Psychometric assessment scores are often reasonably stable, but they are not permanent labels. A score reflects a person’s responses under particular conditions: at a specific time, with particular instructions, using a particular instrument, and sometimes in a particular emotional or physical state.
The most accurate answer is:
Score stability depends on what is being measured, how long the interval is, how reliable the assessment is, and whether the person or circumstances have changed.
A well-designed assessment may produce similar results across repeated administrations, especially when it measures a relatively enduring characteristic. Still, some variation is normal and expected.
What does “stable over time” mean?
In psychometrics, stability usually refers to test–retest reliability. This is the degree to which people receive similar scores when they complete the same assessment on two or more occasions.
A stability estimate is generally expressed as a correlation between scores at different times. A higher correlation means that people tend to maintain their relative standing within the group. For example, if people who score relatively high the first time also tend to score relatively high later, the assessment shows stronger rank-order stability.
This does not mean that every person receives exactly the same score twice. Correlations describe patterns across people, not perfect agreement for each individual.
It is also useful to distinguish two ideas:
Rank-order stability: whether people maintain their relative positions compared with others.
Mean-level stability: whether the average score for a group remains similar over time.
A group can show high rank-order stability while its average score changes. For example, everyone might improve after training while the same people remain relatively higher or lower than one another.
Why do scores change?
Scores can change for several legitimate reasons.
Normal measurement error
Every assessment includes some measurement error. A person’s observed score is commonly understood as a combination of a relatively stable component and error associated with the measurement occasion:
X=T+E
Here, X is the observed score, T is the person’s theoretical true score, and E represents error. The true score is not directly observed; it is a statistical concept used to describe the stable part of measurement.
Sources of error may include item sampling, temporary distraction, unclear instructions, guessing, scoring limitations, and ordinary fluctuations in attention or mood.
Familiarity with the assessment
People may remember questions, become more comfortable with the format, or develop strategies after taking an assessment once. This can affect later scores, particularly when the retest interval is short.
On the other hand, a person may perform worse later because of fatigue, reduced motivation, or less favorable testing conditions.
Changes in the person
Some psychological characteristics are relatively enduring, while others can shift with development, learning, treatment, stress, illness, life events, or changes in the environment. A later score may therefore reflect a genuine change rather than a problem with the assessment.
Changes in the testing conditions
The setting, device, time limit, administrator, accommodations, language, and instructions can all affect performance. Comparisons are most meaningful when the procedures are sufficiently similar.
Regression toward the average
When someone is selected because of an unusually high or low score, a later score will often be closer to the group average partly because of random measurement variation. This statistical tendency is called regression toward the mean. It does not necessarily indicate true improvement or decline.
Are intelligence scores stable?
Cognitive assessment scores can show substantial stability, particularly when the same broad abilities are measured with reliable instruments and the interval is not unusually short or long. However, stability varies by age, ability, instrument, and testing conditions.
Scores during childhood may be less settled than scores later in development because cognitive abilities and related skills are still changing. Longitudinal research has found that intelligence-test scores can shift meaningfully during childhood and adolescence, even though they often show considerable continuity.¹
This is one reason professionals should be cautious about treating a single score as a fixed forecast of future functioning. A score can provide useful information about performance at the time of testing, but interpretation should include age, developmental history, education, health, and the reason for assessment.
Practice effects may also influence repeated cognitive testing. A person may perform better simply because the testing process is more familiar. Repeated assessments should therefore use appropriate intervals, alternate forms when available, and interpretation methods suited to the instrument.
Are personality scores stable?
Personality traits generally show both continuity and change. Research has found that personality stability tends to increase with age, while personality traits can still develop across adulthood.²
This means a personality assessment may be useful for describing relatively consistent patterns without implying that those patterns are unchangeable. Someone’s characteristic tendencies may remain recognizable while also shifting in response to maturity, relationships, work, health, or other life circumstances.
The length of the interval matters. A retest after several days asks a different question from a retest after ten years. Over a short period, differences may reflect mood, context, or measurement error. Over a long period, differences may reflect both measurement issues and genuine development.
