The Evaluating Self

29The Psychology of Change

Geoffrey L. Cohen and David K. Sherman, Annual Review of Psychology 65, pp. 333–371, 2014

Linked to 23 chapters: supports 8 · contradicts 2 · mechanism 17 · same claim 6 · pairing 6

The Psychology of Change, 2014 — the argument in five moves The selfunder threat Howaffirmation works Why a briefeffect lasts Three domainsof evidence When it works,when it does not A need to be adequate;threat sounds analarm; defences blocklearning. Ten minutes on avalue widens the self,quiets the alarm,frees resources. Recursion, interaction,construal: a cycle ofadaptive potentialcarries it forward. Grades for years;openness to healthwarnings; concessionsand warmer couples. Needs a threat andresources beneath it;timing decides;a catalyst, not a cure. Affirmation changes nothing by itself. It releases what threat was suppressing.
Figure 29.0 — The review in five moves: a theory of the self under threat, a mechanism, a claim about how brief effects last, three domains of evidence, and the conditions under which the effect appears at all.

Cohen and Sherman review twenty-five years of work on a small intervention with large claims attached to it. People need to see themselves as adequate, as good and capable, and when an event threatens that view an inner alarm sounds. The alarm narrows attention to the threat and sets off defences, denial, rationalisation, blame, comfort eating, drink, that protect the self in the short run and block learning and change in the long run. Self-affirmation theory, which Claude Steele set out in 1988, says the alarm can be turned down from outside the threatened domain. Ten minutes spent writing about something one values, family, friends, religion, music, reminds a person of the larger self that a bad grade or a doctor's warning does not touch, and with the alarm quieter the evidence in front of them can be seen for what it is. The review's second claim is about time. Interventions this brief have raised the grades of minority students for years afterwards, and the authors explain it by a loop in which a better outcome affirms the person, changes how the school treats them, and changes how they read the next setback, so that the effect propagates long after the writing exercise is forgotten. The review is careful about where this works and where it does not. There must be a threat to lift, and there must be something underneath it, a motive and a capacity, for the lifting to release.

The self under threat

The review opens with two old experiments. Kurt Lewin changed what wartime homemakers bought by having them talk in small groups until a new norm formed. Milton Rokeach strengthened Americans' support for civil rights, durably, by insinuating that they valued their own freedom more than other people's. Both worked on the self, and the review's premise is that the self is both the barrier and the catalyst to change.

Psychological threat is the perception that the environment challenges one's adequacy. It is easy to trigger. People rationalise a trivial choice between two records; partisans reflexively refute contrary evidence; fans eat worse after their team loses; petty insults have led to homicide. Greenwald once compared the self to a totalitarian regime that suppresses unwelcome information, but the authors prefer a storyteller, since people can also be harshly self-critical and can appraise themselves accurately at moments of truth. The goal is not to spin every threat but to keep a narrative of overall adequacy that gives enough optimism to stay in the game. Self-affirmation theory makes three points about that motive. It is global, not domain-specific: the narrative is of being a good person, not a good student, and can draw on many roles. It aims at adequacy, not excellence, so an event need only confirm competence in some valued domain. And it aims at being praiseworthy rather than praised, which is why telling low-self-esteem people to repeat that they are lovable backfires.

Threat, defence and affirmation — the review's mechanism a bad grade,a doctor's warning threat: the alarmattention narrows defences: deny, rationalise,blame, drink, give up protects the self now; blocks learning later a value affirmedoutside the domain the self seen whole; the threatsmaller; the evidence readable turns the alarm down The motive is for adequacy, not excellence; for being praiseworthy, not praised. Any valued domain can supply it.
Figure 29.1 — The mechanism. Threat sounds an alarm and sets off defences that protect the self and block change; an affirmation from outside the threatened domain quiets the alarm and frees the resources the defences were consuming.

A self-affirmation is any act that demonstrates adequacy, and small ones count: reflecting on one's family, exercising a little control over daily visits in a nursing home, a personal note from a doctor. The most studied version has people choose from a list of values the one or few most important to them, with values in the threatened domain removed from the list, and write for about ten minutes about why the value matters and a time it did. The content is self-generated. People most often write about relationships, and also about religion, humour and kindness. The essays the review reproduces are not self-congratulation; they read as a time-out to regain perspective on what matters.

