Does the DASS-21 Really Measure Three Distinct Constructs? Evidence From Pakistani Students

The DASS-21 is designed to measure three related negative emotional states—depression, anxiety, and stress—but evidence from Pakistani university students suggests that the distinction is not as simple as three neatly separated psychological constructs. In a 2026 study of 602 students in Lahore, a three-factor confirmatory model showed adequate overall fit, yet the latent correlations between the DASS-21 factors were extremely high. Depression and stress correlated at 0.939, while anxiety and stress correlated at 0.949. The practical message is therefore nuanced: the DASS-21 can organize symptoms into depression, anxiety, and stress domains, but these domains may substantially overlap in this student population.

By Taimoor Asghar

What Is the DASS-21 Supposed to Measure?

The Depression Anxiety Stress Scales-21, usually abbreviated DASS-21, is a 21-item self-report questionnaire derived from the longer 42-item Depression Anxiety Stress Scales. Its items are divided equally across three seven-item subscales: depression, anxiety, and stress.

The depression items are intended to capture experiences such as low mood, reduced positive affect, hopelessness, loss of interest, and diminished motivation. Anxiety items emphasize manifestations such as physiological arousal, fear, and situational anxiety. Stress items focus more on persistent tension, difficulty relaxing, irritability, agitation, and feeling easily overwhelmed.

Conceptually, this is attractive. Instead of reducing psychological distress to a single score, the DASS-21 attempts to describe different forms of negative emotional experience. A student who primarily reports low mood and lack of interest could theoretically show a different profile from one whose difficulties are dominated by physiological anxiety or persistent tension.

But designing three subscales does not automatically prove that people actually experience or report these symptoms as three clearly separable dimensions. That question requires psychometric testing.

Why Factor Structure Matters

When researchers say that an instrument has a three-factor structure, they are making a statement about the relationships among its items. In the DASS-21, the expected model proposes that depression items tend to cluster together, anxiety items cluster together, and stress items cluster together.

Confirmatory factor analysis, or CFA, can test how well such a proposed structure corresponds to observed response patterns. Researchers specify the expected relationships in advance and then evaluate how closely the statistical model reproduces the data.

This matters because questionnaire scores are interpreted as representations of underlying constructs. If the depression, anxiety, and stress items cannot be meaningfully distinguished statistically, treating the three subscale scores as completely independent psychological quantities becomes harder to justify.

At the same time, complete independence is not expected. Depression, anxiety, and stress commonly co-occur, and the DASS framework was developed around related forms of negative emotional experience. The central psychometric question is therefore not whether the factors correlate, but how strongly they correlate and whether enough unique information remains to justify interpreting them separately.

Evidence From 602 Pakistani University Students

A recently published cross-sectional study examined the psychometric characteristics of the PHQ-9 and DASS-21 among 602 undergraduate students at universities in Lahore, Pakistan. The sample included 424 medical students and 178 non-medical students. The researchers used several analytical approaches, including confirmatory factor analysis, reliability analysis, regression modelling, item response theory, symptom network analysis, and measurement invariance testing.

The full study, Psychometric properties and symptom profiles of the PHQ-9 and DASS-21 among medical and non-medical university students, was published in BMC Psychology in August 2026.

For the DASS-21, CFA produced a Comparative Fit Index of 0.954, Tucker-Lewis Index of 0.948, and Root Mean Square Error of Approximation of 0.052. Taken together, these indices indicated that the proposed factor model provided an adequate representation of the observed responses.

Looking only at those model-fit statistics might lead to a straightforward conclusion: the expected DASS-21 structure worked well. However, another part of the analysis makes the interpretation considerably more interesting.

The Critical Finding: Very High Correlations Between Factors

The estimated latent correlation between DASS-21 depression and stress was 0.939. The correlation between anxiety and stress was even higher at 0.949.

These values are striking because a correlation of 1.0 would indicate statistical identity at the latent-factor level. Values approaching that level suggest that the constructs share a very large amount of variance.

This does not mean that depression, anxiety, and stress are literally the same psychological experience. Nor does it prove that the three DASS-21 subscales should be discarded. It does mean that their empirical distinguishability deserves careful attention in this population.

The distinction between model fit and discriminant validity is crucial here. A three-factor model can fit the data adequately while the factors within that model remain extremely strongly correlated. Model fit tells us whether the proposed pattern reproduces the observed covariance structure reasonably well. It does not, by itself, establish that every latent factor represents a sharply differentiated construct.

Does Good CFA Fit Prove There Are Three Distinct Constructs?

No. Good CFA fit supports the plausibility of the specified measurement structure, but it is not sufficient evidence that the factors are strongly distinct from one another.

Imagine three departments within the same organization. Employees may be assigned correctly to separate departments, meaning the organizational model fits. Yet if the departments perform almost identical work and operate almost entirely together, their practical distinctiveness could still be questioned.