Personality inventories also differ in what they measure. Broad traits may be more stable than narrow states or temporary conditions. A measure of current anxiety, for example, may reasonably fluctuate more than a measure designed to assess a broad personality trait.
Are symptom and mental-health scores stable?
Scores for symptoms such as anxiety, depression, stress, pain, or trauma-related distress are often expected to change. These assessments commonly measure current or recent experiences rather than fixed characteristics.
A change in score may reflect:
actual change in symptoms;
changes in sleep, health, stress, or medication;
differences in the person’s interpretation of questions;
changes in willingness to report symptoms;
differences in the testing setting; or
ordinary measurement error.
For this reason, symptom scores are often most useful when interpreted as part of a broader pattern. A single score should not be treated as a complete account of someone’s mental health, and a change in score should not automatically be interpreted as treatment success or failure.
When repeated measurement is used, the important question is often whether the change is larger than what could reasonably be expected from measurement error. This is sometimes examined with a reliable change index, although the appropriate method depends on the instrument, its reliability, the available norms, and the purpose of testing.
What is the difference between reliability and stability?
Reliability is a broad term that describes the consistency of measurement. It can refer to several different forms of consistency, including:
Internal consistency: how closely related items are within one administration;
Interrater reliability: how similarly different raters score or judge a performance;
Parallel-forms reliability: how similarly equivalent versions of a test perform; and
Test–retest reliability: how consistently scores are obtained across time.
A test may have strong internal consistency without producing highly stable scores over time. For example, its items may measure the same current state very consistently, while that state itself changes from week to week.
Conversely, a measure may show moderate internal consistency but still provide useful information for a particular purpose. Reliability should therefore be considered in relation to the assessment’s intended use.
Does a high reliability coefficient mean that a score will not change?
No. A reliability coefficient describes consistency under particular conditions. It does not guarantee that an individual’s score will remain unchanged.
A person’s observed score can vary even when the assessment is well designed. Also, a genuinely changing characteristic can be measured reliably. For example, an assessment may detect changes in symptoms accurately, even though those symptoms are not expected to remain constant.
Reliability is best understood as a property of scores produced under specified conditions, not as a permanent property of a test in every situation. The Standards for Educational and Psychological Testing emphasize that reliability and precision should be evaluated for the particular interpretation and use being made of test scores.³
How does the time interval affect stability?
The retest interval can strongly influence the result.
Very short intervals
A short interval may increase similarity because the person and circumstances have not changed much. However, memory and familiarity can also raise later scores or otherwise influence responses.
Moderate intervals
An interval of several weeks or months may reduce memory effects while still allowing meaningful comparison. Whether this is appropriate depends on the assessment and the characteristic being measured.
Long intervals
Over longer periods, genuine development, treatment, aging, education, health changes, and life events become more relevant. A lower correlation over a long interval does not necessarily mean that the assessment is poor. It may indicate that the construct or the person has changed.
There is no single ideal retest interval for every assessment. The interval should fit the purpose of testing and the expected rate of change in the characteristic being measured.
How much score change is meaningful?
A numerical difference is not automatically a meaningful difference.
Professionals may consider:
the assessment’s reliability;
the standard error of measurement;
the size of the change;
the time between assessments;
changes in the person’s circumstances;
whether the same procedures were used; and
whether the change has practical or clinical significance.
The standard error of measurement estimates the expected amount of uncertainty around an observed score. A commonly used expression is:
SEM=SD1−r
where SD is the standard deviation of scores and r is a reliability estimate. The exact interpretation depends on the reliability coefficient and the assessment’s scoring model.
A confidence interval based on the standard error can communicate that uncertainty more clearly than a single number. For example, a score should generally be treated as an estimate within a range rather than as a perfectly precise measurement.
For repeated scores, a professional may also examine whether the difference exceeds the amount expected from measurement error. That calculation requires information specific to the instrument and should not be improvised from a general rule.
