RESEARCH DOSSIER
Developmental Measurement and Functional Equivalence
Abstract. Developmental measurement investigates whether psychological scores and observations sustain valid interpretations as task comprehension, behavioral expression, reporting systems, and developmental contexts change across the life course. This fundamental critical narrative monograph synthesizes construct-validation theory, homotypic and heterotypic continuity, longitudinal factorial invariance, differential item functioning, item response models, functional equivalence, developmental scaling, and cross-age assessment. Evidence includes foundational psychometric work and longitudinal investigations of changing depressive-symptom measures, youth-to-adult symptom inventories, early childhood psychopathology, and multi-informant agreement, as well as recent research on item banks. The synthesis separates score comparability, functional similarity, predictive validity, and mechanistic continuity; examines sources of age-dependent bias; and establishes design and reporting requirements for defensible developmental conclusions. It identifies limits in cross-cultural calibration, individual change, response shift, adaptive digital assessment, and transportability. The study is a source-grounded integrative monograph, without newly collected data or a systematic review protocol.
Keywords: developmental measurement; functional equivalence; measurement invariance; heterotypic continuity; developmental psychometrics; differential item functioning; longitudinal assessment; developmental scaling; item response theory; multi-informant assessment; construct validity; response shift; age norms; item banks; lifespan psychology; Artificial Era
Research Object and Scientific Scope
Developmental measurement is the practice of constructing, administering, and interpreting observations of psychological functioning as capacities and contexts change across the life course. Functional equivalence concerns whether age-differing behaviors, response modes, tasks, or informant reports can support a common inference about a psychological function. The principal research problem is inferential: an observed difference between a toddler, adolescent, and older adult could indicate development, altered task demands, changed access to a response, a shift in social opportunity, or a transformed meaning of the behavior. To distinguish these possibilities, a researcher must specify the construct, its expected function, the observation process, and the population within which a comparison will be interpreted. The conceptual foundations of construct validation make score meaning a property of justified inference rather than a property automatically attached to a test Cronbach & Meehl (1955) Messick (1995).
The object examined here includes age-linked task comprehension, motivation, language and motor requirements, observer access, symptom presentation, score scaling, developmental norms, and the modeling of latent change. It encompasses observational and performance measures as well as questionnaires and interviews. Functional equivalence requires consideration of behavior and mechanism: a child's protest and an adolescent's verbal refusal might serve a comparable regulatory function under specific conditions, yet the similarity must be substantiated by converging evidence. The same overt act may serve different functions at different ages. Contemporary analysis of developmental continuity therefore asks investigators to align behaviors, functions, and candidate mechanisms rather than assigning continuity from labels alone Petersen (2024).
This monograph is a critical integrative research synthesis. Its evidence base combines original measurement theory, developmental longitudinal investigations, multi-informant studies, empirical applications of measurement invariance, and recent work on linked item banks. It reports no newly collected participant data, pooled effect-size estimate, or pre-registered systematic review. The central evaluation separates evidence that a measurement model fits, evidence that it functions comparably, evidence that a score predicts external outcomes, and evidence that an underlying process is shared. These form distinct warrants. A model may fit each age separately while its intercepts or thresholds differ; an equivalent prediction may arise from different cognitive strategies. Interpretations must therefore remain matched to the actual tests performed Widaman et al. (2010) Putnick & Bornstein (2016).
Conceptual Foundations of Valid Developmental Inference
A psychological construct is a theoretically articulated attribute or process whose interpretation depends on a network of expected relationships with observations, interventions, developmental precursors, and consequences. Cronbach and Meehl's account of construct validity established the importance of testing that network rather than relying on superficial agreement between a test name and an inferred capacity. A measure of inhibitory control, for example, can depend on working memory, rule comprehension, visual discrimination, motor speed, and willingness to cooperate. Changes in these supporting operations may improve scores even when the target control process changes little. Construct definition therefore precedes instrument selection, and the proposed construct-to-score mapping must be revisited when developmental conditions alter the cognitive work required Cronbach & Meehl (1955) Campbell & Fiske (1959).
Validity is specific to interpretations and uses. Messick's unified account joins structural, substantive, external, and consequential evidence to the question of what a score supports. For developmental research, substantive evidence includes response-process studies demonstrating how children understand instructions, what adults attend to during an assessment, and how task strategies vary with skill. External evidence includes developmental prediction, convergent findings from distinct methods, and discriminatory relations with neighboring constructs. Consequential evaluation examines whether a norm or cutoff produces systematically misleading classifications for populations whose access, language, or opportunities differ from those of the calibration sample. These requirements are compatible with published testing standards, which organize validation around intended interpretation and use Messick (1995) AERA, APA, & NCME (2014).
Functional equivalence is best formulated as a testable relation among observed behaviors and an explicitly defined function under specified conditions. It does not assert identity of surface form, latent mechanism, or evolutionary origin. One operational strategy is to define a functional criterion independently, such as regulation of approach toward a goal under competing demands, and then evaluate different age-appropriate indicators against that criterion. Another is to construct a common latent scale from overlapping anchors and estimate whether age-specific items contribute comparable information. The first addresses function and explanatory context; the second addresses statistical comparability. Their conclusions complement one another because a common score metric cannot by itself demonstrate shared mechanism, and a compelling functional analogy cannot establish numerically interchangeable scores Petersen (2024) Millsap (2011).
Historical Development of the Measurement Problem
Questions about comparable psychological measurement developed alongside factor analysis, classical test theory, psychometrics, and developmental observation. The early statistical approach emphasized standardization, reliability, and norms, which made aggregate comparison possible while leaving open the meaning of a score across different developmental periods. Mid-twentieth-century construct validation clarified the need to justify psychological interpretation through theory and multiple forms of evidence. Campbell and Fiske's multitrait–multimethod approach then offered an influential logic for distinguishing variation attributable to traits from variation attributable to the mode of assessment. This shift made observer and method effects part of the scientific question rather than incidental imperfections to be ignored Cronbach & Meehl (1955) Campbell & Fiske (1959).
Later research formalized the demand for invariance. Horn and McArdle's analysis of aging linked valid inference about changing intellectual ability to questions about whether factor structures and their measurement properties remain comparable across age groups. Meredith's 1993 formal treatment developed conditions for factorial invariance and the statistical meaning of measurement bias. Vandenberg and Lance subsequently synthesized applied testing procedures for comparing groups and occasions. This tradition made explicit that matching instrument wording is not sufficient: different ages can attach different meanings or response probabilities to identical items, and different instruments may sometimes be aligned through shared latent parameters Horn & McArdle (1992) Meredith (1993) Vandenberg & Lance (2000).
