Myśli zapisane

This text was written from the perspective of a neurobiologist rather than a personal-development coach. What interests me here is not encouragement but mechanism – and the point at which the evidence stops and my own interpretation begins.

Introduction. The Instrument in Its Case

Almost anyone past forty who has worked with their mind for twenty years owns an object like this. A case on top of the wardrobe. A box of drawing materials. The manual for a camera nobody makes any more. A notebook in a language that once made poetry legible. The object itself means nothing; what means something is that it was once in daily use and then stopped being used, and that the passage from one state to the other happened without a decision. Nobody resolves to give up the guitar. There is simply a week without playing, then a second one, and then you notice that four years have gone by.

What interests me is what happens afterwards. Because usually, after some years, a moment arrives – triggered by a conversation, a recording stumbled upon, somebody’s exhibition – when the whole thing comes back with startling force. For a few hours, sometimes for a few days, the thought of going back is bright, warm and obvious. And then something else arrives: hard to name, soundless, and thoroughly effective. It does not sound like ‘I don’t want to’. It sounds like ‘not now’, ‘not this week’, ‘I’d have to buy new strings first’, ‘and I’ll never play the way I used to anyway’. And usually that second thing wins. We go back to the groove we were already travelling in – to work, to the streaming service, to whatever requires taking nothing up.

The question I want to put here is this: what exactly is this pair, the excitement and the resistance? Are they two forces of the same nature, contending on a single set of scales, or two entirely different processes that happen to meet in the same week? And a second question, more important: what has actually survived of that old skill, and what merely seems to have survived?

I will give away the conclusion now, because I have no intention of building suspense at the reader’s expense. I think the excitement and the resistance are not opponents in a fair fight, because they are not pricing the same object. The excitement concerns an imagined end state – me, playing. The resistance concerns the next hour – me, fumbling at an instrument with fingers that do not remember. These are not two assessments of one thing, so neither can persuade the other. And the trace that really did survive is invisible to both, because it only becomes visible in the doing. That last point is the best documented and least used observation in the whole affair.

So let me begin with the question that looks the most technical and is in fact the most practical: what physically remains in a person after a skill they have not used for ten years.

I. What Actually Remains

1. The Oldest Result in the Psychology of Memory

In the 1880s Hermann Ebbinghaus did something no ethics committee would approve today: for seven months he memorised lists of nonsense syllables and then measured how fast he forgot them. He was the only participant. The forgetting curve we inherit from him is therefore a description of one man, which is worth remembering before invoking it as a law of nature.

More important than the curve, though, is what Ebbinghaus did with material he had judged forgotten. Instead of asking how much he could recall, he learned each list again and counted how many fewer repetitions it took. He called that difference savings. Jan Murre and Joeri Dros, who repeated the entire experiment in 2015 (again with n = 1, seventy lists of 104 syllables, seventy hours of work across seventy-five days), set his figures alongside their own: after a day Ebbinghaus’s savings stood at 0.337, meaning that relearning a list he could no longer reproduce took a third less work than learning it from scratch; after thirty-one days the figure was 0.211. The shape of the curve reproduced with striking fidelity, with one clear divergence at precisely the thirty-day point.

The savings mechanism itself was later confirmed in a way that is still impressive. Thomas Nelson showed in 1978 that savings appear even for items the participant neither recalled nor recognised – that is, exactly where every standard measure of memory reads zero. Colin MacLeod confirmed this in 1988 for words, line drawings and photographs. He also supplied the limitation that popular summaries leave out: for savings to appear, the test after relearning must be a recall test. With a recognition test the effect vanished.

This is the first thing worth taking from the essay. There exists a well-documented class of memory that introspection cannot measure. It does not answer the question ‘do I remember?’, because to that question the answer is no. It answers only the question ‘how much work will this cost me the second time?’ – and that answer cannot be obtained except by starting.

2. Skills Erode More Than We Like to Think

Here I have to spoil the mood, because this is where popular literature says things the data do not support.

The claim that you never forget how to ride a bicycle has taken on a life of its own and gets stretched to cover everything – playing an instrument, a foreign language, drawing. The quantitative review by Winfred Arthur and colleagues from 1998, covering fifty-three papers and 189 independent effect sizes, paints a considerably less comforting picture. After a break of more than a year, the loss of performance corresponded to an effect size of d = −1.4 – more than a full standard deviation. Physical tasks eroded more slowly than cognitive ones (d = −0.75 against −1.15), and speed-based tasks more slowly than accuracy-based ones (−0.33 against −1.00).

The newest and largest meta-analysis in the field – Corey Tatel and Phillip Ackerman, Psychological Bulletin 2025, 1,344 effect sizes from 457 reports – lets us convert this into time. The rate of decay runs at roughly 0.08 standard deviations per month for accuracy measures and 0.06 for both speed-based and mixed measures. In practice that means losing half of a learned skill after about six and a half months of inactivity on the accuracy dimension, after thirteen months on the speed dimension, and after eleven months on mixed measures.

It is also worth recording what did not hold up. The popular distinction according to which continuous skills (swimming, cycling) are more durable than sequential ones made of discrete steps looks persuasive and has support in the older literature. Arthur and colleagues tested that hypothesis directly and obtained the opposite of what they expected: discrete tasks eroded more slowly (d = −0.71) than continuous ones (d = −1.04). The authors themselves did not treat the matter as settled and pointed to a possible confound – the continuous category contained a disproportionate number of cognitive tasks. The honest conclusion is that the difference between ‘the bicycle’ and ‘the procedure’ is less obvious than the proverb claims, and that nobody has yet resolved it properly.