How it works

Three effects follow, in the authors' account. Affirmed people take a wider view and the threat commands less vigilance; unaffirmed students judged a caged tarantula to be closer than it was, and affirmed ones judged its distance accurately. A threat has less impact on well-being; a poor grade did less lasting damage to affirmed minority students' sense of belonging. And people approach rather than avoid: affirmed participants rated the threatened domain as more important, sought threatening health information rather than shunning it, and showed a stronger neural error signal on a cognitive task, which the authors read as greater attention to their mistakes. Two routes carry these effects. One is buffering: students who had written about a value showed no cortisol rise before a stressful speech, as people with social support feel less pain under shock. The other is reduced defensiveness: affirmed people were more open to a report linking their behaviour to cancer. Affirmation does not change anything by itself. It releases forces, cognitive aptitude or persuasive evidence, that threat was suppressing.

The values-affirmation exercise a list of valuesthreat domain left off choose the onethat matters most write ten minuteswhy, and one occasion nobody grades itpurpose left unstated Most people write about family and friends; religion, humour and kindness follow. The essays read as a time-out to regain perspective, not as self-praise. Two to five times a year in the school studies; once or twice elsewhere. Told what it is for, people gain less; given a choice whether to do it, the effect returns.
Figure 29.2 — The values-affirmation exercise as it is usually run. The list excludes the threatened domain; the writing is brief and self-generated.

Why a brief effect can last

The review's most distinctive argument is about persistence. Many social-psychological interventions have lasting effects when they reach a process at the right point; teachers told to expect certain students to bloom drew better performance from them, but not if told after impressions had formed. A successful intervention is a turning point in a story rather than an isolated event. Three principles explain the propagation. Recursion: an output cycles back as an input, so better performance affirms the self and produces better performance. Interaction: the output feeds other processes in the system, so a student who does better is held to higher expectations and placed in a higher track. Subjective construal: the intervention shifts how experience is filtered, so adversity is read as an isolated event rather than a verdict. The authors' parable is the professional hockey player. Children born just after the age cut-off are the oldest in their cohort, look more talented, get more practice and better teams, and grow into the identity, so a birthday with no causal power of its own becomes life-altering by putting a child on what the authors call a cycle of adaptive potential.

The cycle of adaptive potential — after the review's figure 1 the self-systemfeels adequate, or not adaptive outcomesa better test, a first win the social systemteachers, tracks, friends a. acts, performs c. is noticed e. is affirmed by others b. feels affirmed d. is expected more of Recursion: the output returns as input. Interaction: it feeds other processes. Construal: the next setback is read differently. The hockey player's birthday has no power of its own; it starts this loop.
Figure 29.3 — The cycle, after the review's Figure 1. The self acts, the social system reacts, and each pass strengthens the next; an intervention only has to start it.

Education

The first field experiments were in three racially mixed middle schools, where half the students were African American or Latino American. Teachers handed out the exercises as ordinary assignments, two to five times a year, starting in about the fourth week and usually just before a test, without knowing who was in which condition. The affirmation raised the grade point average of the minority students and not of the white students, and so closed the gap: it halved the share of African American students earning a D or F in the first term, and the closure came to about 30 per cent at two sites and 22 per cent at a third with economically disadvantaged Latino students. The effect persisted for two years and, where records allowed, into high school. It did not lift grades so much as slow the decline that is normal in middle school, so the trajectories diverged. The most threatened subgroup, those with a history of poor performance who felt least sure they belonged, gained almost a full grade point. Fewer affirmed minority students were placed in remedial tracks, 3 per cent against 9, and more took advanced mathematics. Boosters in the second year added nothing; the first year's effect explained the second's.

Later studies extended the pattern. Among first-generation college students, over nine in ten of them white, two affirmations in an introductory biology course cut the gap with their better-off peers by half, and 86 per cent went on to the next course against 66 per cent. Women in introductory physics scored higher on exams, with the largest gains among those who most wondered whether the stereotype about women and science might be true. Affirmed adolescents with grandiose self-views were less likely to respond to a blow to their self-esteem by hitting, name-calling or spreading rumours. One British study of medical students found mixed effects, which the authors attribute to instructors reading and grading the essays, which made the exercise evaluative and public.