Developmental psychology brought these issues into longitudinal research, where the substantive object is change itself. Widaman, Ferrer, and Conger demonstrated how factor invariance can be examined inside longitudinal structural equation models. Putnick and Bornstein later documented considerable variation in reporting and testing conventions across published psychological research. Methodological progress has therefore involved both stronger quantitative models and stricter interpretation: investigators must document which constraints were examined, why the chosen model represents the construct, and which comparisons remain warranted after noninvariance. Recent linked-measure studies reopen an older developmental dilemma by investigating how changing instruments can preserve meaningful longitudinal inference Widaman et al. (2010) Putnick & Bornstein (2016) Tyrell et al. (2019).
Homotypic and Heterotypic Continuity
Developmental continuity can concern persistence of observable form, persistence of function, or continuity of underlying processes. Homotypic continuity generally concerns the recurrence of similar behavioral manifestations. Heterotypic continuity concerns the relation of differing manifestations across development that may express a related organization or function. These distinctions are research hypotheses, and simple predictive association does not settle the matter. A preschool behavior predicting adolescent functioning could reflect stable predisposition, common environments, a developmental cascade, or measurement artifact. Petersen's contemporary review argues for explicit testing of individual behaviors, the functions they serve, and the cognitive or biological processes connecting them Petersen (2024).
An equivalent function may be expressed through distinctly age-specific forms. Infants communicate attention and distress through gaze, vocalization, and action; older children may use verbal descriptions and rule-guided choices; adults may report internal states whose meaning reflects accumulated experience. A developmental study of emotion regulation must therefore decide whether it is following the same response, the same adaptive goal, or an inferred regulatory capacity. Each requires a different design. Behavior sampled in an environment with abundant adult scaffolding may show adaptive regulation that disappears in an independent task, even when the person possesses the necessary skill. Assessment must describe the task environment alongside the respondent's developmental status Petersen (2024) AERA, APA, & NCME (2014).
The methodological challenge is particularly visible when researchers preserve identical items across time merely for convenience. An item can become trivial, unfamiliar, or culturally displaced as the respondent develops. Conversely, age-adjusted items may increase validity for each stage but complicate longitudinal linking. Research using developmentally differentiated depressive-symptom scales has shown that common and unique items can sometimes be combined to test longitudinal invariance. This result supports a practical middle path: maintain defensible anchors while allowing age-specific indicators to represent evolving manifestations. Equivalence remains an empirical outcome of that model and its external validation, never an automatic consequence of attaching the same construct label to two tests Tyrell et al. (2019) Petersen et al. (2026).
Developmental Task Comprehension and Response Processes
Every measurement instrument recruits abilities beyond its nominal target. Instructions require perception, language comprehension, memory for rules, familiarity with examination conventions, and sometimes the capacity to make comparative judgments. In childhood, these prerequisites change rapidly; at later ages, sensory or motor changes may alter access to otherwise familiar tasks. Researchers can isolate these sources by using comprehension checks, interviews, graded practice trials, response-time decompositions, and alternate input modes. The scientific objective is to identify which operations are necessary for a valid response at each age, not to assume that a standardized instruction operates identically for all respondents. A theoretically sound instrument may still fail its intended interpretation if the response process differs systematically from the one used for validation AERA, APA, & NCME (2014) Messick (1995).
Age changes the opportunity to exhibit an ability as well as the ability itself. Recalling spoken words depends partly on hearing and vocabulary; selecting an answer on a touchscreen depends partly on motor precision and prior device exposure. A child who understands a social scenario may not possess the linguistic resources to explain it, while an older adult may answer accurately after a longer processing interval. Adaptive timing, alternate response channels, and calibrated difficulty can improve accessibility, but every adaptation requires evidence that the relevant construct remains represented. Testing the new modality against independent criteria and quantifying whether item parameters move are essential parts of demonstrating comparability Horn & McArdle (1992) Domingue (2026).
Cognitive interviews and observational process data make developmental assumptions visible. Investigators can ask older children to paraphrase items, study preverbal children's gaze and behavior, or compare alternative task versions within randomized designs. Such studies should record the developmental and contextual characteristics that shape comprehension, including language exposure, schooling, disability, and fatigue. If understanding differs across groups, a factor-analytic solution will not repair the interpretative problem automatically. The response-process evidence must influence the construct model and item set, after which quantitative linking can assess what remains comparable. Multiple methods are especially important when a single method confounds task mastery with the psychological phenomenon under investigation Campbell & Fiske (1959) AERA, APA, & NCME (2014).
Measurement Models and Levels of Invariance
Under a common-factor model, an observed response is represented by a relation between an underlying factor, an item loading, an item intercept or threshold, and residual variation. Configural invariance indicates a compatible broad pattern of relations between items and factors. Metric invariance constrains the factor loadings and supports particular comparisons of associations or slopes on a common scale. Scalar invariance further constrains intercepts for continuous indicators or appropriate thresholds in categorical models, enabling defensible comparisons of latent mean levels under the model. Residual invariance imposes stronger constraints on residual variation. Each claim is conditional on the specification, identification, estimation method, and fit of the model Meredith (1993) Widaman et al. (2010).
Testing proceeds by comparing a theoretically justified series of increasingly constrained models, but the hierarchy is not a mechanical ritual. An estimated change in fit depends on sample size, item distributions, number of groups, estimation method, and pattern of departures from invariance. Chen's simulation research found different fit indices respond differently to violations in loadings, intercepts, and residuals; it also showed that the distribution of noninvariant items matters. Putnick and Bornstein describe reporting conventions and unresolved issues rather than one universally applicable threshold. Investigators should provide full model specifications, parameter estimates, uncertainty, localized diagnostics, and an explanation of which inference changes if a constraint is relaxed Chen (2007) Putnick & Bornstein (2016).
Longitudinal applications require additional attention to repeated observations and dependence. The same respondent contributes correlated measurements across occasions, and stable item-specific influences may persist even after controlling the target latent factor. A repeated-item design can require correlated uniquenesses, appropriate ordinal thresholds, or time-specific nuisance factors. When the tested indicator's meaning changes continuously with age, the choice of widely separated age bands can hide transitional patterns; unnecessarily narrow bands can weaken estimation. Models can incorporate moderators of item properties and explicit temporal variation, provided sufficiently informative designs support the additional parameters. The inferential target is valid developmental comparison, and fit statistics alone are subordinate to that target Widaman et al. (2010) Chen & Bauer (2024).