So if someone returns to an instrument after ten years and finds that they play badly – they are right. They really do play badly. This is a measurement, not a distortion of perception or excessive self-criticism.

3. And Yet Something Remains – Just Not What We Were Looking For

The strongest evidence that the past leaves a usable trace comes not from music or sport but from phonetics. In 2009 Jeffrey Bowers, Sven Mattys and Suzanne Gage studied British adults who as small children had been exposed to Hindi or Zulu and had since lost all contact with those languages. None of them remembered a word. They were trained to discriminate subtle sound contrasts that do not exist in English. Participants under forty learned those contrasts selectively – they recovered precisely the contrasts they had encountered in childhood. A control group with no prior exposure did not learn them at all. The caveat one must not omit: participants over forty showed no such effect.

To this can be added two case studies described by Ashleigh Maxcey and colleagues in 2022 – savings after sixty-seven years for word sequences and after twenty-two years for a visual-search skill, with no conscious memory of either material. These are single cases, not population evidence, but they indicate the scale at which the phenomenon has been observed.

And one more caveat, this time theoretical, because it is more honest than the convenient story. Alkis Hadjiosif, Richard Morehead and Maurice Smith demonstrated in 2023 a double dissociation between savings and long-term memory in motor learning: memory components that decay within tens of seconds drive savings but contribute nothing to twenty-four-hour retention, while durable components support long-term memory yet display something like anti-savings. This concerns a narrow visuomotor-adaptation paradigm and should not be generalised to playing the piano. But it is enough to stop us writing that savings ‘prove the memory trace survived’. We do not know what survived. We know the second attempt is cheaper.

Let me compress this into a single sentence, because the rest of the essay follows from it. What remains after an abandoned skill is not competence – competence erodes quickly and measurably. What remains is a reduced cost of acquiring it again. And because that cost is a difference rather than a state, it cannot be felt or estimated without beginning the work. A person sitting in an armchair wondering whether they have ‘still got it’ is asking a question that has no answer in the armchair.

II. Why It Attracts

4. The Decade Memory Treats Differently

There is an asymmetry in what people remember. Memories from roughly the tenth to the thirtieth year of life are systematically over-represented in autobiographical memory – the phenomenon known as the reminiscence bump. David Rubin and Matthew Schulkind described it in 1997, comparing the distribution of memories in people in their twenties and in their seventies.

One has to add straight away what popular versions lose. Where the bump falls depends on how the memories are elicited. The review by Jonathan Koppel and Dorthe Berntsen from 2015 reports that with word cues the bump sits on average between ages 8.7 and 22.5, and with requests for the most important memories between 15.1 and 27.9. The systematic review by Khadeeja Munawar and colleagues from 2018, covering sixty-eight studies drawn from 523 records, gives wider ranges still. Koppel and Rubin go further and suggest we are dealing not with one bump but with at least two different phenomena. This is not a law with sharp boundaries, then, but a well-documented irregularity.

Even in that cautious form, however, something follows from it. Most abandoned passions come from exactly that stretch of life. That is when we were being formed, and autobiographical memory preferentially retains material from the period of formation; being better then has nothing to do with it. That is why the thought of returning to a guitar abandoned at twenty-three has a different temperature from the thought of returning to a spreadsheet course taken in 2014. The first concerns the period out of which the story of who we are is built. The second concerns a Tuesday.

5. Possible Selves and Their Limited Power

Hazel Markus and Paula Nurius gave the name possible selves in 1986 to our conceptions of what we might become, what we would like to become and what we are afraid of becoming. Their article was a theoretical proposal resting on illustrative data from several small student samples – it was not a report of rigorous experiments and should be read accordingly.

Later experimental work, however, said something more important than the mere existence of such images. In the randomised school intervention run by Daphna Oyserman and colleagues in 2006 (N = 264 eighth-graders), the effects of work on possible selves were real but moderate: absences d = 0.27, grades d = 0.25, test pass rates d = 0.36. The authors draw a conclusion that matters for our subject: possible selves do not always sustain self-regulatory action. They begin to work only once they are tied to concrete strategies and to who the person is today.

Put differently: the image of yourself playing the guitar a year from now is pleasant and, on its own, largely inert. That is information about where the bottleneck sits, not a charge against imagination.

6. Excitement Measures Wanting, Not Pleasure

Here comes the distinction I consider the most important in the essay, and the one popular writing most reliably blurs.

Since the 1990s Kent Berridge and Terry Robinson have separated two things usually fused into one: ‘wanting’ (the motivational pull they call incentive salience) and ‘liking’ (the experience of pleasure). The key experiment – rats depleted of dopamine stop working for sugar, while their hedonic facial reactions to its taste remain intact – was conducted in rodents, and that has to be said plainly. In humans the evidence is weaker and indirect: lowering dopamine availability does not reduce reported pleasure from food or from a psychoactive substance, though it does reduce the urge to take it, and in positron emission tomography studies dopamine release correlates more strongly with ratings of ‘how much I want this’ than with ratings of ‘how much I like this’. That is enough to abandon the common description of dopamine as a pleasure chemical, which is simply untrue.

What one must not do is make a second, very common shortcut: wanting in Berridge’s sense is not the same thing as the reward prediction error described by Wolfram Schultz, Peter Dayan and Read Montague in 1997. That result – dopaminergic neurons in awake monkeys respond not to reward but to the difference between reward received and reward expected, and when a promised reward is withheld their activity drops below baseline in precisely the second the reward was due – is among the most frequently replicated in neuroscience. But Berridge devoted a separate paper in 2023 to separating desire from the prediction of outcome value. These are two different frameworks, partly in competition, and welding them into a single sentence is an error.

