How to Measure Anxiety: Scales, Tests, and Devices

Anxiety can be measured through questionnaires you fill out yourself, structured interviews conducted by a clinician, physiological signals picked up by sensors, and even patterns in how you use your smartphone. No single tool captures the full picture, which is why researchers and clinicians typically combine methods depending on the setting and the question being asked. The landscape ranges from paper-and-pencil scales developed decades ago to machine-learning algorithms analyzing your heart rhythm in real time, and each approach has trade-offs worth understanding.

Self-Report Questionnaires

The most common way anxiety gets measured in both clinical practice and research is through short questionnaires you complete yourself. The GAD-7 (Generalized Anxiety Disorder 7-item scale) is probably the one you are most likely to encounter at a doctor’s office. It asks you to rate how often you have been bothered by symptoms like feeling nervous, being unable to stop worrying, or having trouble relaxing over the past two weeks, on a scale from “not at all” to “nearly every day.” It takes under two minutes, and scores range from 0 to 21, with cutoffs at 5, 10, and 15 marking mild, moderate, and severe anxiety.

The State-Trait Anxiety Inventory (STAI) takes a different approach by separating how you feel right now (state anxiety) from how anxious you tend to feel in general (trait anxiety). The full version has 40 items, though abbreviated 8-item versions for each subscale have been validated with strong internal consistency and balanced factorial structures.¹ The STAI, Beck Anxiety Inventory (BAI), and Hospital Anxiety and Depression Scale (HADS-A) are among the most widely used instruments in clinical research.²

The BAI leans more heavily on physical symptoms of anxiety, asking about numbness, dizziness, heart pounding, and sweating. When researchers compared the BAI and the clinician-rated Hamilton scale in a large sample of people with diagnosed anxiety disorders, they found the BAI captured four components: neurophysiological symptoms, subjective feelings, panic-related symptoms, and autonomic symptoms. The Hamilton scale, by contrast, loaded onto just two: psychic anxiety and somatic anxiety.³ This matters because the tool you choose shapes what kind of anxiety you detect. If someone’s anxiety is mostly cognitive worry without many physical symptoms, the BAI might undercount it.

Clinician-Administered Scales

When a trained professional rates your anxiety rather than you rating it yourself, the Hamilton Anxiety Rating Scale (HARS) is the classic choice. A clinician conducts a structured or semi-structured interview and scores 14 items covering both psychological symptoms (anxious mood, tension, fears) and physical ones (cardiovascular, respiratory, gastrointestinal). The HARS has been validated across different populations, showing good construct validity against independent self-report measures of anxiety.⁴ It has also been validated alongside the BAI and HADS in people with Parkinson’s disease, where separating anxiety symptoms from movement-disorder symptoms presents an extra challenge.⁵

The advantage of clinician-rated scales is that a trained interviewer can probe vague answers, notice nonverbal cues, and judge whether a symptom is better explained by a medical condition. The downside is cost and time. An HARS interview takes longer than handing someone a questionnaire, and it requires a trained rater. That is why self-report tools dominate primary care and large-scale research, while clinician-rated instruments tend to appear in specialized settings and clinical trials.

Measuring Anxiety in Children and Teens

Kids and teenagers pose a unique challenge because they often lack the vocabulary or self-awareness to report their internal states accurately. The Screen for Child Anxiety Related Emotional Disorders (SCARED) addresses this by offering both a child-report and a parent-report version. But the two versions frequently disagree. In one study, parent-child agreement on SCARED scores was weak across the board, with parents consistently under-reporting symptoms compared to their children. The gap was significantly larger in families where the child had a diagnosed anxiety disorder.⁶

Interestingly, this disagreement improves as children age. A longitudinal study following families from when children were 9 to 15 found that parent-child agreement on anxiety symptoms was weak at age 9 but moderate by ages 12 and 15. Mother-father agreement was consistently stronger than either parent’s agreement with the child at every age.⁷ The practical takeaway is that relying on a single informant, especially a parent of a younger child, risks missing real anxiety. Best practice involves collecting both perspectives and using a clinician interview as a tiebreaker.

What Counts as a Meaningful Change in Scores

Knowing your anxiety score dropped by a few points after treatment sounds encouraging, but is that drop large enough to reflect a real improvement you would actually notice? Researchers answer this with a concept called the minimally clinically important difference (MCID), essentially the smallest score change that corresponds to a perceptible shift in how you feel. For the GAD-7, the MCID is estimated at about 4 points. In a study tracking patients over a year of treatment, those who improved showed significantly lower GAD-7 scores at both 12 and 48 weeks, while those who worsened showed higher scores, and those who stayed the same showed no meaningful shift.⁸ A comparison across several common anxiety scales found MCIDs of 3 for the GAD-7, 4 for PROMIS anxiety scales, 6 for the PTSD checklist, and 0.3 for the SCL anxiety scale, reflecting the different scoring ranges of each instrument.⁹

For children, the Pediatric Anxiety Rating Scale (PARS) has its own benchmarks. A reduction of 4 or more points (or at least 20 percent) qualifies as a meaningful change. A larger drop of 8 or more points (or 43 percent) defines treatment response, and post-treatment scores of 0 to 5 indicate full remission.¹⁰ These numbers give clinicians and families a concrete way to judge whether treatment is working rather than relying on a vague sense that things are “a little better.”