Differential Item Functioning and Measurement Bias
Differential item functioning, or DIF, occurs when respondents with comparable positions on a specified latent attribute have different conditional probabilities of responding to an item because they belong to different groups or measurement circumstances. In developmental research the grouping variable can be age, pubertal status, schooling stage, linguistic context, cohort, or testing mode. Uniform DIF changes item difficulty or threshold systematically; nonuniform DIF alters discrimination or a related interaction. DIF is evidence about the measurement relation, not a diagnosis of unfair intention or proof that the construct itself differs. Researchers must distinguish item-level departures from broader construct heterogeneity and from genuinely different distributions of the attribute across populations Millsap (2011) Domingue (2026).
Item response theory permits developmental comparisons through explicit models of item characteristic curves and information across levels of an underlying trait. Unlike simple total-score comparisons, such models can evaluate whether a short task is informative for one age and saturated for another. Common-item designs can connect different instruments, while common-person designs allow participants to take overlapping forms; both require careful examination of anchor stability and population assumptions. Linking may produce numerically comparable scores even while some domains receive weaker content coverage in one age group. Item selection thus combines psychometric efficiency with theoretical coverage and transparent evidence about subgroup functioning Millsap (2011) Domingue (2026).
Biased anchors can compromise an entire linking strategy. If the items treated as common have acquired new meanings through schooling, culture, or maturation, parameter calibration may make differences appear to be latent growth. Anchor purification, sensitivity analyses, alternative anchor sets, and external outcomes help detect this problem. Distinguishing DIF from meaningful developmental change also requires substantive investigation: an item about independent travel may legitimately become relevant only after autonomy increases, so forcing strict equality may erase a real structural change. The appropriate result might be a bounded comparison, a stage-specific scale, or an explicitly linked metric with restricted interpretations Tyrell et al. (2019) Petersen et al. (2026).
Latent Growth and Distorted Developmental Trajectories
A developmental trajectory estimated from observed sum scores mixes changes in the target attribute with possible changes in item sensitivity, response standards, and task demands. A rise in average responses can occur if a question becomes easier to endorse, even without an underlying latent increase. Widaman and colleagues recommend measurement models with multiple indicators across occasions so that the scale of the latent factor can be evaluated. The subsequent growth process can then be defined over latent scores rather than presupposing that repeated sums share an unchanging metric. This distinction matters whenever researchers interpret the timing, magnitude, or shape of individual development Widaman et al. (2010) Horn & McArdle (1992).
Chen and Bauer's longitudinal moderated factor analysis approach addresses potential changes in measurement while modeling construct development. It emphasizes the risks of growth models fitted to sums or means when items exhibit time-related DIF. Moderated measurement parameters offer a means to investigate age or context effects continuously, though estimation and model identification must be considered. This research does not imply that every developmental process can be reduced to one continuous score; it instead clarifies when numerical growth estimates require measurement qualifications. If a construct changes qualitatively, the underlying model itself may require a different representation before growth parameters can be meaningful Chen & Bauer (2024) Petersen (2024).
A rigorous longitudinal analysis also separates age, calendar time, repeated exposure, and selective retention. Reusing the same assessment may induce learning or memory for items; changing instruments can reduce practice effects while introducing new linking uncertainty. Participants who remain in a study may differ from those who leave, and item functioning can vary with those differences. Investigators should jointly examine missingness assumptions, attrition patterns, occasion-specific methods, and the sensitivity of conclusions to alternative measurement structures. A smooth growth curve supplies a compact statistical description; an explanatory account requires additional temporal and causal evidence identifying why a change occurred Widaman et al. (2010) Moriarity et al. (2025).
Multi-Informant Assessment and Contextual Specificity
Developmental assessment often combines reports from parents, teachers, clinicians, peers, and young people themselves because behavioral opportunities differ across home, school, and other settings. These observers sample partly different events and apply different comparison standards. Achenbach, McConaughy, and Howell's 1987 synthesis reported stronger agreement between similar informants than between different informant classes. The study established that disagreement cannot be treated simply as interchangeable noise when settings and perspectives differ. Modern interpretation distinguishes limitations in observation, measurement method, contextual variation, and potential discrepancies in the actual behavior being sampled. The construct needs an explicit observational frame, such as classroom disruption or distress across contexts, before informant agreement can be judged Achenbach et al. (1987) De Los Reyes et al. (2015).
De Los Reyes and colleagues' 2015 review develops an evidence-based rationale for integrating multiple perspectives rather than assuming one informant is the definitive measurement source. Reports become especially complex across development because parental access to private experiences declines, self-report abilities improve, teachers change, and the setting in which the behavior occurs evolves. A construct measured by parent report during childhood and self-report during adolescence changes both the respondent and the observation system. Method factors, latent trait–state models, and contextually anchored validation can help determine whether longitudinal differences represent psychological development, informant substitution, or a combination. Claims based on cross-informant averages should disclose their weighting and the justification for combining information De Los Reyes et al. (2015) Campbell & Fiske (1959).
Disagreement can contain substantive information when behavior is context dependent. A child may manage frustration effectively at school while struggling in unstructured family settings; a parent and teacher can both describe their observations accurately. Psychometric validation should therefore test convergence within shared contexts and discriminant patterns across contexts, rather than demand perfect agreement everywhere. Multi-informant invariance also requires asking whether the item has the same referent to each respondent. Wording such as often, rarely, and severe may invoke different opportunity counts and expectations. Age-linked changes in access and interpretation become part of a developmental research design, with carefully defined estimands for within-person and across-setting change Achenbach et al. (1987) De Los Reyes et al. (2015).
Adolescence-to-Adulthood Harmonization Evidence
One direct test of developmental comparability involved five measurement waves from early adolescence through early adulthood. Tyrell and colleagues examined depressive symptoms across changing instruments in 392 families, including European American and Mexican American participants. The investigators used common and unique items to evaluate whether different age-appropriate scales could support longitudinal comparison. Their analysis found evidence for strong invariance within the individual measures and partial strong invariance across waves. The contribution is methodological and substantive: instrument modification can be accommodated when empirical anchors and a defensible measurement model support the interpretation, rather than assuming that any change in questionnaires destroys comparability Tyrell et al. (2019).