What does this mean for the instrument in its case? That the wave arriving after you watch somebody else’s concert is a wave of wanting, not a forecast of pleasure. It is a motivational signal attached to an image – and the image is of a finished, competent self. It is not a promise that Tuesday’s forty minutes with a metronome will be enjoyable. It carries no such information. The absence is built into the signal, and one has to know how to work with it.

7. An Illusion Best Not Treated as a Law

There is one more element, tempting and contested. Jordi Quoidbach’s team described in Science in 2013 a phenomenon they called the end of history illusion: people at every age concede that they changed a great deal in the last decade while assuming they will change little in the next. The study covered more than nineteen thousand people aged 18 to 68.

The trouble is that it was cross-sectional – the same people were not followed over time; the recollections of one group were compared with the forecasts of another. When Heather Harris and Michael Busseri tested this in 2019 on a sample of 2,390 adults observed over eighteen years, the result for life satisfaction went rather the other way: people more often expected too much change than too little. A polemical exchange between the two camps ran for another year, and a cross-cultural replication by Brian Haas and Kazufumi Omura in 2022 found the effect markedly stronger in the United States than in Japan.

So I leave it as an image with partial support, not as a law. The image goes like this: someone contemplating a return tends to treat their self of fifteen years ago as a stranger, and their present self as the final version. The first half of that stance makes returning harder, because it treats the old skill as somebody else’s property. The second half makes it harder still, because it assumes nothing further will happen. This is my interpretation, not a research finding.

III. Why It Repels

8. Effort Does Not Just Tire – It Marks Down the Reward

The preference for not doing can be measured, and has been measured repeatedly. In a series of six experiments Wouter Kool and colleagues offered students a choice between two decks of cards, one of which required more frequent switching between rules. The average participant chose the easier deck on roughly two-thirds of trials, even though both took the same time and paid the same. Two caveats are obligatory here: only students were studied, and the effect is a group effect rather than an individual one – in one experiment only seven of sixteen participants showed it, and later attempts to repeat the study with the same instrument sometimes failed.

Stronger support for the underlying claim, that thinking is experienced as a cost, comes from the meta-analysis by Louise David, Eliana Vassena and Erik Bijleveld from 2024: 170 studies, 358 cognitive tasks, 4,670 participants from twenty-nine countries. The greater the mental effort, the greater the reported unpleasantness, with the effect weaker in Asian samples than in European and North American ones. A 2026 comment by Grund, Roelle and Schmid in the same journal narrows what this can be taken to show: the NASA‑TLX item used to index effort registers task difficulty rather than the intensity of thinking, and it is unclear how many of the pooled tasks were undertaken voluntarily.

The most instructive study for our purposes, though, is the 2013 study by Andrew Westbrook, Dascha Kester and Todd Braver. Participants performed an n-back task of increasing difficulty and then chose between repeating an easier version for less money and a harder one for more; the offer was adjusted until they were indifferent. It turned out that the subjective value of the same reward melts systematically as difficulty rises: at the fourth load level, two dollars were worth on average $0.98 to young adults. In twenty-five participants aged 63 to 88 the discount was far steeper – $0.40 – and it persisted even after actual performance was controlled for (area under the discounting curve 0.72 in the young against 0.36 in the older group; the rise in cost with load, measured separately within each group, came to η² = 0.46 against 0.17).

This is, to my mind, the single most important number in the essay. Effort is not subtracted from the reward – it multiplies the reward by a fraction. And that fraction shrinks with age. What this means is that the same prospect of a difficult beginning is priced lower for older people than it was twenty years earlier, with an identical reward at the end and an identical capacity to reach it. Laziness has nothing to do with it.

9. The Groove Is Held by Context

The second pillar of the resistance is more banal, and more effective for being so: the current shape of the day is the default option, and defaults win.

William Samuelson and Richard Zeckhauser named this the status quo bias in 1988. In their studies 486 students received the same decision options, once presented neutrally and once with one option labelled as the current state. That label alone added about seventeen percentage points of support. To the questionnaire data they added field data: in the TIAA-CREF retirement programme between 1981 and 1986, only 28% of participants ever changed their contribution allocation once set, with an annual change rate below 2.5%. Thirty years later a meta-analysis by Jon Jachimowicz and colleagues – fifty-eight datasets, N = 73,675 – put the related default effect at d = 0.68, which for binary outcomes amounts to a shift of some twenty-seven percentage points.

To this must be added the mechanics of habit itself. Wendy Wood and Dennis Rünger describe habit as a default mode of responding, cued by context and operating relatively independently of current intentions. Those same authors stress something popular summaries lose: habits and conscious goal pursuit work synergistically rather than instead of one another – a habit is formed precisely out of repeated goal pursuit and only gradually becomes independent of it. The familiar figure according to which some forty per cent of daily behaviour is habitual comes from the diary studies of Wood, Jeffrey Quinn and Deborah Kashy in 2002: students wore watches that chimed hourly and noted what they were doing; a habit was defined as behaviour performed roughly daily and usually in the same place. On that definition the first study yielded 35% and the second 43%. That is a lot, and it is well documented – but what was measured was frequency and constancy of place, not automaticity, and the participants were students at one university.

The practical conclusion is that the groove is a property of a room, a time of day and an order of operations. Character comes into it much less than we assume. That is good news, and I will come back to it.