Telling Anxiety Apart from Depression

Anxiety and depression overlap enough in their symptoms that most standard scales struggle to cleanly separate them. If you score high on a general anxiety questionnaire, some of that score might be driven by low mood, fatigue, or irritability that depression and anxiety share. Researchers have proposed extended models that pull apart the overlapping symptoms. One approach identified five separable dimensions: depressed mood, lack of positive affect, somatic arousal, phobic fear, and hostility. The resulting scales could accurately differentiate between patients with mood disorders and those with anxiety disorders, and even distinguish between different anxiety disorders.¹¹

A more recent effort, the State-Trait Anxiety-Depression Inventory, was built from the ground up to measure anxiety and depression as distinct but related constructs. Tested in a representative German sample of over 3,000 people, it showed strong construct validity for separating anxiety from depression at the item level.¹² This distinction matters beyond academic tidiness. If your assessment tool conflates anxiety with depression, treatment decisions based on that score might lead you toward an antidepressant when the primary problem is anxiety, or vice versa.

Physiological Signals and Stress Hormones

Your body produces measurable signals when you are anxious, and researchers have been working to turn those signals into objective anxiety measures. Heart rate variability (HRV), the variation in time between consecutive heartbeats, reflects the balance between the sympathetic (“fight or flight”) and parasympathetic (“rest and digest”) branches of your nervous system. Lower HRV is associated with anxiety, making it a promising biomarker under active investigation.¹³ Electrodermal activity (EDA), which measures tiny changes in skin conductance caused by sweat gland activity, is another commonly studied signal.¹⁴

Cortisol, the stress hormone produced by the adrenal glands, offers another window. Researchers often measure the cortisol awakening response (CAR), the spike in cortisol that normally occurs in the first 30 to 60 minutes after waking up. In people with anxiety disorders, two atypical patterns have been identified: a flattened response (little or no rise after waking) and an exaggerated response (a rise above 75 percent).¹⁵ The relationship between cortisol and anxiety is not as straightforward as “more anxiety equals more cortisol,” though. Older adults with long-lasting anxiety disorders actually showed a lower cortisol awakening response than those without anxiety, consistent with the idea that chronic anxiety can burn out the stress-hormone system over time.¹⁶ Prospective data add nuance: a higher CAR predicted increased first onsets of anxiety disorders, with a particularly strong link to social anxiety disorder.¹⁷

So cortisol can be elevated before an anxiety disorder develops, then depleted in people who have had one for years. This bidirectional pattern means a single cortisol reading is not a reliable diagnostic on its own.

Wearable Devices

Consumer wearables like smartwatches and fitness bands already track HRV, heart rate, and sometimes skin conductance, which has sparked interest in using them for anxiety detection. A systematic review of wearable anxiety-detection studies found that combining multiple signals outperformed relying on any one signal alone, with a pooled accuracy of about 82 percent for multi-signal approaches compared to roughly 77 percent for single signals. Among individual signals, ECG (electrocardiography) performed best, with a pooled accuracy around 80 percent. Electrodermal activity came next at about 79 percent, while PPG (photoplethysmography, the green-light sensor on most smartwatches) lagged behind at around 69 percent.¹⁸

Those numbers sound promising, but they come from controlled study conditions where participants often performed specific tasks designed to induce stress. In daily life, your heart rate rises when you climb stairs or drink coffee, not just when you are anxious. The gap between lab accuracy and real-world usefulness is the central challenge. A platform designed to combine wearable physiological data with self-report for mental health monitoring found a substantial rate of false positives in a healthy sample, though users responded positively to the concept.¹⁹ The technology works well enough to flag patterns worth investigating, but it is not yet reliable enough to replace a clinical assessment.

Regulatory hurdles add another layer. Wearable software intended for clinical anxiety assessment would need to navigate FDA classification as a Software Medical Device, demonstrating not just that it accurately measures physiology, but that it correlates with actual anxiety states and improves patient outcomes.²⁰ No consumer wearable has cleared that bar for anxiety specifically.

Tracking Anxiety in Real Time

Traditional questionnaires ask you to think back over the past week or two and summarize how you felt. Memory is unreliable, and people tend to overweight recent or extreme moments. Ecological momentary assessment (EMA) avoids this by pinging you multiple times a day on your phone and asking how you feel right now. This approach captures the natural ups and downs of anxiety in real life, revealing patterns that retrospective questionnaires miss entirely: which situations trigger spikes, how quickly symptoms resolve, and how anxiety interacts with sleep, activity, and social contact.²¹ EMA has also been adapted for youth, where it shows promise for clinical trials studying treatment effects in real time.²²

Digital phenotyping goes further by passively collecting data from your phone’s sensors, things like GPS patterns, screen-on time, typing speed, and call frequency, without requiring you to answer any questions at all. Researchers have shown that smartphone sensor data can predict the severity of social anxiety symptoms with a strong correlation, and customized deep-learning models can forecast hour-by-hour changes in a person’s anxiety levels. Longer-term prediction is also possible: wearable motion-sensing data significantly predicted which people would experience symptom deterioration over a period of many years.²³ The accuracy is impressive in research settings, though the privacy implications of continuous behavioral monitoring remain a serious concern that the field has not fully resolved.