Health

Health information threatens the self as much as a grade does, and defences block the simple steps that would save lives. Smokers who had first listed their good qualities found graphic warnings more distressing and were more motivated to quit a week later; among low-income smokers, 59 per cent of the affirmed took leaflets on quitting against 37 per cent. People at risk of diabetes who had reflected on their values were more willing to be screened. Drinkers in a bar who completed an affirmation before reading about alcohol's risks drank less a month later, and the share within recommended limits rose by 32 percentage points. Effects on attitudes tend to last; effects on behaviour have been mixed, which the authors take as a reason to pair affirmation with concrete plans and with early behavioural wins, and to place the affirmation at the moment of temptation, with a bracelet, a key chain or an if-then rule.

Five outcomes, control against affirmation — per cent low-income smokerstaking a leaflet on quitting first-generation studentsenrolling in the next course cardiac patients reachingrecommended activity hypertensive patientsusing their medication minority students placed inremedial track or held back 3759 6686 3755 3642 93 grey: control teal: affirmed
Figure 29.4 — Five outcomes the review reports, control against affirmation. All are from randomised studies; the school and clinic figures are from field trials.

Affirmation also buffers stress. Students facing the exam they had named as most stressful showed no rise in adrenaline if they had written about their values, and again the effect was concentrated among the most threatened. In a reanalysis of an expressive-writing trial with breast cancer patients, the active ingredient was self-affirmation rather than reappraisal of the illness. Women concerned about their weight who wrote about a value before being weighed had lost more weight two and a half months later, with lower body mass and smaller waists, and the effect ran through improved working memory. Three large trials in 2012 combined affirmation with a standard regimen for patients with chronic conditions: medication use among hypertensive African American patients rose from 36 to 42 per cent at a year, and the share of cardiac patients reaching recommended activity rose from 37 to 55 per cent, with a null result in young asthma patients except among those who had been hospitalised. Those trials also included mood-lifting gifts, so the authors cite a cleaner double-blind study in which African American patients who wrote about a value before an appointment with a white doctor gave and asked for more information, and their exchanges were warmer, more attentive and less distressed.

Conflict and relationships

Affirmed Israelis and Bosnians were more willing to acknowledge their own side's wrongs and to support reparations. Affirmed partisans judged a presidential debate less partisanly, and affirmed Republicans expected the winner to govern more fairly ten days later. Arab American men got prejudiced listeners to consider their perspective by first asking them self-affirming questions. In negotiation, affirmation reduces the escalation of commitment that comes when identity binds to a position, and people who had been reminded of their political commitments made more concessions on abortion legislation and trusted the other side more. In marriage, defensiveness is among the strongest predictors of distress, and happy couples break the tit-for-tat spiral with a disarming compliment or a shared joke. Experiments show the spiral can be interrupted from outside: affirmed low-self-esteem partners derogated their partners less, insecure people behaved more easily with a stranger weeks later, and students asked to explain why their partner had once complimented them felt more secure two weeks on and reported their partners behaving more supportively.

When it works and when it does not

The review's closing section is a list of conditions, and the authors say plainly that affirmation is a catalyst, not a panacea. There must be resources for growth once the threat is lifted, an ability inhibited by stress or information the environment supplies; without them affirmation can produce disengagement. There must be a threat: where there is none, affirmation can raise confidence and resistance to change, and where threat and defensiveness are what produce the good outcome, affirmation reverses it, so that people whose identity as open-minded negotiators had been made salient became more closed-minded when affirmed. Timing governs everything. In the laboratory an affirmation given before threatening information, or soon after, reduces defensiveness, but once a person has rationalised, it cannot undo the rationalisation. In the schools, affirmation in the first week of term helped as much again as affirmation a few weeks later, because early outcomes compound. Values that transcend the self, compassion, service, being loved, work better than power, status or approval, and affirming oneself inside the threatened domain can increase defensiveness. The exercise loses force when people are told what it is for, unless they are also given a choice about doing it. And it is indirect: it does not address the problem, which the authors note is true of many effective interventions, from volunteering to meditation.

Timing — the window and the compounding in the laboratory affirmation works here: before, or soon after the threat arrives rationalised: too late to undo in the school year first week: as much again as a few weeks later early outcomes compound; the second year needs no booster Cook et al. 2012; Cohen et al. 2009
Figure 29.5 — Timing, in the laboratory and in the school year. The window closes once a person has rationalised the threat, and early effects are worth more because they compound.
When affirmation does nothing, or reverses no threat presentnothing to lift nothing underneathno motive, means or kindling the defence is usefulthreat driving the good outcome value inside the domainor self-centred: status, approval confidence rises, and resistance to change with it Briñol et al. 2007 disengagement; the effect reveals untapped potential, not its own power Vohs and Schmeichel 2013 the open-minded negotiator, affirmed, became more closed-minded Cohen et al. 2007 defensiveness increases; self-transcending values work best Blanton et al. 1997; Burson et al. 2012 And subtlety: told that the exercise is meant to help them, people gain less, unless they are free to decline it.
Figure 29.6 — The boundary conditions the review lists. Each row is a case where affirmation does nothing or reverses.