10. Not Having Returned Earlier Makes Returning Now Harder

There is one particularly perverse mechanism here, because it grows stronger the longer the abstention lasts.

Orit Tykocinski, Thane Pittman and Erin Tuttle described in 1995 a phenomenon they called inaction inertia: having missed a genuinely good opportunity, we are less likely to take up a worse one that appears later, even though the later one is still worth taking on its own terms. A preregistered replication run by Jiayu Chen’s team in 2021 on 1,555 participants confirmed the effect, but conditionally: it was clear for high-stakes decisions (the car scenario, d between −0.57 and −0.68; the ski scenario between −0.18 and −0.67) and shrank to zero for small ones. The authors recommend that further work use only the scenarios in which the effect held.

Applying this to an abandoned passion is my interpretation rather than a research finding – but it seems warranted, because we are talking about a decision with a high personal stake. Someone who could have returned to drawing five years ago now has a worse version of the same opportunity at their disposal: less time, more obligations, more lost facility. And it is precisely the comparison with that better, missed version that makes today’s version look not worth the trouble – even though, had it come first, it would have been taken without hesitation.

11. The Real Weight: The Rival Is Your Own Former Version

Everything above concerns costs. But the heaviest part of the resistance has, in my view, to do with identity.

E. Tory Higgins described in 1987 the mechanics he called self-discrepancy: we suffer not from who we are but from the distance between the actual self and the ideal or ought self. The general claim has solid support – the meta-analysis by Tyler Mason and colleagues from 2019, covering seventy studies and ninety-four samples, found moderate associations between self-discrepancy and depression (r = 0.27) and anxiety (r = 0.23). Higgins, however, claimed something more specific: that discrepancy from the ideal self produces dejection, and discrepancy from the ought self produces agitation. That part the data did not confirm; both kinds of discrepancy relate to both outcomes, and actual–ideal discrepancy proved the stronger of the two in each case.

For the returner this has a consequence the beginner does not face. The beginner compares themselves with zero, and every step forward is a gain. The returner compares themselves with their own best version – and every step forward is a reminder of the distance. The same thirty minutes of practice hands two people opposite feedback. Mental toughness has nothing to do with it; the difference lies in their reference points, and the reference point is what determines the sign of the result.

Loss aversion can be added here, though with care, since the concept has been through a decade of fire. David Gal and Derek Rucker announced ‘the loss of loss aversion’ in 2018, arguing that the classic paradigms confuse loss with inaction. The reply from Kellen Mrkva and colleagues in 2020, resting on five samples and 17,720 participants, showed that the phenomenon has moderators – it weakens with knowledge and experience, and strengthens with age – but appears at every level of them. The broadest quantitative basis so far comes from the meta-analysis by Alexander Brown and colleagues in 2024: 607 estimates from 150 studies, a coefficient of λ = 1.955 with a narrow confidence interval [1.820, 2.102]. Yechiam and Zeif re-analysed that same body of estimates in 2025 and reported that with symmetrical stakes and no ordering of items λ falls to about 1.07, no longer differing significantly from 1. The dispute is therefore still open.

My interpretation runs like this. As long as the instrument stays in its case, a person owns the memory of having been good. That memory is unassailable because it is untestable. The first hour of practice is the only event that can invalidate it. The resistance is not protecting us from effort – we accept effort readily enough elsewhere in life. The resistance is protecting a particular record in memory from contact with a measurement.

IV. Why the Resistance Usually Wins

12. A Gap That Deserves More Airtime

Paschal Sheeran and Thomas Webb summarised in 2016 what is known about the translation of intention into action with a figure that ought to be better known. They cite a synthesis of ten earlier meta-analyses, covering 422 studies in total, which produced a weighted correlation between intention and subsequent behaviour of r = 0.53 – about 28% of the variance explained. Close to three-quarters of what we do happens outside what we decided.

More eloquent still is their earlier experimental meta-analysis: across forty-seven tests, interventions that changed intentions by a medium-to-large amount (d = 0.66) changed actual behaviour only by a small-to-medium amount (d = 0.36). Sheeran and Webb use the term ‘inclined abstainers’ for people who intend and do not act.

These numbers describe how the mechanism is built; no reproach is intended.

13. Two Valuations That Never Meet

Here I arrive at the thesis I announced at the start. It is an interpretation rather than a research finding, but one drawn from the evidence set out above and testable against one’s own experience.

The excitement we feel when we think about returning is a wanting signal attached to an imagined end state. Its reach is the reach of imagination, not of the calendar. The resistance, by contrast, is a valuation of one particular, imminent hour – and that valuation takes into account the effort (which multiplies the reward by a fraction), the cost of departing from the default option (seventeen to twenty-seven percentage points), the comparison with a better opportunity already missed, and finally the risk to identity.

The two assessments have different objects, so they can neither persuade each other nor be summed. The excitement has no access to the information the resistance holds, and the resistance does not price what the excitement can see. What looks like an argument is two parties talking about two different things in parallel. And because what actually happens on Wednesday at eight is decided by the valuation of Wednesday and not by the valuation of an imagined year, the resistance wins. Nearly always.

And one further consequence, this time straight out of Part I. Both valuations are blind to the one quantity that would genuinely change the arithmetic: the savings. The fact that the second attempt is cheaper than the first. That information exists in neither of them, because by definition it comes into being only in the course of relearning. The person deliberating in the armchair is therefore keeping an account with the most important entry missing – and has no way of supplying it other than by standing up.

V. What Follows

I do not intend to supply a numbered list of techniques, because most such lists in this field have weaker support than their form suggests. I will instead offer five things that do follow from the evidence above, together with an honest statement of how firm their foundations are.