The sample composition, attrition, scale content, and identification strategy matter for interpreting that finding. An invariance result within a particular developmental and population range does not demonstrate transportability to every later adult age or to all symptom dimensions. Item parceling, used in the analysis, can stabilize estimation but may conceal heterogeneous functioning among constituent items. A stronger replication program would compare item-level and parcel-level results, test independent cohorts, and examine whether linked scores predict functionally relevant outcomes consistently across transition periods. Such triangulation is important when the construct's behavioral expression changes in ways that originally motivated the use of different questionnaires Tyrell et al. (2019) Putnick & Bornstein (2016).
A further longitudinal study by Moriarity and colleagues evaluated the Achenbach youth and adult self-report systems across the transition to adulthood in a sample of 1,309 people followed across six waves, with mean ages approximately 11 and 26 years at the endpoints. The findings differed by subscale: evidence was stronger for some anxiety, attention-deficit/hyperactivity, and cognitive depressive-symptom dimensions, and weaker for particular somatic depressive-symptom dimensions. The study shows why invariance must be assessed by scale and purpose rather than assigned to an instrument family in general. Its sampling frame, Dutch setting, and specific report forms delineate the populations and constructs to which the findings directly speak Moriarity et al. (2025).
Developmental Scaling With Changing Symptom Indicators
Petersen, Demko, Lee, and Oleson's 2026 longitudinal study operationalized heterotypic continuity in early childhood by comparing ways of selecting and linking age-specific indicators. The sample comprised 231 children observed across seven occasions from ages three to seven and a half. Externalizing, internalizing, and thought-disordered dimensions were assessed through reports from mothers, fathers, and teachers or caregivers. The authors compared common-item, upward-or-downward extension, and construct-valid item-selection approaches; they also compared average scoring with developmental scaling. This constitutes an unusually direct empirical test of how measurement architecture alters developmental inference, rather than merely demonstrating an abstract problem with changing instruments Petersen et al. (2026).
For externalizing behavior, the study used observations of compliance and task attention as external criteria. Age-appropriate construct-valid items showed modestly greater criterion validity than the alternative item-selection strategies, with reported differences in correlation of approximately 0.07 to 0.13. Developmental scaling also showed modestly greater criterion validity than average scoring, with correlation differences around 0.03 to 0.17. These values describe particular contrasts in this longitudinal design and should not be read as universal improvements in developmental measurement. The investigation's strongest contribution is the integrated design: item validity, common scale construction, multiple informants, and external observations were examined together Petersen et al. (2026).
The results also illustrate why validation must continue beyond fit to a latent measurement model. A scale can connect scores across age periods yet perform poorly against observed behavior, and item selection can optimize within-age description without supporting longitudinal linking. By testing both dimensions, the study offers an empirical basis for joining developmental sensitivity with quantitative comparability. Replication is needed in independent samples, later life stages, other construct domains, and different cultural or educational settings. Its evidence supports the feasibility of developmental scaling in the examined context; it does not establish a universal method for all changing constructs or causal equivalence of the behaviors involved Petersen et al. (2026) Petersen (2024).
Interpretation of Conflicting Empirical Findings
Apparently conflicting results in invariance studies can arise because researchers test different levels of equality, use different sample populations, analyze different items, or allow different nuisance parameters. One investigation may report an acceptable approximate fit under a metric model while another concludes scalar invariance fails. These outcomes need not contradict one another if the first concerns associations and the second concerns latent mean differences. Similarly, a model with a small number of freed intercepts may support partial mean comparability under qualified assumptions, whereas an unmodified sum score remains contaminated. The correct unit of synthesis is the particular inference each model licenses, with attention to parameter-level diagnostics and substantive significance Putnick & Bornstein (2016) Chen (2007).
Evidence quality also varies between controlled calibration studies, observational cohorts, clinician-rated outcomes, and theoretical demonstrations. A Monte Carlo study evaluates model behavior under specified simulated conditions; it cannot establish that a developmental construct is truly invariant in a population. A cohort investigation can show measurement comparability across its observed interval, while unmeasured time-varying confounding and selective participation may limit explanatory conclusions. A review may organize evidence or propose standards without itself providing direct replication. Preserving these distinctions prevents a methodological possibility from being misreported as a settled empirical property of a specific psychological process Chen (2007) Moriarity et al. (2025).
An integrated evidence assessment examines construct definition, item selection, external validation, sample transportability, and the consequences of alternate scoring decisions. When a model produces stable results after plausible changes in anchors, estimator, or grouping strategy, confidence in the restricted comparison increases. When conclusions reverse under mild changes, the uncertainty belongs in the substantive report. The developmental researcher should describe the extent of comparability rather than assign a binary pass or fail to an entire instrument. This approach also clarifies what independent replication would have to reproduce: specific score interpretations under stated developmental and contextual conditions, not merely a favorable global fit statistic Petersen et al. (2026) Putnick & Bornstein (2016).
Age Norms, Reference Groups, and Developmental Meaning
Norm-referenced scores locate an observation within a specified reference distribution, whereas criterion-referenced scores express performance against a defined capability or standard. Developmental age norms are useful for describing relative standing at a particular time, but their interpretation changes when the reference population changes. A percentile obtained at age five and the same percentile at age ten indicate a similar rank position within different age distributions; they do not imply unchanged absolute functioning. Norms additionally depend on sample recruitment, calendar period, language, schooling, disability inclusion, and statistical smoothing procedures. The measurement specification must distinguish chronological age, developmental opportunity, and the scale on which the observed response has been calibrated AERA, APA, & NCME (2014) Horn & McArdle (1992).
Developmental screening, diagnostic assessment, and longitudinal research pose different inferential demands. Screening may prioritize sensitivity within an age band and accept uncertain identification requiring follow-up; research on individual change requires a stable or defensibly linked metric; diagnostic interpretation demands evidence about the consequences of classification. A cutoff calibrated in one population cannot simply be transported to a different generation or educational context without checking its functioning. When distributions change because opportunities or health conditions change, apparent differences in prevalence can reflect altered norms or response procedures. Technical manuals should therefore report norming dates and population coverage, and researchers should avoid treating normative rank as a directly measured developmental mechanism Messick (1995) AERA, APA, & NCME (2014).