What Brain Imaging Reveals

Functional and structural brain imaging is not used to diagnose anxiety in clinical practice, but it has taught researchers a great deal about what anxiety looks like in the brain. The core finding across studies is an imbalance between the amygdala, a region that generates fear responses, and the prefrontal cortex, which normally keeps those responses in check. In people with pathological anxiety, the amygdala tends to be overactive while the prefrontal cortex is underactive.²⁴

More refined imaging work in generalized anxiety disorder (GAD) shows that the picture is more specific than a blanket “too much amygdala, too little prefrontal.” The amygdala shows reduced connectivity with the ventral prefrontal cortex (a region involved in dampening fear) but increased connectivity with the dorsal prefrontal cortex (which amplifies aversive responses).²⁵ A meta-analysis of structural and functional MRI studies confirmed that key nodes in the brain’s resting-state networks, including the hippocampus, anterior cingulate cortex, and amygdala, show reduced volume in people with GAD, and the connectivity between the prefrontal cortex and amygdala is weakened.²⁶

EEG (electroencephalography) offers a cheaper, more portable alternative. Frontal alpha asymmetry, a measure of the relative electrical activity in the left versus right frontal lobes, has been explored as a potential marker. Researchers have tested latent factorial approaches to measuring this asymmetry, extracting distinct frontal and parietal factors from resting-state EEG data.²⁷ The appeal is that EEG can be done with relatively inexpensive equipment and repeated frequently, but the findings are still too variable across individuals to serve as a standalone diagnostic.

Inflammatory Blood Markers

A growing body of evidence links anxiety to low-grade inflammation throughout the body. A meta-analysis found significantly raised levels of C-reactive protein (CRP), interferon-γ, and tumor necrosis factor-α in people with generalized anxiety disorder compared to healthy controls.²⁸ A large study using UK Biobank data from up to about 145,000 participants found that CRP concentrations were associated with anxiety symptom scores in a dose-response fashion, meaning that higher inflammation tracked with worse anxiety in a graded way, not just above a threshold. Genetic analyses in the same study suggested the relationship is at least partly causal: genetically predicted higher levels of the inflammatory molecule IL-6 were associated with both depression and anxiety.²⁹

The relationship runs in both directions. Longitudinal data show that current anxiety disorders, particularly agoraphobia, predicted a steeper rise in CRP levels over the follow-up period, suggesting that chronic anxiety itself may drive inflammation upward over time.³⁰ None of this means a blood test for CRP can diagnose anxiety. CRP rises with infections, obesity, and autoimmune conditions too. But inflammatory markers may eventually contribute to a broader panel that helps clinicians assess risk or monitor treatment response.

Cognitive Tasks That Measure Attentional Bias

Anxious people tend to notice threats faster than non-anxious people and have a harder time looking away from them. Researchers measure this attentional bias with tasks like the dot-probe, where pairs of images (one threatening, one neutral) flash briefly on a screen, and you press a button as fast as you can when a dot appears in one location. The dot-probe task is often considered the gold standard for studying attentional bias to threat in anxiety research.³¹ Newer versions of the task combine reaction time data with eye-tracking and EEG to measure both where your eyes go and how your brain responds to the threatening image.³²

The catch is that the dot-probe task has struggled with poor reliability. Someone who shows a strong bias toward threat on one testing day might not show it a week later. This makes it a useful group-level research tool but a questionable individual-level diagnostic. The hope is that combining behavioral, eye-tracking, and neural measures in a single session will produce more stable results, but this remains an active area of work.

Cultural Dimensions of Anxiety Assessment

Most widely used anxiety scales were developed and validated in Western, English-speaking populations. When translated into other languages and cultures, the same questions can mean different things. Some cultures express distress primarily through physical symptoms rather than psychological ones, or use idioms of distress that do not map neatly onto Western categories. A systematic review of studies that incorporated culturally specific idioms of distress into mental health assessments found that measures integrating these idioms generally showed good psychometric properties, suggesting the approach is feasible.³³ If you are being assessed using a translated scale and feel the questions miss how you actually experience anxiety, that disconnect is a recognized limitation of the instrument, not a sign that your experience is not real.

How Animal Models Inform the Science

Much of what we know about the neurobiology of anxiety comes from rodent models, and those models rely on their own behavioral “anxiety tests.” The elevated plus maze, for instance, places a rat or mouse at the junction of four arms, two enclosed and two open. Because rodents naturally avoid open, elevated spaces, the amount of time they spend in the open arms reflects how anxious they are. The test takes just five minutes per animal and has been validated for evaluating anti-anxiety drugs and identifying brain regions involved in anxiety-related behavior.³⁴ These preclinical tests are how researchers screen candidate anxiety medications before human trials, and they are why the drugs you might eventually be prescribed were originally considered worth testing in people at all.