14. Stop Asking Whether You Have Still Got It

This follows directly from Part I and is the best-documented claim in the essay. Savings are a difference in the cost of relearning, not a state available to introspection. Nelson showed that they appear even where the participant does not recognise the material. Bowers showed that childhood exposure to a language confers a selective advantage in people who remember nothing of it.

The consequence is simple and uncomfortable: the question ‘can I still do this?’ is badly posed, because its answer comes into existence only in the doing. A conviction on this matter – the belief that you can, and the certainty that you no longer can, alike – is worthless as data.

15. Do Not Wait for the Urge, but Do Not Believe the Slogan That Action Creates It Either

Behavioural activation – a therapeutic approach in which one first plans and performs specific activities and only then waits for the inclination – is a well-documented method. The meta-analysis by David Ekers and colleagues from 2014 (twenty-six randomised trials, 1,524 participants) produced an SMD of −0.74 against control conditions, and the British COBRA trial run by David Richards and colleagues, on 440 patients, showed the method to be equivalent to cognitive behavioural therapy and considerably cheaper.

It is tempting to extract a general maxim from this: that action precedes motivation. Here one has to stop. The systematic review of mediators by Noortje Janssen and colleagues from 2021 showed that the evidence for the mechanism itself is weak: of eight studies testing whether increased activity is what accounts for the improvement, only two confirmed it, and the largest and methodologically strongest did not. We know it works in the treatment of depression. We do not know well why, and we have no grounds for extending it into a general psychological law covering healthy people and their hobbies.

The honest version of the recommendation is therefore more modest: waiting for the urge is a strategy of documented low effectiveness, and starting without it is a practice that works in a clinical population for reasons not yet established. That is enough not to wait. It is not enough to promise that enthusiasm will arrive after the first ten minutes.

16. Change the Context, Not the Character

If the groove is a property of a room and a time, then the logical move is to attack the room and the time.

Bas Verplanken and colleagues named this, in 2008, the habit discontinuity hypothesis: when the context breaks – a house move, a change of office, a change in the rhythm of the day – the old cues stop firing and for a while we genuinely decide again instead of simply doing. The strongest evidence comes from the randomised field experiment by Verplanken and Deborah Roy in 2016 on 800 participants: the same intervention worked on people who had recently moved house and did not work on those who had stayed put, with the optimal window estimated at around three months after the change. Data from a British panel study of 18,053 adults point the same way: the influence of stated values on transport choice was strongest in the first month after a move (a difference of thirteen percentage points) and melted to five points after ten years.

Two honest caveats. The effects are real but small – shifts of a few percentage points, at most into the low teens, not a breakthrough. And the mechanism is unsettled: in a study of students by Paul Haggar and colleagues, what mattered was prior planning of the new route rather than value activation or the change of surroundings as such.

In practice: a return to an old passion stands a better chance if it is folded into a moment when the context is changing anyway, and if the place where the activity happens changes too. The same goes for the tool. An instrument taken out of its case and set on a stand in plain sight stops being an object that requires a decision.

It is worth adding here the one planning mechanism with decent support – and giving its current rather than its archival magnitude. Plans of the form ‘if situation X, then I will do Y’ were valued at d = 0.65 in Peter Gollwitzer and Paschal Sheeran’s 2006 meta-analysis of ninety-four studies. That figure still circulates in popular books and is out of date. Sheeran and Gollwitzer returned to the question in 2025, now with Listrom as a third author, in a meta-analysis covering 642 tests: the effect came out at d = 0.36, and after Bayesian correction for publication bias it fell to d = 0.15, with very strong evidence of selective publication. If–then planning still works – more clearly in the laboratory (d = 0.49) than in the field (d = 0.27), and most strongly in people who are already motivated (d = 0.79). It is a small, real improvement for someone who already wants to, not a lever that manufactures wanting.

17. Separate Returning from Regaining Your Old Level

This follows from the section on self-discrepancy and is the advice I regard as the most useful in practice, though it has the weakest direct support – I know of no study that tests it head-on.

If the returner’s burden is comparison with their own best version, then the sensible move is to redefine the activity so that the old version stops being the reference point. The way to do that is to enter an adjacent territory in which one has no history, without pretending it never existed. Someone who played classical can come back through the blues. Someone who drew from life in pencil can come back through a travel sketchbook filled in ballpoint. Someone who shot in a studio can come back with a single prime lens on the street. The activity stays the same in the sense that matters for skill – so the savings apply. What changes is what the result is compared against, and that is the only variable determining the sign of the feedback.

To this can be added a warning from the research on passion. Robert Vallerand and colleagues proposed in 2003 a distinction between harmonious passion, which one chooses, and obsessive passion, which chooses us. The meta-analysis by Thomas Curran and colleagues from 2015 – ninety-four studies, 1,308 effect coefficients – bore out the picture: harmonious passion is associated with flow (r = 0.48) and protects against burnout (r = −0.47), while obsessive passion relates weakly to both and more clearly to negative affect. These are correlational data and should be read as such. But they suggest something worth bearing in mind: a return converted into a project with a schedule, a target and an accounting stands a fair chance of recreating exactly the conditions under which the passion was abandoned the first time.

A related observation comes from research on mere exposure. Robert Zajonc showed in 1968 that repetition alone increases liking; Robert Bornstein’s 1989 meta-analysis (134 articles, 208 independent results) put the effect at r = 0.26. Bornstein already pointed out that the relationship is not linear: with satiation and boredom it takes the shape of an inverted U. The meta-analysis by Matthew Montoya and colleagues from 2017 confirmed that shape parametrically, on 268 curves from 81 articles. Liking rises, peaks, and then falls with excessive repetition. For the returner this means that an intensive burst is a worse strategy than regular, moderate contact – not for reasons of discipline, but because of the shape of the curve.

