Self-Determination Theory

What it is

Motivation is not one substance in different amounts. That is the move self-determination theory makes, and everything else follows from it. Two people can be equally motivated to run four mornings a week and be running on different machinery, one because a doctor has frightened them and one because they have started to think of themselves as somebody who runs, and the second arrangement will still be there in September. SDT is the attempt to say what the difference is made of.

Edward Deci and Richard Ryan set it out in Intrinsic Motivation and Self-Determination in Human Behavior (Plenum Press, 1985), building on experimental work through the 1970s and extending it across education, health, work and sport for four decades since. Their claim is that three things are needs rather than preferences. Autonomy is acting in line with your own values, experienced as endorsing what you are doing rather than being moved through it. Competence is the felt sense of getting better at something. Relatedness is connection to other people that carries some weight. Thwart any of the three and motivation degrades, no matter how much incentive is stacked on top.

The corollary is the part that surprises people. A reward can reduce motivation, if it is experienced as control. Nir Eyal's account of Quora's Views feature is a compact illustration: a metric switched on for users rather than offered to them, which removed a choice people had not realised they valued until it was gone. Reactance is what fires at that moment, and what it is defending is the same need that sustains motivation in the first place. Rewards that are offered leave autonomy intact. Rewards that are imposed spend it.

A word about how this page states its own evidence, since the tag and the research are separate questions. Our vault carried this note as unverified for months, which reads like a verdict on self-determination theory and is not one. In this publication's vocabulary unverified means nobody here had opened the paper, and in this case what had gone untraced was a figure our own note had picked up from a trade book. The theory was never in doubt. The sourcing was. The note was checked on 1 September 2026 and now carries robust, with one caveat and one deletion, both below.

The caveat is autonomy. Decades of converging evidence support the three-need structure across education, health, exercise and work, but whether autonomy is a universal human need or a description of what motivation looks like in individualist societies is genuinely open, and the argument is live in the primary literature rather than settled in a footnote. Competence is the least disputed of the three. Anyone building on SDT in a cross-cultural setting should treat the autonomy leg as the load-bearing assumption it is.

The deletion is ours. Our note used to state that the twenty per cent of gym joiners who keep showing up have either built identity-based intrinsic motivation or found a context that satisfies all three needs. A targeted search found nothing matching that claim in any study. The nearest real figure is a gym-industry statistic about how often members attend, which measures attendance frequency rather than motivational cause and comes from retention marketing. The sentence may still be a reasonable inference. It is not evidence, and it is not repeated here as one. The same discipline applies to the phrase in our own note that intrinsic motivation outlasts extrinsic motivation "by orders of magnitude": that is a formulation, not a measurement, and no effect size sits behind it.

In effect

Research: Teixeira's review, and why exercise is the field's best test case

The most useful application of SDT to a real behaviour is Pedro Teixeira and colleagues' systematic review of exercise and physical activity, published in the International Journal of Behavioral Nutrition and Physical Activity in 2012, with Richard Ryan among the authors. Exercise is a good testing ground because the failure is so visible. People sign up in numbers, know what they should do, and stop. Reviewing that literature through the three needs reorganises it: what predicts someone still training a year later is not how badly they wanted to start.

Our own reporting used to lean hardest on the idea that relatedness is the leg fitness culture under-serves: autonomy and competence are what the industry sells, personalised programmes and visible progress, and relatedness is left to happen by accident among strangers in headphones.

Corrected 11 September 2026, and the correction runs the other way. In the exercise evidence relatedness is the leg that holds up least. Nikos Ntoumanis and colleagues (2021), Health Psychology Review 15(2), 214 to 244, found SDT-informed interventions moved autonomy (g = 0.37) and competence (g = 0.31) but not relatedness (g = 0.202, not significant, falling to g = 0.027 at follow-up). Marlene Kritz and colleagues (2026), Nature Human Behaviour 10(6), 1109 to 1121, pooled 71 studies and 22,042 participants and found group against individual exercise null, g = 0.086, p = 0.249.