Age grading also raises questions about timing and within-age heterogeneity. Two children of the same chronological age may differ substantially in language exposure, access to instruction, health, or opportunities to practice an assessed task. Broad age norms can conceal meaningful differences in developmental experience; extremely narrow bands can yield unstable estimates. A defensible approach combines adequate sample sizes, flexible models of age, reporting of conditional standard errors, and sensitivity analyses for population composition. Where equitable comparison requires age-specific adaptations, researchers should demonstrate which measurement properties remain shared. Functional equivalence pertains to the interpretation of performance, and norms are one component of that argument rather than its substitute Domingue (2026) Millsap (2011).
Language, Culture, Disability, and Assessment Accessibility
Developmental measures are embedded in linguistic and cultural routines. The vocabulary needed to understand an item, familiar examples used in a scenario, and the relevance of independence or deference vary across environments. A cross-language translation that preserves literal wording may change pragmatic meaning or difficulty. Children growing up multilingual may respond differently depending on the language of administration and which language they use for particular domains. Consequently, age invariance established in one language does not guarantee invariance in another. Researchers need translation review, response-process investigation, item-level DIF analysis, and external validation within the populations for which claims will be made AERA, APA, & NCME (2014) Putnick & Bornstein (2016).
Disability and sensory accessibility create parallel design problems. A child with a motor disability may have difficulty completing a timed drawing task while possessing the intended conceptual competence; a respondent with limited hearing may miss information in a verbally administered memory task. Accommodations can remove construct-irrelevant barriers, yet they can also change score distributions or the strategy required to perform the task. The researcher must identify whether speed, manual action, or auditory perception is a legitimate component of the target construct. Alternate modalities deserve empirical linking and documentation, with adapted norms where necessary, rather than simple assurances of equivalence Messick (1995) AERA, APA, & NCME (2014).
Cultural fairness and developmental validity intersect because social expectations often change with age. For example, a questionnaire about autonomy or peer relationships can reflect different institutional opportunities across communities. Distinguishing genuine developmental differences from measurement bias requires information about schooling, family roles, and the context of observable behavior. Statistical invariance provides evidence about some comparative assumptions; substantive inquiry explains why a parameter behaves as observed. Instrument developers should invite expertise from the populations being studied, examine possible differential consequences of decisions, and make explicit the boundaries of any standardized comparison. Such procedures strengthen both scientific interpretation and responsible use of developmental measures Messick (1995) De Los Reyes et al. (2015).
Item Banks, Adaptive Testing, and Linked Forms
Item banking stores item content and calibrated response information so that researchers can select questions appropriate to a respondent while preserving a shared interpretative framework. Item response models make it possible to choose items that are informative at different levels of ability or symptom severity and to evaluate whether parameters remain stable between groups. A linked item bank can improve efficiency in longitudinal developmental studies in which the easiest tasks become uninformative and harder tasks become accessible only later. The central statistical requirement is sufficiently stable measurement relationships for the intended linking operation; the central scientific requirement is continued coverage of the construct as its manifestations evolve Domingue (2026) Millsap (2011).
Domingue's 2026 review discusses the promise of developmental item banks and surveys examples including the NIH Infant and Toddler Toolbox, LEVANTE, and the D-score system. These initiatives illustrate attempts to coordinate measurement across instruments and developmental periods while reducing unnecessary respondent burden. The review is a methodological synthesis and should not be read as demonstrating uniform measurement invariance or predictive accuracy for all items in each system. Each bank has its own development history, task domains, calibration populations, and operational purposes. Researchers adopting an item bank still need to document differential functioning, the logic of linking, uncertainty in individual estimates, and restrictions on comparisons between the scales and respondents involved Domingue (2026).
Adaptive testing can create a common latent-score framework while administering different tasks to different people. This advantage also changes the observation mechanism: selection rules, stopping criteria, exposure to earlier items, and platform effects may influence the final estimate. Longitudinal researchers should preserve versioned item parameters, document algorithm changes, and check whether the resulting scores remain comparable across software releases and cohorts. They should report uncertainty instead of interpreting small score differences as genuine development when those differences lie within measurement error. Where no secure common scale can be established, linked-but-partial reporting or construct-specific profiles may be scientifically preferable to a misleading single developmental index Domingue (2026) AERA, APA, & NCME (2014).
Reliability, Precision, and Individual Change
Reliability concerns consistency or precision under a specified measurement design. It does not establish that an item measures the intended psychological process or functions equally across ages. A scale can have high internal consistency in childhood and adulthood while differing in factor structure, thresholds, or relevance to external outcomes. For developmental inference, conditional precision is especially important: a measure may distinguish low ability levels well and perform poorly near the upper end, or show less information in a subgroup with unusual response patterns. Reports should distinguish population-level reliability estimates from individual-level measurement uncertainty and explain whether observed change exceeds plausible measurement variability Millsap (2011) Meredith (1993).
Repeated assessment introduces errors that may correlate over time. Familiarity with tasks, changes in testing environment, memory for prior responses, changes in informants, and seasonal opportunities can all affect observed stability. Test–retest correlations can therefore combine true developmental persistence, method stability, and systematic practice effects. Designs with alternative forms, counterbalanced administration, repeated short-term calibration, or multiple indicators can help estimate these components. Such controls should be chosen to fit the developmental question; an alternate form that is dramatically harder may reduce practice effects but create a new problem of linking. A credible reliability analysis is conditional on the score's intended use and the population represented Widaman et al. (2010) AERA, APA, & NCME (2014).
Individual change requires attention to uncertainty at two levels: variation in each person's estimated score and uncertainty in the model used to compare occasions. When the inference concerns whether one child has improved, confidence or credible intervals around latent change, sensitivity to measurement invariance, and meaningful external criteria matter more than the mere sign of a difference. When comparing developmental trends across groups, uncertainty in age-specific calibration and selection into the sample becomes important. Using a single common scoring algorithm across the life course may simplify reporting, but it cannot replace empirical checks of measurement structure and precision at each developmental interval Chen & Bauer (2024) Putnick & Bornstein (2016).
Developmental Psychopathology and Clinical Interpretation
Clinical developmental measurement adds the problem of distinguishing age-typical expression from clinically significant impairment. A behavior can be common in one developmental context and concerning in another; reports of intensity, frequency, duration, and impairment must be considered relative to opportunities and demands. In childhood, adult informants frequently provide the primary evidence, while adolescent and adult self-reports introduce different access to internal experience. Consequently, changes in symptom scores can arise from developmental shifts in expression, the observer system, and genuine differences in distress or impairment. Evaluating measurement invariance by specific symptom dimension is necessary before interpreting apparent stability or change in clinical severity Moriarity et al. (2025) De Los Reyes et al. (2015).