18. Do Not Justify the Return by Cognitive Benefits

Finally, the matter on which popular writing diverges most sharply from the data, and which people readily invoke as a self-sufficient argument: that learning new things ‘sharpens the mind’.

The most comprehensive review of cognitive training – Daniel Simons and colleagues, Psychological Science in the Public Interest 2016 – arranges the findings in a very clear gradient. Evidence that practising a given task makes you better at that task is abundant. Evidence that this transfers to closely related tasks is thinner. Evidence that it transfers to distantly related tasks or to everyday cognitive functioning is close to non-existent. The authors add that none of the studies reviewed fully met the standards of good methodological practice.

The same applies to exercise. The umbrella review by Luis Ciria and colleagues in Nature Human Behaviour in 2023 gathered twenty-four meta-analyses of randomised trials in healthy populations and recalculated them: the raw effect was d = 0.22, dropping to 0.13 once active control groups were taken into account, and to 0.05 after correction for publication bias, with a confidence interval spanning zero. A year later Olivier Dupuy and colleagues replied that the review has methodological and theoretical flaws of its own. The dispute continues.

The closest thing to evidence that demanding learning yields something cognitively is Denise Park’s Synapse Project from 2014. Two hundred and twenty-one people aged 60 to 90 spent three months learning digital photography or quilting, at an average of 16.51 hours a week. Improvement showed up in one thing – episodic memory – and more clearly in the photography group (d = 0.54) than in the others; in the quilting group the effect was not significant. Processing speed and visuospatial ability did not move. The study has not since been replicated. The group that simply met socially gained nothing cognitively – so if anything is working here, it is the difficulty rather than the activity.

And wellbeing? The largest study of the question – Hei Wan Mak and colleagues, Nature Medicine 2023 – covered 93,263 people over sixty-five in sixteen countries, in a model where each participant serves as their own control over time. In periods when people had a hobby, they reported fewer depressive symptoms (coefficient −0.10) and higher life satisfaction (0.10). The authors state plainly that with observational data causality cannot be established even with the most sophisticated modelling.

The conclusion, to my mind, is a liberating one. There are no good grounds for returning to drawing in order to have a better memory. There are entirely sufficient grounds for returning to drawing because one wants to draw. It is the only justification that needs no footnote – and the only one that will not collapse when the next meta-analysis appears three years from now.

Conclusion. An Account with an Entry Missing

I wrote this essay convinced that more consolation than knowledge circulates on the subject of returning to abandoned passions, and that consolation is a worse companion here than precision. Precision says uncomfortable things. That after ten years you really do play badly, rather than merely feeling that you do. That the same difficulty is priced worse with age, measurably so. That the resistance has on its side the architecture of the day, the arithmetic of effort, and the protection of a memory in which we were good.

And yet that same precision leads to a conclusion that is neither comforting nor discouraging, only practical. All these valuations – the warm one and the cold one alike – are produced in an armchair. And the one quantity that could overturn them is not produced in an armchair at all. Savings are not a state one can find out about. They are a difference that appears in the course of the work, and only there.

Which means that deliberation on this particular question is structurally defective – not because we are weak, but because we are keeping an account with the most important entry missing, and there is no way of supplying it without getting up. There is no wisdom here to be acquired. There is a piece of information to be fetched, and the only place it is lying is the first awkward hour.

The instrument in its case is not a keepsake. It is an unresolved measurement.

A Note on Sources

A few words about what the essay above rests on, because in this field the difference between kinds of evidence is often the difference between a conclusion and a misunderstanding. Where I write about the state of knowledge, a meta-analysis or a systematic review stands behind it. Where I refer to a single study – the work on effort discounting, say, or the Synapse Project – I give the sample size and the population, because both matter. Classic results are given in their current rather than their archival form: I price the if–then planning effect at d = 0.36 on the 2025 meta-analysis rather than at d = 0.65 on the 2006 one, because the latter figure, still ubiquitous in popular books, did not survive correction for publication bias.

Several claims that circulate in popular accounts as settled I have left with the disagreement marked: the end of history illusion, loss aversion as a general principle, the cognitive benefits of physical exercise, and the distinction between continuous and sequential skills. That is what the state of knowledge looks like today, and I see no reason to smooth it over. I have also kept animal-model data consistently separate from human data – this matters especially for the wanting–liking distinction, whose evidential core consists of experiments in rodents. Passages marked as my interpretation are interpretation, and no study tests them directly.

Every work cited appears in Sources with a full reference, so that anyone can check whether it says what I have taken it to say.

A note of caution: this essay is not medical or psychotherapeutic advice. The passages on behavioural activation describe the state of research on a method used in the treatment of depression, not recommendations for any particular person.

Sources

Arthur, W., Jr., Bennett, W., Jr., Stanush, P. L., & McNelly, T. L. (1998). Factors that influence skill decay and retention: A quantitative review and analysis. Human Performance, 11(1), 57–101.

Berridge, K. C. (2023). Separating desire from prediction of outcome value. Trends in Cognitive Sciences, 27(10), 932–946.

Berridge, K. C., & Robinson, T. E. (2016). Liking, wanting, and the incentive-sensitization theory of addiction. American Psychologist, 71(8), 670–679.