There is a serious argument on the other side, and it belongs to the same researchers. Ntoumanis and Pedro Teixeira both hold that the weak relatedness result may reflect how thinly relatedness is measured, and how rarely interventions target it directly, rather than showing it does not matter. That is a reasonable reading of why the number is small, and we keep it open.

Where we land: generic belonging is the weak thing, and identification with a specific group is what survives. Xiaoxue Gao and colleagues (2026), Behavioral Sciences 16(6), 882, found group identification, social attraction and social integration predicted adherence in adult group fitness classes where broad cohesion scores did not, behind self-efficacy. Our article on group fitness now runs on that narrower finding.

Everyday life: the January joiner, and the Wednesday-night fixture

Set two cases side by side. The January gym joiner has a goal weight, a deadline in the form of a holiday, and a programme somebody else designed. Autonomy is low, because the plan is not theirs. Competence is fragile, because they are new and everything is hard in public. Relatedness is absent. All three needs are unmet at once, and by March the motivation that was doing the work, which came from outside, has been withdrawn.

Then the men in their fifties who have played the same Wednesday-night football fixture for thirty years. Nobody is running a discipline challenge. The match is not in a calendar app. They come back because the team would notice if they did not, because the dressing room is a place where part of them is still alive, and because the two beers afterwards are part of the evening. Relatedness is not a supplement to that arrangement. It is the arrangement. Read as the source of the difference in who lasts, though, that is an illustration rather than a finding, and the correction above applies to it: what the football fixture supplies is not company but a specific group that a specific person belongs to. What the first case has to manufacture, the second has by default, which is the whole argument of identity-based habits and of belonging approached from the motivational side.

Commerce and product design: the needs, sold back

SDT has been absorbed by the engagement industry more completely than almost any other academic framework, and in two directions at once. In Hooked, Eyal grounds his third category of variable reward, rewards of the self, in Deci and Ryan's work on competency: the pull of inbox zero, of levelling up, of a completed streak, sought for its own sake rather than for anything it buys. That is the competence need turned into a retention mechanic. Our own audit of that book records the honest limit, which is that the three-reward taxonomy is a taxonomy Eyal proposes, each type anchored to a real literature that was not built to support the split.

In Indistractable, five years later, the same author uses the same theory to argue the opposite case. Autonomy, competence and relatedness become psychological nutrients, and Ryan's need density hypothesis supplies the mechanism: the more of those needs go unmet in life, the more they get met in virtual realities. A child playing eleven hours of a game is described as a child whose needs are being served somewhere, rather than as a child with a screen problem. Ryan blurbed the book, which is worth knowing and does not make the reading wrong.

What it does not say

It does not say the three needs are equally settled. Competence is the least disputed. Autonomy is the one with a live argument attached, on whether it is a universal need or a Western individualist framing, and any application that assumes the first should say so.

It does not say twenty per cent of gym joiners persist because of SDT. That sentence was in our own note and has been removed from citable status. No study supports it, and the industry number nearest to it measures something else.

It does not license a magnitude. "Intrinsic motivation outlasts extrinsic motivation" is supported. Any specific multiplier attached to that sentence, including the one our note used to carry, has no measurement behind it.

It does not transfer its evidence to everything built on it. SDT being robust does not make the rewards-of-the-self taxonomy robust, or the need density hypothesis, or any product that has read the theory and is quoting it in a pitch deck. Each of those borrows the vocabulary. Only the core theory carries the evidence.

It is not a treatment. Motivational interviewing is the clinical method that comes closest to operationalising autonomous motivation, and it is a practitioner discipline with its own evidence record and its own limits, not something a reader applies to themselves after reading a theory.

Who developed it

Edward L. Deci and Richard M. Ryan, jointly, in Intrinsic Motivation and Self-Determination in Human Behavior (1985), extended across decades of work since.

Our vault credited Deci alone until 1 September 2026. That correction belongs on the page rather than in a changelog, because it is the tenth instance of a single failure pattern in this vault: the credit going to the more famous name in the neighbourhood. Deci and Ryan are one of the rare cases where the famous name is genuinely first author and the error was still made, by dropping the second. Neither has a page on this site yet, and nor does Pedro Teixeira.

---