A similar issue can arise in psychological measurement. Items may cluster according to theoretically expected depression, anxiety, and stress domains, while the underlying factors themselves remain tightly interconnected.

Researchers therefore examine more than fit indices when evaluating dimensionality. Depending on the research question, relevant evidence can include:

• latent-factor correlations;
• factor loadings;
• reliability coefficients;
• discriminant and convergent validity;
• measurement invariance;
• comparisons between one-factor, correlated-factor, hierarchical, and bifactor models;
• relationships with external clinical or behavioral variables; and
• whether separate subscales predict meaningfully different outcomes.

The Pakistani study provides evidence supporting an adequately fitting DASS-21 factor model, while simultaneously highlighting unusually strong relationships among its latent dimensions. Those findings should be considered together rather than selecting whichever result supports a preferred interpretation.

DASS-21 Factor Structure: Distinct but Closely Related May Be the Better Description

One way to interpret the results is to move away from a binary question—either three completely separate constructs or one completely unified construct.

Psychological symptoms frequently have hierarchical structures. At one level, people may experience a broad tendency toward psychological or emotional distress. At another level, that distress may manifest through somewhat more specific patterns associated with depressive affect, anxious arousal, or tension and stress.

This possibility has appeared in previous DASS-21 research. For example, work by Henry and Crawford found evidence that DASS-21 responses could contain both specific depression, anxiety, and stress dimensions and a broader general psychological-distress component. Large cross-cultural investigations have likewise examined whether the intended dimensions remain interpretable across different countries and populations.

The Pakistani findings are compatible with the broader observation that DASS-21 domains can be statistically structured as depression, anxiety, and stress while sharing considerable common variance. However, the 2026 study should not be interpreted as demonstrating that a particular alternative model—such as a bifactor or single general-distress model—is superior unless those models are directly compared within the relevant analysis.

Why Might Depression, Anxiety, and Stress Overlap So Strongly?

There are several plausible explanations, and a cross-sectional psychometric study cannot establish which is responsible.

1. The experiences themselves genuinely overlap

Depressive symptoms, anxiety symptoms, and stress responses are not isolated biological or psychological systems. A student facing persistent academic or financial pressure may simultaneously experience tension, worry, impaired concentration, poor sleep, reduced motivation, and low mood.

Strong correlations can therefore reflect genuine co-occurrence rather than a defective questionnaire.

2. A broader negative-affect or distress dimension may influence responses

Many psychiatric symptoms share a general component of emotional distress. Someone experiencing high overall distress may endorse items across multiple DASS-21 domains, creating strong relationships between the subscales even when each domain retains some specific content.

3. Student populations may experience blended symptom profiles

University students face overlapping academic, social, financial, sleep-related, and transitional pressures. In a non-clinical student sample, emotional difficulties may not organize themselves into the same sharply differentiated patterns that diagnostic labels sometimes imply.

4. Cultural and linguistic context can affect symptom interpretation

Psychological constructs are measured through language and self-report. The meaning respondents assign to concepts such as agitation, nervousness, lack of enthusiasm, or difficulty relaxing may vary across populations. That is one reason instruments validated elsewhere should still be psychometrically examined when applied in new cultural contexts.

None of these possibilities should be assumed to explain the Pakistani results without further study. They illustrate why very high latent correlations can have several plausible sources.

Reliability Was Good, but Reliability Is Not the Same as Distinctiveness

The Pakistani study reported Cronbach’s alpha coefficients between 0.82 and 0.88 across the assessed scales, indicating good internal consistency in this sample.

Reliability answers a different question from factor distinctiveness. A scale can contain items that consistently move together and therefore have high internal consistency while still overlapping strongly with another reliable scale.

For example, if both an anxiety scale and a stress scale are individually reliable but correlate extremely strongly with each other, their reliability does not establish that they measure entirely separate phenomena.

This distinction is relevant whenever questionnaires are evaluated. Statements such as “the scale was reliable” should not automatically be translated into “the scale definitively measures a unique construct.” Reliability is necessary for useful measurement, but validity requires a wider body of evidence.

Measurement Invariance Adds Another Important Piece

The study also reported support for measurement invariance across academic discipline. This is particularly relevant because the sample contained both medical and non-medical students.

Measurement invariance asks whether an instrument functions comparably across groups. Without sufficient invariance, an observed difference between groups could partly reflect the questionnaire behaving differently rather than a genuine difference in the underlying construct.

Support for invariance therefore strengthens the rationale for comparing DASS-21 scores between medical and non-medical students in this dataset. It does not, however, resolve the separate question of whether depression, anxiety, and stress are sufficiently distinct from one another. Group comparability and discriminant validity are different psychometric issues.

What Did the Study Find About Medical and Non-Medical Students?