The authors connect the work to resilience research, whose adaptive strategies match affirmation's effects, and to the study of social ties, since most people affirm themselves through relationships and much self-affirmation is social affirmation. They see the mechanism at work in behavioural activation therapy, motivational interviewing, expressive writing, a Superman cape that tripled preschoolers' delay of gratification, and a survey that raised turnout by asking how important it was to be a voter rather than to vote. They end with Raymond Carver's baker, whose cinnamon rolls given to grieving parents are a small, good thing at precisely the right moment.

Since 2014

The review appeared at the peak of confidence in brief psychological interventions, and the decade since has tested it. Attempts to reproduce the school results at scale, across whole districts and many teachers, found effects that were smaller and more variable than the original trials, present in some sites and absent in others, and largest where the identity threat was most acute, which is what the review's own boundary conditions predict and also what makes the intervention hard to deploy as a policy. Some laboratory findings the review leans on, including the neural error signal and several health results, have been reported with smaller effects or not at all in larger replications. The response from the field has been to reframe the intervention as a seed that depends on the soil: an affirmation does nothing unless the setting affords the change it makes possible, and identifying that affordance is now treated as the first task. The theory of the self under threat, the observation that defences protect adequacy at the cost of learning, and the account of how early outcomes compound through a system have held up better than any single effect size, and the practical lesson the review ends on, that timing and the presence of a real threat decide whether a small good thing does anything, has become the standard caveat for the whole genre.

Claims drawn from this chapter · 8

opinionWhy a brief effect can last (the cycle of adaptive potential, figure 29.3; the hockey birthday parable)

A short exercise lasts only when it sets off a feedback loop

Cohen and Sherman argue a brief exercise has lasting effects only when it starts a loop: the person does better, other people treat them better, and they read the next setback differently. A second year of booster sessions added nothing.

Becausethe loop is an explanation offered after the fact, consistent with the null from the second year's boosters but not tested against an alternative.
LimitsRecursion, interaction and construal name propagation rather than measure it, and the same account would explain a null as a system that failed to react.
PracticePlace a small intervention where the system will react to its first result, and stop repeating it once the loop is running; repetition is not what carried the original effect.
After AIThe loop runs in both directions. An engineer who reads one delegated success as evidence of his own competence takes on harder work; one who reads it as evidence of his redundancy takes on less. The tool supplies the same first outcome either way.
evidenceWhen it works and when it does not (timing, figure 29.5)

Affirmation works before a threat, not after you have explained it away

Cohen and Sherman find that writing about your values before a threat, or soon after it, lowers defensiveness, while the same exercise given after the person has already talked themselves round has no effect.

Becausethe laboratory timing manipulation and the school-year comparison point the same way, and the second is a field test of the first.
LimitsThe school comparison is between weeks of a term rather than a randomised timing manipulation, and having rationalised is observed only through the failure of the affirmation.
PracticePut whatever is meant to change how a person reads an output before they open it. Cognitive forcing is that rule in its strongest form: write your own answer first.
After AIBy the time a person has explained to themselves why the model is wrong, the intervention that would have helped does nothing. The window is before the output is opened, which is also when nobody thinks they need it.
evidenceWhen it works and when it does not (the boundary conditions, figure 29.6)

Affirming your values helps only where a threat and a resource exist

Cohen and Sherman find that writing about your own values only helps when there is a real threat and some ability or knowledge waiting to be used. Without both, it does nothing, and it can make people more closed-minded than before.

Becausethe reversal is demonstrated rather than inferred — people whose identity as open-minded negotiators had been made salient became more closed-minded when affirmed — and the decade since has reframed the intervention as a seed that depends on the soil.
LimitsThe absence of a resource is identified after a null result, which makes the condition hard to apply in advance, and the review concedes the intervention is indirect and does not address the problem itself.
PracticeBefore installing anything meant to make review calmer, check that there is both a real threat and something to release — a skill, a checklist, evidence the person can act on. Witness practice in review assumes the review is still happening.
After AIA reviewer whose defences have been quieted still needs something to review with. Where the failure-mode knowledge is missing, the affirmation releases nothing and the output goes through anyway.
evidenceThe self under threat, and How it works (the exercise, figure 29.2)

People defend being a good person in general, not good at something

What the defences protect is a broad sense of being a good and capable person, not being good at any one thing. That is why writing about a value from an unrelated part of your life calms a threat to the part under attack.