The harmonization study of Tyrell and colleagues addresses depression as a developmentally changing construct. Its demonstration that instruments with overlapping and distinct items can be linked under specific invariance assumptions is particularly relevant to longitudinal mental-health research. Meanwhile, Moriarity and colleagues found that evidence for invariance differed across symptom subscales rather than holding uniformly across the youth-to-adult transition. Together these studies caution against treating a diagnostic domain as a homogeneous measurement target. They also show that age-appropriate measures can support research when the measurement link is explicitly tested. Clinical decisions require additional evidence about diagnostic accuracy, impairment, and outcomes beyond factorial model fit Tyrell et al. (2019) Moriarity et al. (2025).
A scientifically defensible clinical interpretation follows a sequence: establish the construct and age-relevant manifestations; select instruments with suitable developmental validation; document who observes the behavior and under what conditions; assess linking across age and informants; then evaluate functional impairment and decision consequences. The present monograph supplies methodological synthesis rather than individualized diagnosis or intervention guidance. It does not assign clinical cutoffs or treatment recommendations. The broader validity framework requires that classifications be justified by evidence about their intended use and the likely consequences of misclassification, especially when developmental or cultural circumstances alter the meaning of a response Messick (1995) AERA, APA, & NCME (2014).
Response Shift, Self-Concept, and Lifespan Reporting
Self-report can change in meaning when people's standards, values, or conceptualizations of an experience evolve. Sprangers and Schwartz proposed response shift as a framework for understanding how appraisal and internal reference standards can change during health-related transitions. Although the original model was developed in quality-of-life research, its distinctions are relevant to developmental self-evaluation: a child's description of being capable, an adolescent's judgment of autonomy, and an older adult's judgment of well-being may invoke different standards and life circumstances. Longitudinal differences in reported levels can therefore include response recalibration as well as underlying psychological change Sprangers & Schwartz (1999).
Response shift encompasses possible recalibration of internal standards, changes in the priority attached to dimensions, and reconceptualization of the target domain. These forms are conceptually distinct from ordinary random measurement error and from changes in actual function, although empirical separation may be difficult. Schwartz and Sprangers discussed methodological strategies for identifying response shift in longitudinal research. Investigators can examine item-level changes, shifts in the structure of responses, external anchors, qualitative accounts of how standards changed, and the stability of links to observable outcomes. Such evidence is strongest when several independently informative methods agree, because a single parameter change may have multiple plausible explanations Sprangers & Schwartz (1999) Schwartz & Sprangers (1999).
Age-linked self-evaluation also intersects with social comparison. A teenager can rate competence against peers facing similar tasks; an adult may compare performance with professional or family responsibilities; an older person may assess autonomy relative to health limitations. Equal questionnaire responses need not represent equal opportunities or comparable reference groups. Researchers should characterize the subjective comparison frame and distinguish the respondent's lived appraisal from a standardized performance metric. Both can be legitimate constructs, but they answer different questions. Some developmental transformations are substantively important changes in the meaning of well-being or agency and should be modeled as such rather than forcibly erased by measurement constraints Messick (1995) Sprangers & Schwartz (1999).
Designing Developmentally Informative Studies
A strong developmental measurement study begins with an explicit research claim: whether a function is present, whether a latent mean changes, whether people retain relative standing, or whether different age-specific behaviors share a predictive pathway. The study design follows from that claim. Cross-sectional age comparisons offer broad coverage but mix developmental differences with cohort and selection. Longitudinal studies provide within-person observations but add attrition, repeated-test learning, and instrument changes. Cohort-sequential designs can help examine age and historical context when properly identified. Whatever design is selected, construct coverage, task demands, and reporting methods should be pre-specified or transparently documented before fitting growth models Widaman et al. (2010) AERA, APA, & NCME (2014).
Planning should include calibration overlap across developmental transitions. Participants in neighboring age ranges can complete parts of both instruments, permitting common-person linking; shared anchor items can support common-item linking. Randomized order reduces contamination from practice or fatigue, and response-process interviews test whether the anchor items retain a stable meaning. Cross-informant assessment can sample the same behavior in distinct contexts, while independent behavioral tasks provide external criteria. Researchers should simulate expected ceiling effects and score precision before data collection and plan sensitivity analyses for noninvariant items and informative dropout. This makes measurement equivalence a central design issue rather than a post hoc adjustment Domingue (2026) Petersen et al. (2026).
Analysis should match the measurement scale to the inferential target. Ordinal ratings may require threshold models rather than treating categories as equally spaced; counts and response times may require different likelihoods; performance outcomes can show speed–accuracy tradeoffs or floor and ceiling effects. Researchers should disclose item distributions, missingness, estimation methods, model identification, equality constraints, local misfit, alternate anchor sets, and uncertainty around linked means or growth parameters. External validity tests ought to include functionally meaningful outcomes and negative controls where feasible. When a single metric cannot be justified across all ages, partial links or age-specific conclusions preserve information without implying unsupported precision Millsap (2011) Chen (2007).
Causal Inference, Intervention, and Mechanism
Evidence of measurement comparability is necessary for many claims about developmental change, but measurement comparability itself does not identify causes. A longitudinal correlation may reflect stable family characteristics, reciprocal effects, schooling, biological maturation, or common changes in the observational environment. Random assignment of an intervention can strengthen causal inference about the intervention's effect, yet it does not ensure that the outcome scale has the same meaning across ages or treatment conditions. Researchers should test whether an intervention changes response strategy, social desirability, symptom labeling, or the task environment independently of the target function. The causal estimand and the measurement model should therefore be specified jointly Messick (1995) Widaman et al. (2010).
Mechanistic equivalence is a particularly demanding claim. Two age-differing behaviors may achieve similar outcomes through different learning histories, neural operations, or social supports. Establishing a shared mechanism requires stronger designs than finding equal latent scores or similar predictive correlations. Useful approaches include manipulating candidate task requirements, measuring mediating processes, testing invariance of response patterns after an intervention, and examining the boundaries of the purported function across contexts. When direct manipulation is inappropriate, longitudinal temporal ordering and convergent methods can constrain hypotheses without settling causation. Petersen's reanalysis of continuity arguments makes explicit why behavior, function, and mechanism must be separately specified Petersen (2024).
