The question
What evidence justifies the conclusion that a person has changed?
Definition
A claim that a person has changed is justified when the evidence is appropriate to the specified outcome, level and timescale and provides sufficient reason to conclude that the observed difference reflects the change being claimed rather than measurement error, ordinary variability or a plausible alternative interpretation.
A person says they feel different.
Someone close to them notices fewer arguments. A questionnaire score has improved. Their behaviour in a difficult situation looks different from before.
Has the person changed?
Perhaps.
But those observations do not all support the same conclusion, and none can be interpreted properly until we know what change is actually being claimed.
If the claim is that behaviour changed, behavioural evidence matters directly. If the claim is that subjective anxiety decreased, first-person report becomes highly relevant. If the claim is that the person has undergone a lasting psychological transformation, the evidential burden is much greater.
The problem is therefore not simply whether we have evidence.
It is whether the evidence we have is appropriate to the conclusion we want to draw.
Evidence begins with the claim
Different things can change.
Behaviour can change. Experience can change. Cognitive performance can change. Interpersonal patterns can change. Bodily and neural processes can change.
These outcomes may be related, but they are not interchangeable.
Evidence that someone behaved differently tells us something about behaviour. It does not automatically establish that their identity changed.
A person's report that they feel less anxious may be highly informative about their experience. It does not establish that their behaviour, physiology or long-term pattern changed to the same degree.
A neural measure may reveal a neural difference. Its psychological meaning still has to be justified.
The evidential value of an observation therefore depends partly on the claim it is being asked to support.
The first question should be:
What exactly are we claiming changed?
Only then can we ask whether the available evidence fits the claim.
What was actually observed?
Evidence often becomes inflated as we move from observation to interpretation.
Suppose a person's questionnaire score is lower than it was three months ago.
What we know most directly is that the recorded score differs.
We may then infer that the psychological construct represented by the questionnaire has changed. That inference may be well supported, but it still depends on what the measure captures and how confidently the observed difference can be interpreted.
The same logic applies elsewhere.
A person reports feeling calmer. Someone observes fewer arguments. Performance on a cognitive task improves. Physiological activity differs. A neural signal changes.
Each observation can be informative. Broader conclusions require additional justification.
The further a conclusion moves beyond what was observed, the more support is required for that move.
This does not make external observation inherently superior to subjective evidence. Some phenomena are partly defined by experience itself. Pain, meaning or felt anxiety cannot simply be replaced by behavioural or biological proxies without changing the question.
The relevant distinction is between what was observed and what that observation is being used to establish.
Measures do not remove that distinction.
A questionnaire is not identical to the construct it is intended to represent. A cognitive task is not the entirety of a cognitive capacity. A physiological indicator does not carry a fixed psychological meaning. Behavioural observation captures behaviour, not every process that produced it.
A changed measure may therefore support a claim of underlying change without proving it automatically.
The reverse is also possible. A person may change while a particular measure fails to detect that change because it is noisy, insensitive, too narrow or poorly matched to the phenomenon.
Both errors matter: a measure changed, therefore the broader construct definitely changed and the measure did not change, therefore nothing changed go beyond what the observation alone can establish.
Change relative to what?
A claim about within-person change requires comparison.
If someone exercises twice this week, that observation means little by itself.
If they have exercised twice every week for years, there may be no meaningful departure from their prior pattern. If they had not exercised for a year, the same week may tell us something very different.
The reference point matters.
But even "baseline" can sound more secure than it is. A single earlier observation may capture an unusual day rather than a person's ordinary functioning. Mood, sleep, performance, behaviour and physiological measures can all fluctuate.
The better question is:
What prior functioning or trajectory is the new observation being compared with?
Sometimes one earlier observation is informative enough. In other cases, repeated observations are needed to characterise what is normal for that person before the newer pattern can be interpreted.
This is also why between-person difference and within-person change must remain separate.
If one person's score is lower than another person's today, the two people differ.
That does not tell us whether either person changed.
A recorded difference is not automatically meaningful change
Suppose a score moves from 52 to 55.
The recorded values differ.
What remains uncertain is what produced that difference and what conclusion it warrants.
Measurements contain error. People also fluctuate.
Fatigue, stress, recent events, practice effects, context and temporary states can alter an observation without establishing the broader change being claimed.
Methodological approaches to reliable change address one part of this problem by asking whether a score movement is large enough to exceed what would reasonably be expected from measurement unreliability.
But a second question remains.
A statistically reliable difference and a substantively meaningful change are not identical.
A score may move beyond expected measurement error while the difference has little practical or psychological importance. An apparently important difference may also remain difficult to interpret when the measure itself is imprecise.
So we need to distinguish: Is the observed difference sufficiently reliable? from: Does that difference justify the substantive claim being made?
There is no universal threshold that answers both questions across all forms of human change.
The broader standard is more useful: the evidence should give us adequate reason to distinguish the claimed change from ordinary variability, measurement uncertainty and plausible alternative interpretations.
The timescale of the evidence must match the timescale of the claim
Consider one calm response during conflict.
That observation can support: this person responded calmly in this situation.
It cannot by itself establish: this person now generally responds differently during conflict.
Still less: this person's emotional functioning has changed in a lasting way.
Each statement extends the claim further through time.
The evidence has to extend with it.
