Two years ago you answered a long list of questions about yourself and got a tidy row of scores. This week you answered them again, and one number moved. It is an odd moment. You are looking at a difference of a few points and asking something that sounds philosophical but is really a measurement question: has my personality changed?
Yes, personality can change. That is the direct answer, not a hedge. Personality traits can be relatively stable in one sense and change in another, and those two facts sit together without contradiction because they answer different questions. Stability usually means people keep roughly their relative position. Change usually means a group average shifts, or one person's level moves across years. A trait can be a durable pattern without being a life sentence.
What follows is not a promise that you can redesign yourself by wanting it hard enough. It is an explanation of what researchers measure when they say "change," where the evidence is stronger and where it thins out, and how to hold a test result without letting it hold you.
In short
- Traits do change across adulthood, and that change is not confined to your twenties.
- "Stable" usually means stable relative to other people, not frozen in place.
- Personality measures have changed during clinical interventions and programs designed to target personality, with important limits on generalization.
- Wanting to change alone has only a weak and mixed relationship with later change.
- A score difference alone does not establish lasting personal change.
What a trait is, and what would have to change for it to "change"
A trait is a recurring pattern, not a rule that fires every time. Someone high in conscientiousness still misses deadlines; the pattern is that they usually do not, across settings and over time. The Big Five describe five of these broad patterns — openness, conscientiousness, extraversion, agreeableness, and neuroticism — as continuous dimensions rather than categories.
That gap between "summary" and "script" is where most confusion starts, because three quite different things get called change:
- A changed behavior. You prepared an agenda before three meetings this month.
- A changed label. You now describe yourself as organized, or someone else does.
- Evidence of change in measured traits. A study examines whether trait measures change over a stated period, using an appropriate design and analysis.
These observations answer different questions. A changed behavior or label does not by itself establish enduring change in a person. Evaluating that claim requires comparable measurement, attention to uncertainty, and observations over time. Studies of group-average change address a different level: they do not require every participant to exceed an individual reliable-change threshold.
The five domains are broad, and beneath them sit narrower facets. A person can shift noticeably on one facet and look almost unchanged at the domain level, which is why a single broad score can hide the very movement you are asking about. The Big Five overview of domains and facets is a useful map here.
Three different questions are hiding inside "did I change?"
Researchers keep three questions apart because they can give different answers at the same time, and none of them is the "real" one (Roberts & Mroczek, 2008).
Mean-level change asks whether the average level in a population rises or falls over a defined period. It is a statement about a group, and it says nothing about who moved.
Rank-order stability asks whether people keep their relative positions. If people throughout a group broadly retain their relative positions over fifteen years, rank-order stability can be high even if the group's average level has risen.
Individual differences in change asks how far people depart from the average pattern. Some rise more than the average, some fall while the average rises, some barely move. This is the level you actually care about when the question is about you.
A fictional example with arbitrary units
Here is a small illustration. The units below are arbitrary. They are not FermatMind scores, not a norm, and not a scale anyone administers.
| Person | First occasion | Later occasion |
|---|---|---|
| A | 20 | 30 |
| B | 40 | 50 |
| C | 60 | 70 |
Check the arithmetic. The first average is (20 + 40 + 60) ÷ 3 = 40. The later average is (30 + 50 + 70) ÷ 3 = 50. The group mean rose by 10 units. And the ordering was A < B < C on the first occasion and A < B < C on the second. Nothing was violated: the average moved and every relative position held.
Two cautions. First, this is the tidiest possible case, because all three people moved by exactly the same amount. In real samples people move by different amounts, and some move downward, so the ordering can reshuffle even while the average climbs. Second, the "+10" is not a threshold. It tells you nothing about whether a 10-point move on any real instrument is meaningful.
What the example shows is logical compatibility: average change and relative stability are not rivals. What it does not show is a stability coefficient, a reliable-change cutoff, or anything about any actual person.
What long-term studies find in adulthood
Long-term studies track the same people over years. Roberts, Walton, and Viechtbauer (2006) synthesized 92 longitudinal samples and found a pattern that is more specific than the usual summary suggests.
People tend to increase in social dominance — one facet of extraversion, covering assertiveness and confidence — along with conscientiousness and emotional stability. In that synthesis, increases were especially apparent in young adulthood, roughly ages 20 to 40, and four of the six trait categories also showed significant change in middle and old age. The second extraversion facet, social vitality, behaves differently: it rises through adolescence and then declines in old age, and openness follows a similar arc. The analysis detected agreeableness change only in old age; that finding is not a rule that no younger individual can change. Longer studies and studies of younger cohorts showed larger change.