A developmental construct can also exhibit compensatory organization. Older adults may maintain task performance using additional time, external aids, or alternative strategies; children may achieve success through adult scaffolding. Such performance is real, but the functional interpretation depends on whether the target is independent capability, adaptive achievement, or effortful control under a given environment. Experimental removal or addition of support may help reveal which components govern performance, provided ecological relevance is retained. The best causal claim is limited to the manipulations and populations studied, and it states what would count as evidence of an alternate mechanism rather than assuming behavioral similarity settles the question Horn & McArdle (1992) Petersen (2024).
Digital Measurement, Ecological Observation, and Artificial Era Research
Contemporary developmental assessment increasingly includes digital tasks, experience sampling, passive contextual records, and adaptive interfaces. These modes can capture behaviors repeatedly in everyday situations and reduce dependence on retrospective questionnaires. They also introduce new sources of variation: device performance, software revisions, accessibility differences, platform literacy, privacy-related selection, and context-dependent use. A mobile questionnaire completed by a younger child with family assistance and an adult completing it independently may generate identical recorded responses through different response processes. Therefore, ecological and digital measurement demands explicit documentation of device, interface, assistance, timing, and observation context alongside traditional reliability and invariance analyses AERA, APA, & NCME (2014) Domingue (2026).
Longitudinal digital studies need version control over the measurement procedure itself. An algorithm that changes its adaptive item selection, a redesigned interface, or a modified passive-data classification rule may alter scores independently of development. Calibration bridge studies can administer old and new versions together and investigate whether model parameters can be linked. Distinctions between reproducibility of computation, stability of item functioning, and validity of developmental interpretation are essential. Even when millions of records are available, a biased selection process or shifting platform behavior can produce precise but misleading age trends. Scientific transparency requires traceable measurement specifications and independent validation against relevant behavioral criteria Millsap (2011) Messick (1995).
The Artificial Era introduces opportunities to compare developmental organization in human and artificial systems, but the meaning of development must be explicitly operationalized for each system. A software model's performance after retraining, a child's acquired capability, and an adult's compensatory adaptation arise through different histories and processes. Functional comparison can ask whether varied systems perform a specified operation under matched perturbations and resource constraints; psychological equivalence requires additional constructs and evidence. A shared benchmark score demonstrates a restricted behavioral outcome under specified conditions, whereas claims about artificial mental life or identity require separate research. The appropriate methodological contribution of developmental measurement is to clarify task equivalence, environment equivalence, and the exact limits of construct transfer Petersen (2024) Messick (1995).
Critical Appraisal of the Evidence Base
The literature provides robust methodological reasons to test invariance and substantive examples in which instrument linking supports restricted longitudinal conclusions. The evidence is uneven across constructs, age transitions, countries, and populations. Much of the foundational literature is theoretical or simulation-based, establishing the conditions under which comparisons are valid rather than documenting invariance in every empirical setting. Original longitudinal investigations offer more direct evidence but differ in sample composition and clinical or educational context. The Tyrell study addressed depressive symptoms with changing measures during the adolescent transition; Moriarity and colleagues found subscale-specific differences over a longer interval; Petersen and colleagues compared linking strategies in early childhood. Their combined contribution is a research program, not a single universal empirical calibration Tyrell et al. (2019) Moriarity et al. (2025) Petersen et al. (2026).
Common threats include underpowered subgroup tests, excessive dependence on global fit statistics, unstable anchors, exploratory model modifications, and treating partial invariance as permission for unrestricted comparisons. Informant effects and measurement drift can be especially difficult to separate when a new assessment method is introduced exactly as participants enter a new developmental period. Investigators should report both the strongest supported and weakest unsupported interpretations, with sensitivity analyses showing how results vary under different linking assumptions. Analytic choices such as parceling or allowing correlated residuals need a substantive rationale because they can alter both model fit and the visibility of item-level bias. Replication should aim at the inferential conclusion, including its known limitations Chen (2007) Putnick & Bornstein (2016).
External validity remains an important check. A measurement model may reproduce item covariances yet fail to predict independently observed functional behavior, while a functionally useful scale may require partially noninvariant indicators. The study of early childhood developmental scaling illustrates why criterion validity can inform decisions about age-differing items. Cross-informant research additionally reveals that a lack of agreement can reflect true contextual specificity. Evidence assessment must therefore use multiple standards: model adequacy, response processes, developmental content representation, criterion relations, transportability, and consequences. This combined approach is consistent with the historic construct-validation tradition and avoids elevating one statistical test into a complete theory of psychological development Petersen et al. (2026) De Los Reyes et al. (2015) Campbell & Fiske (1959).
Open Problems and Cumulative Research Priorities
A first open problem is constructing scales that remain meaningful across qualitative developmental reorganizations. When early capacities are expressed through gaze and action while later capacities appear through explicit verbal reasoning, common indicators may be scarce or conceptually narrow. Linking through short periods of overlap, common external criteria, and longitudinal response-process studies may provide partial solutions, but the identifiability and robustness of long-range connections remain unsettled. Future work should test whether a chain of local links preserves a stable interpretation when its endpoints are far apart. Reporting uncertainty accumulated across links is essential, since a numerical developmental score can appear far more precise than its evidential foundations warrant Domingue (2026) Petersen (2024).
A second problem is the integration of age, cohort, culture, language, informant, and testing context in a single measurement framework. Flexible item-response and moderated factor models can represent some of these dimensions, but complex models require large, informative samples and substantively justified constraints. Independent replications across languages and educational systems are needed to establish how portable item calibrations really are. Research should compare alternative model families on held-out groups and test whether they preserve the same conclusions about meaningful developmental outcomes. Rather than demand universal invariance, such programs can map where equivalence is supported and where a distinct age- or context-specific construct is required Chen & Bauer (2024) Putnick & Bornstein (2016).
A third priority is to advance functional validation beyond questionnaire covariation. Experimental manipulations of demands and supports, observational criteria in ecologically meaningful settings, mechanistic process measures, and pre-specified competing explanations can distinguish similar performances produced through different organization. Sensitive clinical and educational decisions require special attention to misclassification, fairness, and the consequences of using norms beyond their calibration populations. Open materials, item provenance, software versioning, precise reporting of missingness and attrition, and replication of both positive and negative invariance results would make developmental measurement more cumulative. The strongest future studies will treat the measurement relation as an object of explanation in its own right Messick (1995) AERA, APA, & NCME (2014).
