediger and Karpicke (2006) showed that self-testing consolidates long-term memory far better than restudying for the same amount of time. Karpicke and Blunt (2011) pitted retrieval against concept mapping — a method widely regarded as deep — and retrieval still won.
For vocabulary this settles how a deck should be built. A card that asks you to recognise a word is comfortable and nearly useless. A card that forces you to produce it is the one that makes the memory.
The illusion of competence. This may be Oakley's most practically useful contribution: the feeling of understanding and the fact of remembering are different things, and they often move in opposite directions.
Karpicke, Butler and Roediger (2009) asked students how they studied on their own. Most reread. Very few tested themselves — the more effective method. Bjork and Bjork (2011) named the phenomenon desirable difficulties: conditions that slow learning down and make it feel worse in the moment often produce more durable memory later.
For a language learner the illusion takes a specific form: you read a page of English, understand nearly all of it, and conclude that you know the words on that page. You do not. You recognise them. The gap between those two things is the gap between reading and speaking.
Spacing. Cepeda et al. (2006): for the same total time, spreading sessions apart produces far better retention than massing them. Within language specifically, Nakata (2015) found a clear benefit for spacing — along with something worth knowing: equal intervals worked about as well as expanding ones, though flashcard software usually markets the latter as an advantage.
Interleaving. Rohrer and Taylor (2007), then Rohrer, Dedrick and Stershic (2015), demonstrated in mathematics that students who interleaved outperformed those who worked in blocks on later tests. Scope it correctly: the evidence is strongest in mathematics and motor skills; elsewhere results are less uniform. And one thing is reliably true — interleaving feels worse while you are doing it. That is exactly why learners abandon it.
Oakley states plainly that the idea of individual learning styles — visual learners, auditory learners — has no basis. She is right.
Pashler et al. (2008) reviewed the literature and concluded that the crucial hypothesis had almost never been tested properly, and that where studies were designed rigorously enough to test it, the results ran against it. Newton and Salvi (2020) pooled surveys and found that most teachers worldwide still believe it.
That someone with 3.6 million learners is willing to contradict so popular a belief deserves acknowledgement.
Focused and diffuse mode. This is Oakley's most famous idea: the brain has two modes, a focused one for concentrated work and a diffuse one that operates when you relax and links disparate ideas. The trick is to alternate.
The underlying neuroscience is real. Raichle et al. (2001) described the default mode network — a set of regions that becomes more active when attention is not on an external task.
But this must be said clearly: diffuse mode as a switch you flip in order to learn is a metaphor, not an established mechanism. The default mode network is not a learning mode, and the brain does not have two settings.
What actually supports the practical advice lies in a different literature: incubation. Sio and Ormerod (2009) meta-analysed studies of stepping away from an unsolved problem and returning to it, and found a small but reliable positive effect — larger for open-ended problems, and larger when the break was filled with an undemanding task rather than complete rest.
In other words: the advice is sound; the picture should not be taken literally. When stuck, get up and walk — but not because you have just switched the brain into another mode.
Chunking. Oakley teaches chunking: binding scattered pieces into a single unit that working memory can handle as one. The concept is solid, and its real foundation is Chase and Simon (1973) on chess players — strong players do not remember more pieces, they see positions.
What needs correcting is the number usually attached. The figure seven plus or minus two comes from Miller (1956) and still circulates everywhere, but Cowan (2001) reviewed the evidence and put the working capacity closer to four chunks. That does not undermine chunking — it makes it more important, because the compartment is smaller than we assume.
For anyone learning Chinese or Japanese, chunking is not abstract theory. A new character is not twelve separate strokes; it is two or three components you already know. Fluent readers do not remember more strokes — they see fewer pieces.
This is the most instructive part, and it is not Oakley's fault.
Explaining why sleep matters for learning, Oakley cites Xie et al. (2013) in Science: a study in mice reporting that the spaces between brain cells expand during sleep, letting cerebrospinal fluid flush metabolic waste away more quickly. The image of sleep washing the brain clean spread from there.