That distinction matters. In this review, social dominance and social vitality showed different age-related patterns. Saying simply that "extraversion increases with age" misses those differences. These were the review's subdomains, not a claim that every extraversion instrument has exactly two halves. Broad domain scores can conceal differences among narrower tendencies.
What this kind of research cannot do is equally specific. It cannot give you a timetable, because the averages describe samples. The patterns alone do not establish why change occurred. Differences by age or sample do not, by themselves, separate maturation from environmental influences or historical conditions. Nor can this evidence show that a particular job, habit, or event caused a particular person's shift.
What intervention studies add
Developmental studies observe patterns over time. Intervention studies include treatments targeting mental-health problems, during which personality may also change, and programs specifically designed to support desired trait changes. These are related but distinct research questions.
Roberts and colleagues (2017) reviewed 207 studies that tracked personality during interventions. Those 207 are not 207 randomized trials; the set deliberately includes true experiments alongside simpler pre-post designs, which matters for how much causal weight the total can carry. Interventions averaged about 24 weeks and were associated with an average standardized change of roughly d = 0.37. Most were clinical. Emotional stability moved most, then extraversion, and the type of therapy was not strongly related to how much changed. Those numbers are standardized mean changes, not percentages of people who improved.
Stieger and colleagues (2021) studied PEACH, a three-month smartphone program combining goal-linked behavioral practice and self-reflection. Of 1,523 consenters who passed screening and completed an initial assessment, 875 started the program; these are not completion counts.
The randomized study compared change across the intervention group's full program with change during the waitlist group's one-month wait. Self-reported change was greater in the intended directions in the intervention group. These were unequal observation windows: the waitlist then received the program, so there was no three-month untreated comparison.
From baseline to the follow-up three months after the program ended, participants reported desired changes. Observer reports detected intended increases, while aggregate intended decreases were not statistically significant. Nonsignificance does not prove no change, and group findings do not establish that each participant changed.
This was a specific program with motivated, largely young, German-reading nonclinical adults. Its findings do not validate an ordinary habit, this website's test, or permanent change.
What the 2024 review adds
Haehner, Wright, and Bleidorn (2024) reviewed the volitional change literature with a search cutoff of 1 October 2024: 30 longitudinal studies, 26 independent samples, 7,719 people. Samples were mostly from the United States and Europe, mean age about 26, and largely students or convenience samples.
Their headline number is the one most easily misread. The reported d = 0.22 is a within-group pre-to-post change across intervention studies — 7 studies, 26 effect sizes, with a confidence interval from 0.005 to 0.433 that nearly touches zero. It is not an intervention-minus-control effect. The review could not pool control-group change at all, because too few studies reported it, and it says so. It also did not perform a risk-of-bias assessment of the individual studies. An exploratory, non-preregistered follow-up analysis of 4 studies put change from pre-intervention to follow-up at d = 0.37, which is a loose signal rather than a settled estimate.
The review also separates wanting to change from participating in a studied intervention. Associations between change goals and later changes were generally modest and mixed, with some studies finding no association or movement in the opposite direction. A large combined analysis included some intervention samples, so it did not isolate the effect of wanting change without intervention. A desire can guide a goal; it does not establish that the desired change will occur, or which support will help a particular person.
A fictional case: correcting a conclusion in both directions
The following person is invented to show the reasoning, not reported from an interview, a user's result, or a trial.
Priya describes herself as relatively quiet. She takes a Big Five assessment and scores toward the lower end of extraversion. She draws a conclusion that feels obvious: this is how I am, and people like me do not become different. Eight months later her team reorganizes, she volunteers to run the weekly cross-team meeting because nobody else will, and she finds it fine. Six weeks in, she has a second conclusion: I have become an extrovert.
Both conclusions are too fast, in opposite directions.
What she can actually see. She ran six meetings. She spoke in each of them. She did not dread them as much as she expected. She also has one score placing her toward the lower end of a dimension, taken on a form whose details she no longer remembers.
What nobody has measured. Whether the score and any second score came from the same form, the same conditions, or the same comparison group. Whether her behavior changed anywhere other than a meeting she was running. Whether her experience in unstructured social settings — a crowded lunch room, a party — shifted at all. How much of the ease came from the agenda, which gave her a reason to speak and an audience expecting her to. What the meetings cost her afterward in energy, and whether that cost changed over the six weeks.