Conclusion
Developmental Measurement and Functional Equivalence establishes that valid developmental inference depends on the relation among construct definition, observed behavior, response process, task environment, and the measurement model. Psychological capacities may change in degree, expression, organization, and social function. An unchanged item can acquire a different meaning across life stages, while changing instruments can sometimes be placed on a defensible shared metric. These possibilities are resolved through substantive research and explicit model tests. Construct validation, multi-method assessment, age-aware calibration, and external functional criteria jointly determine what a score can mean. The historical trajectory from construct theory through factorial invariance to developmental scaling has progressively clarified this central requirement Cronbach & Meehl (1955) Meredith (1993) Petersen et al. (2026).
The reviewed evidence identifies several practical conclusions. Longitudinal invariance testing is important before treating score differences as latent developmental growth; scalar or otherwise appropriate linking assumptions are required for many mean comparisons. Heterotypic continuity demands evidence about functions and mechanisms that extends beyond mere predictive correlation. Multi-informant disagreement can reveal the contexts in which behavior is expressed. Age-appropriate item design and linked scoring can improve criterion validity in studied populations, but no linking model automatically extends across all ages, constructs, or cultural groups. Measurement uncertainty and subgroup heterogeneity belong inside the interpretation of each reported developmental finding Widaman et al. (2010) De Los Reyes et al. (2015) Petersen et al. (2026).
The scientific horizon extends from longitudinal childhood assessment to lifespan research, clinical measurement, digital observation, and rigorous comparison of distinct cognitive systems. The governing principle is explicit inferential scope: define the psychological function, demonstrate that observations represent it, establish the comparisons the scale permits, and disclose where the interpretation ends. This principle supports cumulative developmental science because it makes competing claims testable and makes apparent continuity or change accountable to evidence. The work synthesized here remains a critical narrative monograph grounded in the cited scientific record; it offers methodological integration and research priorities rather than presenting an untested authorial framework as established empirical knowledge Messick (1995) Putnick & Bornstein (2016).
References
- Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/adolescent behavioral and emotional problems: Implications of cross-informant correlations for situational specificity. Psychological Bulletin, 101(2), 213–232. https://doi.org/10.1037/0033-2909.101.2.213
- American Educational Research Association, American Psychological Association, & National Council on Measurement in Education. (2014). Standards for educational and psychological testing. American Educational Research Association. https://ncme.org/resources/books/testing-standards/
- Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56(2), 81–105. https://doi.org/10.1037/h0046016
- Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 14(3), 464–504. https://doi.org/10.1080/10705510701301834
- Chen, S. M., & Bauer, D. J. (2024). Modeling construct change over time amidst potential changes in construct measurement: A longitudinal moderated factor analysis approach. Psychological Methods. Advance online publication. https://doi.org/10.1037/met0000685
- Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological Bulletin, 52(4), 281–302. https://doi.org/10.1037/h0040957
- De Los Reyes, A., Augenstein, T. M., Wang, M., Thomas, S. A., Drabick, D. A. G., Burgers, D. E., & Rabinowitz, J. (2015). The validity of the multi-informant approach to assessing child and adolescent mental health. Psychological Bulletin, 141(4), 858–900. https://doi.org/10.1037/a0038498
- Domingue, B. W. (2026). Psychometric rigor for the real world: Item banking in developmental research. Annual Review of Developmental Psychology, 8. Advance online publication, July 1, 2026. https://doi.org/10.1146/annurev-devpsych-111124-041520
- Horn, J. L., & McArdle, J. J. (1992). A practical and theoretical guide to measurement invariance in aging research. Experimental Aging Research, 18(3–4), 117–144. https://doi.org/10.1080/03610739208253916
- Meredith, W. (1993). Measurement invariance, factor analysis and factorial invariance. Psychometrika, 58(4), 525–543. https://doi.org/10.1007/BF02294825
- Messick, S. (1995). Validity of psychological assessment: Validation of inferences from persons’ responses and performances as scientific inquiry into score meaning. American Psychologist, 50(9), 741–749. https://doi.org/10.1037/0003-066X.50.9.741
- Millsap, R. E. (2011). Statistical approaches to measurement invariance. Routledge. https://www.routledge.com/Statistical-Approaches-to--Measurement-Invariance/Millsap/p/book/9780203821961
- Moriarity, D. P., Mac Giollabhui, N., Melo, D. C., & Hartman, C. (2025). Longitudinal measurement invariance of the ASEBA youth/adult self-reports across the transition from adolescence to adulthood. Assessment, 32(3), 356–367. https://doi.org/10.1177/10731911241245875
- Petersen, I. T. (2024). Reexamining developmental continuity and discontinuity in the 21st century: Better aligning behaviors, functions, and mechanisms. Developmental Psychology, 60(11), 1992–2007. https://doi.org/10.1037/dev0001657
- Petersen, I. T., Demko, Z., Lee, W.-C., & Oleson, J. J. (2026). Studying development of psychopathology using changing measures to account for heterotypic continuity. JAACAP Open, 4(1), 111–123. https://doi.org/10.1016/j.jaacop.2025.10.008
- Putnick, D. L., & Bornstein, M. H. (2016). Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Developmental Review, 41, 71–90. https://doi.org/10.1016/j.dr.2016.06.004
- Schwartz, C. E., & Sprangers, M. A. G. (1999). Methodological approaches for assessing response shift in longitudinal health-related quality-of-life research. Social Science & Medicine, 48(11), 1531–1548. https://doi.org/10.1016/S0277-9536(99)00047-7
- Sprangers, M. A. G., & Schwartz, C. E. (1999). Integrating response shift into health-related quality of life research: A theoretical model. Social Science & Medicine, 48(11), 1507–1515. https://doi.org/10.1016/S0277-9536(99)00045-3
- Tyrell, F. A., Yates, T. M., Widaman, K. F., Reynolds, C. A., & Fabricius, W. V. (2019). Data harmonization: Establishing measurement invariance across different assessments of the same construct across adolescence. Journal of Clinical Child & Adolescent Psychology, 48(4), 555–567. https://doi.org/10.1080/15374416.2019.1622124
- Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3(1), 4–70. https://doi.org/10.1177/109442810031002
- Widaman, K. F., Ferrer, E., & Conger, R. D. (2010). Factorial invariance within longitudinal structural equation models: Measuring the same construct across time. Child Development Perspectives, 4(1), 10–18. https://doi.org/10.1111/j.1750-8606.2009.00110.x