Bornstein, R. F. (1989). Exposure and affect: Overview and meta-analysis of research, 1968–1987. Psychological Bulletin, 106(2), 265–289.

Bowers, J. S., Mattys, S. L., & Gage, S. H. (2009). Preserved implicit knowledge of a forgotten childhood language. Psychological Science, 20(9), 1064–1069.

Brown, A. L., Imai, T., Vieider, F. M., & Camerer, C. F. (2024). Meta-analysis of empirical estimates of loss aversion. Journal of Economic Literature, 62(2), 485–516.

Chen, J., Hui, L. S., Yu, T., Feldman, G., et al. (2021). Foregone opportunities and choosing not to act: Replications of inaction inertia effect. Social Psychological and Personality Science, 12(3), 333–345.

Ciria, L. F., Román-Caballero, R., Vadillo, M. A., Holgado, D., Luque-Casado, A., Perakakis, P., & Sanabria, D. (2023). An umbrella review of randomized control trials on the effects of physical exercise on cognition. Nature Human Behaviour, 7(6), 928–941.

Curran, T., Hill, A. P., Appleton, P. R., Vallerand, R. J., & Standage, M. (2015). The psychology of passion: A meta-analytical review of a decade of research on intrapersonal outcomes. Motivation and Emotion, 39(5), 631–655.

David, L., Vassena, E., & Bijleveld, E. (2024). The unpleasantness of thinking: A meta-analytic review of the association between mental effort and negative affect. Psychological Bulletin, 150(9), 1070–1093.

Dupuy, O., Ludyga, S., Ortega, F. B., et al. (2024). Do not underestimate the cognitive benefits of exercise. Nature Human Behaviour, 8(8), 1460–1463.

Ebbinghaus, H. (1885). Über das Gedächtnis: Untersuchungen zur experimentellen Psychologie [On Memory: A Contribution to Experimental Psychology]. Leipzig: Duncker & Humblot.

Ekers, D., Webster, L., Van Straten, A., Cuijpers, P., Richards, D., & Gilbody, S. (2014). Behavioural activation for depression: An update of meta-analysis of effectiveness and sub group analysis. PLoS ONE, 9(6), e100100.

Gal, D., & Rucker, D. D. (2018). The loss of loss aversion: Will it loom larger than its gain? Journal of Consumer Psychology, 28(3), 497–516.

Gollwitzer, P. M., & Sheeran, P. (2006). Implementation intentions and goal achievement: A meta-analysis of effects and processes. Advances in Experimental Social Psychology, 38, 69–119.

Grund, A., Roelle, J., & Schmid, S. (2026). Is thinking inherently unpleasant? A cautionary note on David, Vassena, and Bijleveld (2024). Psychological Bulletin, 152(2), 201–206.

Haas, B. W., & Omura, K. (2022). Cultural differences in susceptibility to the end of history illusion. Personality and Social Psychology Bulletin, 48(9), 1331–1348.

Hadjiosif, A. M., Morehead, J. R., & Smith, M. A. (2023). A double dissociation between savings and long-term memory in motor learning. PLOS Biology, 21(4), e3001799.

Haggar, P., Whitmarsh, L., & Skippon, S. M. (2019). Habit discontinuity and student travel mode choice. Transportation Research Part F: Traffic Psychology and Behaviour, 64, 1–13.

Harris, H., & Busseri, M. A. (2019). Is there an ‘end of history illusion’ for life satisfaction? Evidence from a three-wave longitudinal study. Journal of Research in Personality, 83, 103869.

Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94(3), 319–340.

Jachimowicz, J. M., Duncan, S., Weber, E. U., & Johnson, E. J. (2019). When and why defaults influence decisions: A meta-analysis of default effects. Behavioural Public Policy, 3(2), 159–186.

Janssen, N. P., Hendriks, G.-J., Baranelli, C. T., Lucassen, P., Oude Voshaar, R., Spijker, J., & Huibers, M. J. H. (2021). How does behavioural activation work? A systematic review of the evidence on potential mediators. Psychotherapy and Psychosomatics, 90(2), 85–93.

Kool, W., McGuire, J. T., Rosen, Z. B., & Botvinick, M. M. (2010). Decision making and the avoidance of cognitive demand. Journal of Experimental Psychology: General, 139(4), 665–682.

Koppel, J., & Berntsen, D. (2015). The peaks of life: The differential temporal locations of the reminiscence bump across disparate cueing methods. Journal of Applied Research in Memory and Cognition, 4(1), 66–80.

Koppel, J., & Rubin, D. C. (2016). Recent advances in understanding the reminiscence bump: The importance of cues in guiding recall from autobiographical memory. Current Directions in Psychological Science, 25(2), 135–140.

MacLeod, C. M. (1988). Forgotten but not gone: Savings for pictures and words in long-term memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 14(2), 195–212.

Mak, H. W., Noguchi, T., Bone, J. K., Wels, J., Gao, Q., Kondo, K., Saito, T., & Fancourt, D. (2023). Hobby engagement and mental wellbeing among people aged 65 years and older in 16 countries. Nature Medicine, 29(9), 2233–2240.

Markus, H., & Nurius, P. (1986). Possible selves. American Psychologist, 41(9), 954–969.

Mason, T. B., Smith, K. E., Engwall, A., Lass, A., Mead, M., Sorby, M., Bjorlie, K., Strauman, T. J., & Wonderlich, S. (2019). Self-discrepancy theory as a transdiagnostic framework: A meta-analysis of self-discrepancy and psychopathology. Psychological Bulletin, 145(4), 372–389.