A temporary state can be supported by evidence appropriate to a temporary state. A recurring pattern requires repeated observations. A claim of stable or lasting change requires evidence capable of supporting stability or durability rather than mere occurrence.
Different outcomes operate on different timescales, so no universal observation period can do this work for every phenomenon.
The principle is simpler:
The timescale of the evidence must match the timescale of the change being claimed.
Without that alignment, the conclusion becomes temporally broader than the evidence.
Different evidence answers different questions
There is a recurring temptation to rank forms of evidence as if one method were intrinsically closer to the truth.
That is usually the wrong comparison.
Self-report can be especially informative when the claim concerns subjective experience, beliefs or perceived states. Its limitations—recall, interpretation, reporting conditions and incomplete introspective access—matter, but they do not make subjective evidence inherently weak. For some questions, subjectivity is part of the phenomenon itself.
Behavioural observation bears closely on behavioural claims. If someone who previously avoided public speaking repeatedly begins giving presentations, that is strong evidence that their behaviour changed. It does not uniquely reveal whether confidence, training, incentives, social pressure or something else produced the change.
Performance measures, physiological measures and neural measures can each provide important evidence at the levels they assess. Their interpretation becomes less direct when they are used to support broader claims about psychological meaning.
A neural difference, for example, can establish a measured neural difference. It does not automatically certify that a psychological change is deeper, more genuine or more important than one supported through another method.
Observer reports can add information about interpersonal or behavioural patterns outside the person's own perspective, while still reflecting the observer's limited contexts and interpretations.
All measurement requires interpretation when it is used to support a broader claim.
That is why there is no universally superior form of evidence independent of the question being asked.
The best evidence is evidence capable of supporting the claim.
Convergence can strengthen evidence
Sometimes several appropriately chosen methods point in the same direction.
A person reports lower anxiety. Their avoidance decreases. Someone close to them notices that they enter situations they previously avoided.
Because these observations arise through different routes, their convergence may strengthen confidence that the broader pattern changed.
This is one reason multimethod evidence can be valuable.
Yet convergence does not mean that every genuine change must appear at every level.
A behavioural change may occur before self-description catches up. A subjective change may be real even when an unrelated biological measure is unchanged. Different aspects of a person can move at different rates or remain partly dissociated.
Additional measurements therefore help only when they actually bear on the claim.
Evidence should converge on the claim, not merely accumulate around the person.
More data is not automatically stronger evidence if the added observations are poorly aligned with the conclusion.
Evidence that change occurred does not tell us why
Suppose someone's anxiety decreases after beginning therapy.
The evidence may strongly support the conclusion that their anxiety changed.
A different question is whether therapy caused the change.
The temporal sequence alone cannot settle that. Other circumstances may have changed during the same period, natural fluctuation may have contributed, and several causes may be involved.
This boundary matters throughout the study of human change.
Evidence that an outcome changed and evidence that identifies its cause are different evidential problems.
This essay is concerned with the first.
The conclusion must not outrun the evidence
Compare these statements:
Her score was lower today.
Her reported anxiety decreased.
Her anxiety has changed.
Her emotional functioning has changed.
She has undergone lasting psychological transformation.
Each conclusion travels farther from the immediate observation.
As the claim becomes broader, more enduring or more abstract, the evidential burden increases.
A score change may require evidence that the measure is reliable and meaningfully related to the construct. A claim of broader psychological change requires evidence extending beyond one indicator. A claim of lasting change requires temporal evidence. A claim of transformation carries further assumptions about the breadth or depth of what changed.
Stronger evidence does not necessarily mean more instruments, more expensive technology or biological measurement.
It means evidence capable of supporting the additional inference.
The strength and breadth of the conclusion should not exceed what the evidence can support.
Human-change language can expand faster than the observations beneath it.
A behavioural difference may be interpreted as identity change. A subjective shift may be treated as evidence of transformation. A short period of improvement may be described as lasting.
Any of those conclusions might eventually be justified.
The evidence has to get there too.
What counts as evidence of change?
The argument can now be stated as a general criterion.
A claim that a person has changed is justified when the evidence is appropriate to the specified outcome, level and timescale and provides sufficient reason to conclude that the observed difference reflects the change being claimed rather than measurement error, ordinary variability or a plausible alternative interpretation.
That criterion does not prescribe one universal method.
It requires alignment.
We need to know what is claimed to have changed, what was actually observed, what comparison makes the difference interpretable and whether the timescale of the evidence matches the timescale of the conclusion.
We also need to recognise how much inference lies between observation and claim.
No single indicator can establish every form of human change because different indicators answer different questions.
Precision rather than certainty
Taking measurement error, ordinary variability and inference seriously can make the evidential standard sound demanding.
It is demanding.
But it does not make human change unknowable.
The purpose of evidential discipline is to make claims about change more precise.
Instead of asking only whether someone has "really changed," we can ask:
What changed?
What evidence supports that conclusion?
Across what time and relative to what prior pattern?
How much does the conclusion extend beyond what was directly observed?
Those questions do not remove uncertainty. They tell us what kind of confidence the evidence justifies.
Once that foundation is established, later questions can be asked more rigorously: how the construct should be measured, what caused the change, what mechanism produced it, and how robust the scientific conclusion is.
Those are downstream problems.
This essay establishes the condition that comes first: before explaining human change, we need to know what our evidence actually allows us to say changed.