Which inference gets withdrawn. "People like me do not change," and "I have become an extrovert." Her first conclusion treats a tendency as an unchangeable identity, which neither her score nor the research justifies. Her second conclusion is not established by these observations: greater participation in one structured setting could occur with or without a broader trait change.
A better bounded conclusion. Something like: I chaired six meetings and felt less dread than I had expected. I do not yet know whether my broader tendencies changed, or what the meetings cost me afterward.
A plausible alternative explanation. Roles carry scripts. A meeting chair has a defined reason to speak first, a list of items, and a room that expects them to talk. That structure could make initiation easier; it is an explanation to check, not a known cause in her case. Her experience in the chair's seat might differ from other settings. Novelty could also matter, and her later experience may differ from these first six weeks.
A voluntary next step, and an explicit option to stop. If she is curious, she could pick one setting with no script — a break-room conversation, an unfamiliar social invitation — and notice what she does and what it costs, on a few occasions, without scoring anything. If the noticing itself makes her more self-conscious, stopping is a legitimate outcome. Nothing here requires follow-through, and there is no schedule to keep.
A score difference alone does not establish lasting change
A difference tells you that the recorded scores differed. Interpreting it as personal change requires comparable measures and evidence about measurement uncertainty; establishing durability requires observation over time. Different forms, comparison groups, response conditions and ordinary fluctuation can also contribute. A pair of scores with no such context cannot separate these explanations. The question deserves more room than a paragraph, so it has its own article on why retest results differ and what can still be compared.
What a test result can and cannot tell you about change
A few product facts are relevant here. FermatMind offers 90-item and 120-item forms. Do not directly compare or convert their scores without an applicable linking method. Its public explanation describes reverse coding, facet and domain means, and conversions when reference parameters are available. Under that disclosed calculation, percentiles are estimated with a normal-distribution model rather than counted directly from an observed population. When no usable norms are available, the compatibility path uses a linear score that is not a population percentile. Norm status is labelled CALIBRATED, PROVISIONAL, or MISSING, and that label should be read alongside any number.
This is calculation disclosure, not evidence that the instrument can detect a given person's change; a disclosed method is a different kind of evidence from a validation study. The guide to reading Big Five results lays out the scoring and the remaining evidence gaps, and the Big Five test page explains the two forms and the reference-status labels in full.
Where this leaves you
The question at the start has a cleaner answer than it first appears. Your traits can change: group averages move across adulthood, and structured programs have been followed by measurable movement in many people who wanted it. Your traits are also stable: relative position tends to hold, which is what makes a trait a useful description rather than a mood report. Those statements describe different levels, and both are true at once.
For self-reflection, note the form, comparison basis and conditions when you answer. Read comparable scores alongside repeated behavior across relevant settings and, if you choose, observations from someone who knows you well. These sources can complement one another; informal records do not replace appropriately designed longitudinal measurement or independently establish lasting trait change. You do not need to keep testing or recording when it no longer serves a useful question.
This evidence leaves room for change without promising complete control. It can help you distinguish a useful goal, an observed behavior and a claim that still needs stronger evidence.
References
- Roberts, B. W., Walton, K. E., & Viechtbauer, W. (2006). Patterns of mean-level change in personality traits across the life course: A meta-analysis of longitudinal studies. Psychological Bulletin, 132(1), 1–25. https://doi.org/10.1037/0033-2909.132.1.1
- Roberts, B. W., & Mroczek, D. (2008). Personality trait change in adulthood. Current Directions in Psychological Science, 17(1), 31–35. https://doi.org/10.1111/j.1467-8721.2008.00543.x
- Roberts, B. W., Luo, J., Briley, D. A., Chow, P. I., Su, R., & Hill, P. L. (2017). A systematic review of personality trait change through intervention. Psychological Bulletin, 143(2), 117–141. https://doi.org/10.1037/bul0000088
- Stieger, M., Flückiger, C., Rüegger, D., Kowatsch, T., Roberts, B. W., & Allemand, M. (2021). Changing personality traits with the help of a digital personality change intervention. Proceedings of the National Academy of Sciences, 118(8), e2017548118. https://doi.org/10.1073/pnas.2017548118
- Haehner, P., Wright, A. J., & Bleidorn, W. (2024). A systematic review of volitional personality change research. Communications Psychology, 2, Article 115. https://doi.org/10.1038/s44271-024-00167-5