Four families of evidence to consider together
No single statistic demonstrates validity. The argument is built from complementary findings and explicit hypotheses.
Internal structure
Relationships between items, tasks, or subscores should be consistent with the stated theoretical organization, although statistical fit alone is not enough.
Convergent evidence
The score should show the expected relationships with other measures, observations, or situations that illuminate a related construct.
Discriminant evidence
It should also remain distinct from neighboring concepts. Overly broad correlations may indicate that the interpretation lacks specificity.
Population and use
Evidence concerns a specific version, language, population, and use; its transfer should be studied rather than assumed.
Why is this concept central to psychometrics?
It shifts attention from the score itself to the quality of the inferences we want to draw from it.
It requires distinguishing what is directly observed, a response or performance, from the proposed theoretical explanation.
It makes competing explanations visible: a task can involve several abilities, strategies, or contextual constraints.
It reminds us that the data required depend on the intended interpretation and decision, not only on an instrument's reputation.
From a score to a network of evidence: a simple conceptual outline
The table provides a textual illustration of the path from observation to interpretation. It does not replace a validation study.
Construct
- What it refers to
- A theoretical concept, such as fluid reasoning.
- Useful question
- What exactly are we trying to describe?
- Do not confuse with
- A single item or score.
- Limitation
- The construct must be defined and examined.
Indicator
- What it refers to
- A performance, response, or observation collected using a tool.
- Useful question
- What was observed, and under what conditions?
- Do not confuse with
- The concept itself.
- Limitation
- A task can involve multiple processes.
Nomological network
- What it refers to
- Expected relationships with other concepts, measures, and situations.
- Useful question
- What convergences or divergences should emerge?
- Do not confuse with
- A simple list of correlations.
- Limitation
- Hypotheses must be explicit.
Interpretation
- What it refers to
- The limited conclusion that the evidence supports for a particular use.
- Useful question
- What can reasonably be said, and what must remain open?
- Do not confuse with
- A definitive truth about a person.
- Limitation
- Context and limitations remain necessary.
Educational example: measuring reasoning without reducing it to one task
This example is fictional and does not describe any real test. It shows why an isolated piece of evidence is not enough.
The intended concept
- Purpose
- Describe reasoning about a new problem as a theoretical dimension.
- Audience
- Research or assessment that clearly states this objective.
- Context
- The concept must be distinguished from knowledge, speed, memory, or motivation.
- Professional
- The definition belongs to the theoretical and methodological framework.
- Limitations
- The word “reasoning” can have different meanings depending on the situation.
The indicators collected
- Purpose
- Collect several relevant performances or observations under documented conditions.
- Audience
- The population and context for which the tool is studied.
- Context
- Instructions, language, time, and familiarity with the task can influence responses.
- Professional
- The administration framework depends on the tool.
- Limitations
- A single indicator can be affected by factors that are not the intended focus.
The cautious conclusion
- Purpose
- State what the evidence can illuminate without generalizing beyond the use studied.
- Audience
- Any reporting or communication of results.
- Context
- Convergences, divergences, and limitations must be made visible.
- Professional
- An individual decision requires the appropriate framework.
- Limitations
- Validity supports an interpretation; it does not sum up a person.
Approach
What Cereya does, and what Cereya does not do
The logic of convergence helps explain how Cereya structures multidimensional guidance; it is not evidence of the product's validity.
What Cereya does
- Offer an accessible first step toward finding direction, available immediately.
- Organize several indicators and observations to identify convergences rather than an isolated sign.
- Help formulate questions worth exploring further and prepare for the next step.
What Cereya does not do
- Validate an official test, a translation, or a particular use.
- Infer a profile, condition, or diagnosis from a score.
- Present an isolated correlation as definitive evidence.
How should validity evidence be read?
The useful question concerns the strength of a specific interpretation more than a general verdict.
Identify the intended construct and the hypotheses formulated before examining the results.
Look for several types of evidence, including findings that could contradict the interpretation.
Check the population, version, context, and use to which the data relate.
Distinguish what is documented from what remains debated or unstudied.
Construct validity, reliability, and diagnostic validity
These concepts address different questions and cannot be used as synonyms.
Reliability concerns the consistency or precision of a measure; it does not establish what the score represents.
Construct validity examines the meaning attributed to a score through a network of theoretical and empirical evidence.
Diagnostic validity concerns a tool's ability to support clinical classification, often studied through sensitivity, specificity, and comparison populations.
A tool can have construct evidence without being designed for diagnosis, and an isolated correlation is insufficient for any of these conclusions.
Assessment limitations
A historical concept, extended and debated
Cronbach and Meehl provided an enduring language for thinking about constructs and networks of evidence. Contemporary frameworks have since expanded and refined how validity evidence is gathered, particularly around the interpretation and use of scores. This development calls for avoiding two extremes: treating a test as inherently valid or believing that a single study can settle the question definitively.
- The nomological network is a reasoning tool; it does not remove the need for high-quality data or discussion of alternative explanations.
- Validity evidence may vary across populations, languages, versions, and proposed decisions.
- A psychological construct remains a theoretical representation, not an essence directly observable in a person.
- Validity does not replace questions of fairness, accessibility, reliability, or the consequences of use.
Scientific and institutional references
Primary publications and reviews that put the foundational text and its contemporary developments in context.
- Cronbach, L. J. & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52(4), 281–302. DOI : 10.1037/h0040957.View source
- Cronbach, L. J. & Meehl, P. E. (1955). Notice bibliographique PubMed, PMID: 13245896.View source
- Messick, S. (1995). Validity of psychological assessment. American Psychologist, 50(9), 741–749. DOI : 10.1037/0003-066X.50.9.741.View source
- Kane, M. T. (2013). Validating the interpretations and uses of test scores. Journal of Educational Measurement, 50(1), 1–73. DOI : 10.1111/jedm.12000.View source
- American Educational Research Association, American Psychological Association & National Council on Measurement in Education (2014). Standards for Educational and Psychological Testing.View source
Last editorial update: August 17, 2026.