In 2024, Miao et al. published a contrary result in Nature Neuroscience: by their measurements, clearance of solutes from brain tissue slowed during sleep and under anaesthesia. They argued that earlier work had misread deeper dye penetration as evidence of faster clearance.
The original group and other specialists have objected on methodological grounds, and there have been objections to the objections. The dispute is unresolved. The honest position today is that we do not know enough to say whether sleep plays an important role in clearing waste from the brain.
But here is what matters, and it rescues the advice: sleep's role in memory consolidation is a separate and much better-established body of evidence, and it never depended on the clearance story at all. Rasch and Born (2013) review it: during sleep, memory traces are reactivated and reorganised, and depriving someone of sleep after learning impairs the process.
So the advice stands — sleep properly after a vocabulary session — but the correct reason is not to rinse the brain. It is that the session is not over when you close the book; it is over after you sleep.
That is a lesson in method worth more than any study tip: when good advice is attached to an appealing mechanism, check whether the advice survives if the mechanism collapses. This time it does.
Procrastination is about feeling, not laziness. Oakley uses the Pomodoro technique — set a timer for twenty-five minutes, work, break. Let us be direct: Pomodoro itself has almost no direct research literature. It is a productivity device from the 1980s, not an experimental finding.
The principle underneath does have support. Sirois and Pychyl (2013) argue that procrastination is largely short-term mood repair: we avoid a task not from laziness but because it evokes something unpleasant, and avoiding it removes the feeling immediately.
That explains why a timer works. It does not increase willpower; it shortens the commitment to a length at which the unpleasant feeling never wins.
Start hard, then walk away. For exams Oakley recommends the hard start: begin with the hardest question, work at it for a few minutes, then leave it and move to easier ones, returning later. The rationale is precisely the incubation effect. The critical part is the leaving — sitting on the hard question turns the technique into a disaster.
For JLPT, TOPIK or IELTS the application is concrete: read the longest comprehension passage first, note a few points, move to another section, then come back.
The first-meaning trap. Oakley warns about the Einstellung effect, first described by Luchins (1942): a solution you already know blocks a better one. Bilalić, McLeod and Gobet (2008) tracked chess players' eye movements and revealed the mechanism precisely — once the familiar option was seen, their eyes kept returning to that region of the board even while they believed they were searching elsewhere.
In language this takes a form everyone meets: the first meaning you learn for a word blocks the rest. Learn 気 as spirit and it takes a long time before 気をつける reads as take care. Learn get as obtain and get it keeps sounding odd.
The remedy is not to try to forget, but to meet the word in enough different contexts that no single sense holds a monopoly — exactly as Nation (2013) and Webb and Nation (2017) describe knowing a word as a process of repeated encounters rather than a single act of memorisation.
Memory palaces, and the version Chinese-character learners already have. Oakley teaches the method of loci — attaching things to be remembered to points in a familiar space. It has genuine evidence behind it: Dresler et al. (2017) trained ordinary people for six weeks and found both their memory performance and their brain network connectivity shifting toward that of memory champions.
But it is laborious, and impractical across thousands of words.
Learners of Chinese, Japanese and Korean have a far cheaper version built in: the components of a character are already memory hooks. 忙, busy, is the heart radical 忄 beside 亡, to lose — a heart mislaid. You do not need to construct a palace; the structure sits inside the character, and the Sino-Vietnamese layer gives you one more thread to tie it to.
The science in Oakley's books is knowledge she is relaying from others, and like all popularisation some of it has aged.
But there is one claim she is better placed to make than almost anyone else writing on the subject, and it comes from no laboratory at all.
She was a language specialist, translating Russian on fishing boats in the Bering Sea, who started mathematics again from the beginning in her twenties and became a professor of engineering.
For a thirty-five-year-old wondering whether it is too late to start Japanese, that is a more concrete answer than any citation.
These articles summarize well-established research in learning science and linguistics. Key sources and further reading:
The question assumes focus is a muscle you train. The strongest evidence says the training barely carries over, and the...
No time, no talent, too old — those are explanations, not causes. What can be measured is duller: one sentence with a ti...
Not one event but four, running on clocks a billion-fold apart: a wave at 400 milliseconds, a word locked in overnight,...