The broader study was not limited to factor analysis. After adjustment for the variables included in the regression models, non-medical students had higher DASS-21 depression and anxiety scores than medical students, whereas the difference in stress scores did not reach statistical significance. Female gender was associated with higher scores across the examined symptom domains.

These group-level findings show why preserving domain-specific scores can still have analytical value. Depression, anxiety, and stress did not necessarily behave identically in every comparison, despite their strong latent correlations.

At the same time, the authors reported low model R-squared values for the regression analyses, meaning academic discipline and the included covariates explained only a small proportion of the overall variation in scores. The study therefore does not support simplistic conclusions that a student’s field of study determines their mental-health profile.

What Should Researchers Do With DASS-21 Subscale Scores?

The evidence does not support treating the DASS-21 as useless, nor does it justify assuming that its three scores represent perfectly isolated psychological dimensions. A more defensible approach is to interpret the subscales as related indicators of negative emotional states whose degree of separability should be evaluated within the population being studied.

Researchers using the DASS-21 in Pakistani university populations can improve interpretation by reporting the psychometric evidence observed in their own sample rather than relying only on validation studies from other countries.

Useful practices include reporting subscale reliability, factor correlations, CFA results, and measurement invariance when group comparisons are central to the research question. Where dimensionality is a major scientific objective, comparing competing models may provide additional information beyond testing only the conventional correlated three-factor structure.

It may also be useful to examine whether depression, anxiety, and stress scores show different relationships with external variables. If the three domains predict different outcomes or correlate differently with independent measures, that can provide evidence of practical distinctiveness even when their mutual correlations are high.

What Should Clinicians and Students Take From These Findings?

The most important clinical implication is that DASS-21 numbers should not be treated as diagnoses. The instrument measures self-reported symptom severity across related emotional domains. A high depression, anxiety, or stress score cannot by itself establish a psychiatric disorder, determine its cause, or replace a clinical assessment.

The strong overlap observed in the Pakistani student sample reinforces the importance of looking at the whole person rather than interpreting a subscale score in isolation. Someone reporting high anxiety may also have considerable stress and depressive symptoms, and the relationships among these experiences may be clinically more informative than a single label.

For students completing such questionnaires, scores are better understood as signals about current symptom burden than as definitive statements about identity or diagnosis. Persistent distress, significant impairment in daily functioning, or concerns about safety warrant assessment by an appropriately qualified healthcare or mental-health professional.

What This Study Does—and Does Not—Establish

The study contributes valuable evidence because psychometric properties should not simply be assumed to transfer perfectly across countries and populations. Its sample of Pakistani medical and non-medical university students provides locally relevant data that have historically been less represented in the international psychometric literature.

The results support several conclusions. The DASS-21 showed good internal consistency. Its CFA demonstrated adequate fit. Its measurement properties were sufficiently invariant across academic discipline for the comparisons performed in the study. At the same time, depression, anxiety, and stress showed extremely strong latent relationships, particularly the depression-stress and anxiety-stress pairs.

What the study cannot establish is equally important. Because the design was cross-sectional, it cannot determine how the structure of emotional symptoms changes over time. The findings should not automatically be generalized to all Pakistani students, all age groups, clinical populations, or the Pakistani population as a whole. Nor do high correlations prove that the DASS-21 should be replaced by a single total score.

Determining whether a general-distress model, bifactor structure, alternative shortened scale, or traditional correlated three-factor model provides the most useful representation would require appropriately designed model comparisons and replication in independent samples.

So, Does the DASS-21 Really Measure Three Distinct Constructs?

The most evidence-consistent answer is: it appears to measure three recognizable but very strongly overlapping domains in this Pakistani university-student sample.

The adequate CFA fit supports the intended structural organization of depression, anxiety, and stress items. However, latent correlations of 0.939 between depression and stress and 0.949 between anxiety and stress caution against describing these dimensions as cleanly independent constructs.

This is not necessarily a contradiction. Psychological measurement often involves constructs that are distinguishable in theory and item content while sharing a substantial common core. The DASS-21 may therefore be most informative when its three subscales are interpreted together, with attention to both their domain-specific content and their shared representation of broader emotional distress.

For researchers, the finding is a reminder that reporting good model-fit indices is only the beginning of construct validation. For clinicians and students, it is a reminder that questionnaire categories simplify emotional experiences that are often deeply interconnected.

Medical Disclaimer

This article is for educational and research-information purposes only. The DASS-21 is a symptom-assessment instrument and should not be used on its own to diagnose depression, anxiety disorders, or other mental-health conditions. Individual symptoms and questionnaire results should be interpreted in the appropriate clinical context by a qualified healthcare or mental-health professional. Anyone experiencing severe psychological distress, substantial impairment, or concerns about personal safety should seek appropriate professional assessment promptly.