Can a score become more stable with age?
For some characteristics, stability tends to increase across development. This pattern has been observed in research on personality traits, where rank-order stability generally rises from childhood toward adulthood.²
However, “more stable” does not mean “unchangeable.” Adult personality traits can continue to develop, and individual patterns vary. The same caution applies to cognitive and symptom measures: age can affect stability, but it does not determine a person’s result by itself.
Interpretation should always consider the specific construct, the person’s developmental stage, and the assessment’s technical evidence.
What role do norms play?
Norms allow an individual score to be interpreted in relation to a comparison group. If the person’s score changes over time, the meaning of that change may also depend on which norms are used.
For example, a raw score may stay the same while a person’s percentile changes if the relevant comparison group differs. Conversely, a raw score may change while the person’s relative standing remains similar.
Professionals should use the norms and scoring procedures appropriate for the person’s age, language, education, culture, and assessment purpose. Norms do not remove measurement uncertainty, but they help place scores in context.
Can culture, language, or accommodations affect score comparisons?
Yes. Comparisons across time are less straightforward when the language, cultural context, instructions, accommodations, or testing format changes.
A change may reflect differences in access to the assessment rather than a change in the underlying characteristic. This is especially important when a person completes different language versions, uses different accommodations, or moves between paper and digital formats without evidence that the versions are comparable.
Fair interpretation requires attention to whether the assessment is appropriate for the person and whether the score meaning is comparable across administrations.
Should the same assessment be taken repeatedly?
Not always. Repeated testing can be useful when monitoring change, making decisions, or evaluating an intervention, but frequent testing may introduce memory and practice effects.
The appropriate plan depends on:
the assessment’s purpose;
the expected rate of change;
available alternate forms;
the recommended retest interval;
the importance of detecting small changes; and
the consequences of an incorrect interpretation.
Some situations call for repeated measurement. Others are better served by a single assessment combined with interviews, records, behavioral information, or reports from relevant people.
How should people interpret two different scores?
Start by asking whether the difference is larger than expected from ordinary measurement uncertainty. Then consider the circumstances surrounding each administration.
Useful questions include:
Was the same assessment used?
Were the same instructions and time limits used?
Was the person similarly rested, healthy, and motivated?
Was the interval long enough for real change?
Did treatment, education, stress, or life circumstances change?
Are the scores being compared using the same norms?
Does the assessment manual provide a standard method for interpreting change?
A modest difference may have little practical meaning. A larger difference may be important, but it still requires context. A score should not be interpreted in isolation when important decisions are at stake.
What is the most balanced conclusion?
Psychometric scores are often stable enough to be useful, but rarely so fixed that change should be ignored.
Stability is usually stronger when:
the characteristic is relatively enduring;
the assessment is reliable;
the retest interval is appropriate;
testing conditions are comparable;
the person’s circumstances have not changed substantially; and
the interpretation uses suitable norms and technical guidance.
Stability is usually weaker when:
the assessment measures a temporary state;
the person is undergoing substantial development or change;
the retest interval is very short or very long for the purpose;
practice effects are likely;
testing conditions differ; or
the score is close to the limits of the instrument’s precision.
The most useful approach is neither to dismiss a score as meaningless nor to treat it as permanent. A score is a carefully qualified estimate. Its value comes from combining the number with the assessment’s evidence, the person’s circumstances, and the question the assessment was designed to answer.
Notes
Ian J. Deary, Steve Strand, Pauline Smith, and Chloe Fernandes, “Intelligence and Educational Achievement,” Intelligence 35, no. 1 (2007): 13–21.
William W. Roberts and Wendy F. DelVecchio, “The Rank-Order Consistency of Personality Traits from Childhood to Old Age: A Quantitative Review of Longitudinal Studies,” Psychological Bulletin 126, no. 1 (2000): 3–25.
American Educational Research Association, American Psychological Association, and National Council on Measurement in Education, Standards for Educational and Psychological Testing (Washington, DC: American Educational Research Association, 2014).