Becausethe standard exercise removes values in the threatened domain from the list and still works, while affirming inside the threatened domain raises defensiveness instead.
LimitsThe essays are self-generated, so what did the work in any given case is unknown, and the review's own boundary conditions say values of power, status and approval do not serve.
PracticeWhen someone is about to hear something hard about their work, do not reassure them about their work. Give the ten minutes to something else they are, and then deliver the evidence.
After AIThe reflex answer to the machine is to defend one's competence at the thing the machine did. The evidence says that is the move that increases the defensiveness rather than the one that removes it.
evidenceThe self under threat (figure 29.1)

People defend themselves rather than learn when their identity is threatened

When something threatens a part of the self a person values, the first response is defence: denial, excuses, blaming others. It protects the person's sense of adequacy now and blocks learning later.

Becausethe pattern is produced at will across unrelated domains — people rationalise a trivial choice between two records, partisans reflexively refute contrary evidence, fans eat worse after their team loses — and the affirmation manipulation reverses it in randomised trials.
LimitsSeveral laboratory results the review leans on, including the neural error signal and some health outcomes, have come back smaller or absent in larger replications. The theory of threat has held better than any single effect size.
PracticeTreat the first response to unwelcome evidence as defence rather than as judgement, in yourself first. Witness practice in review is that noticing made into a habit with a log.
After AIA machine that does the work better is a threat of exactly this kind, and the response the literature predicts first is not learning but an account of why the comparison does not count.
evidenceWhen it works and when it does not; and the British medical-student study in Education

Telling people what the writing exercise is for removes its effect

The values-writing exercise loses its force when people are told what it is for, unless they are also given a choice about doing it. British medical students showed no benefit because instructors read and graded their essays.

Becausethe disclosure finding is a direct manipulation, and the medical-student null is explained by the same mechanism — instructors read and graded the essays, making the exercise evaluative and public.
LimitsThis is what makes the intervention hardest to deploy honestly at scale, and the choice qualification rests on a smaller literature than the disclosure effect itself.
PracticeIf a practice has to be announced, make participation a real choice; announced and compulsory is the combination that does nothing.
After AIAny instrument installed to make a team review a model's output more carefully stops working the moment it is introduced as such, unless people are given a genuine choice about using it.
contestedEducation (the three middle schools; figure 29.4)

Ten minutes writing about a personal value lifts threatened students' later grades

Ten minutes spent writing about something you personally value, done just before or soon after a threat to your standing, raises later performance in that area — but only for the people the threat hits hardest.

Becausethe original field trials were randomised and large — the share of African American students earning a D or F in the first term halved, and the effect persisted two years — but district-scale replications since have been smaller, more variable and present in some sites only.
LimitsThe effect is concentrated in the most threatened subgroup and absent where there is no identity threat, and one British medical-school study found mixed effects, attributed to instructors grading the essays.
PracticeIf you try it, try it early on the people whose identity the work threatens, and measure whether anything moved. Witness practice in review is the version that needs no trial to run.
After AIThe most threatened population in an engineering team is the one whose identity rests on writing the code. Whether the intervention survives outside the schools where it was found is the open question, and often it has not.
opinionWhen it works and when it does not

Writing about compassion lowers defensiveness; writing about status does not

Cohen and Sherman report that writing about values that reach beyond yourself — compassion, service, being loved — makes people less defensive about bad news. Writing about power, status or being approved of does not.

Becausethe comparison is assembled across studies in the review rather than run as a designed contrast, alongside the finding that affirming within the threatened domain raises defensiveness.
LimitsValues are self-selected, so the comparison is between people who chose differently rather than between conditions. The aspiration literature in self-determination theory makes the same distinction on stronger evidence.
PracticeWhere worth is being rebuilt, rebuild it on something that is not a ranking. Witness practice in review keeps the attention on the act rather than on the standing it earns.
After AIThe values most available to an engineer under threat are the ones the review says do not work — being the best at this, being the one who is asked. The ones that do work are not about the work at all.

All claims · Source map