Maxcey, A. M., Shiffrin, R. M., Cousineau, D., & Atkinson, R. C. (2022). Two case studies of very long-term retention. Psychonomic Bulletin & Review, 29(2), 563–567.

Montoya, R. M., Horton, R. S., Vevea, J. L., Citkowicz, M., & Lauber, E. A. (2017). A re-examination of the mere exposure effect: The influence of repeated exposure on recognition, familiarity, and liking. Psychological Bulletin, 143(5), 459–498.

Mrkva, K., Johnson, E. J., Gächter, S., & Herrmann, A. (2020). Moderating loss aversion: Loss aversion has moderators, but reports of its death are greatly exaggerated. Journal of Consumer Psychology, 30(3), 407–428.

Munawar, K., Kuhn, S. K., & Haque, S. (2018). Understanding the reminiscence bump: A systematic review. PLOS ONE, 13(12), e0208595.

Murre, J. M. J., & Dros, J. (2015). Replication and analysis of Ebbinghaus’ forgetting curve. PLOS ONE, 10(7), e0120644.

Nelson, T. O. (1978). Detecting small amounts of information in memory: Savings for nonrecognized items. Journal of Experimental Psychology: Human Learning and Memory, 4, 453–468.

Oyserman, D., Bybee, D., & Terry, K. (2006). Possible selves and academic outcomes: How and when possible selves impel action. Journal of Personality and Social Psychology, 91(1), 188–204.

Park, D. C., Lodi-Smith, J., Drew, L., Haber, S., Hebrank, A., Bischof, G. N., & Aamodt, W. (2014). The impact of sustained engagement on cognitive function in older adults: The Synapse Project. Psychological Science, 25(1), 103–112.

Quoidbach, J., Gilbert, D. T., & Wilson, T. D. (2013). The end of history illusion. Science, 339(6115), 96–98.

Richards, D. A., Ekers, D., McMillan, D., Taylor, R. S., Byford, S., Warren, F. C., et al. (2016). Cost and outcome of behavioural activation versus cognitive behavioural therapy for depression (COBRA): A randomised, controlled, non-inferiority trial. The Lancet, 388(10047), 871–880.

Rubin, D. C., & Schulkind, M. D. (1997). The distribution of autobiographical memories across the lifespan. Memory & Cognition, 25(6), 859–866.

Samuelson, W., & Zeckhauser, R. (1988). Status quo bias in decision making. Journal of Risk and Uncertainty, 1(1), 7–59.

Schultz, W., Dayan, P., & Montague, P. R. (1997). A neural substrate of prediction and reward. Science, 275(5306), 1593–1599.

Sheeran, P., Listrom, O., & Gollwitzer, P. M. (2025). The when and how of planning: Meta-analysis of the scope and components of implementation intentions in 642 tests. European Review of Social Psychology, 36(1), 162–194.

Sheeran, P., & Webb, T. L. (2016). The intention–behavior gap. Social and Personality Psychology Compass, 10(9), 503–518.

Simons, D. J., Boot, W. R., Charness, N., Gathercole, S. E., Chabris, C. F., Hambrick, D. Z., & Stine-Morrow, E. A. L. (2016). Do ‘brain-training’ programs work? Psychological Science in the Public Interest, 17(3), 103–186.

Tatel, C. E., & Ackerman, P. L. (2025). Procedural skill retention and decay: A meta-analytic review. Psychological Bulletin, 151(6), 696–736.

Thomas, G. O., Poortinga, W., & Sautkina, E. (2016). Habit discontinuity, self-activation, and the diminishing influence of context change: Evidence from the UK Understanding Society Survey. PLoS ONE, 11(4), e0153490.

Tykocinski, O. E., Pittman, T. S., & Tuttle, E. E. (1995). Inaction inertia: Foregoing future benefits as a result of an initial failure to act. Journal of Personality and Social Psychology, 68(5), 793–803.

Vallerand, R. J., Blanchard, C., Mageau, G. A., Koestner, R., Ratelle, C., Léonard, M., Gagné, M., & Marsolais, J. (2003). Les passions de l’âme: On obsessive and harmonious passion. Journal of Personality and Social Psychology, 85(4), 756–767.

Verplanken, B., & Roy, D. (2016). Empowering interventions to promote sustainable lifestyles: Testing the habit discontinuity hypothesis in a field experiment. Journal of Environmental Psychology, 45, 127–134.

Verplanken, B., Walker, I., Davis, A., & Jurasek, M. (2008). Context change and travel mode choice: Combining the habit discontinuity and self-activation hypotheses. Journal of Environmental Psychology, 28(2), 121–127.

Webb, T. L., & Sheeran, P. (2006). Does changing behavioral intentions engender behavior change? A meta-analysis of the experimental evidence. Psychological Bulletin, 132(2), 249–268.

Westbrook, A., Kester, D., & Braver, T. S. (2013). What is the subjective cost of cognitive effort? Load, trait, and aging effects revealed by economic preference. PLoS ONE, 8(7), e68210.

Wood, W., Quinn, J. M., & Kashy, D. A. (2002). Habits in everyday life: Thought, emotion, and action. Journal of Personality and Social Psychology, 83(6), 1281–1297.

Wood, W., & Rünger, D. (2016). Psychology of habit. Annual Review of Psychology, 67, 289–314.

Yechiam, E., & Zeif, D. (2025). Loss aversion is not robust: A re-meta-analysis. Journal of Economic Psychology, 107, 102801.

Zajonc, R. B. (1968). Attitudinal effects of mere exposure. Journal of Personality and Social Psychology, 9(2, Pt. 2), 1–27.