The Beck Anxiety Inventory (BAI) is a 21-item self-report measure of anxiety symptom severity over the past week. Each item is rated from 0 (not at all) to 3 (severely), giving a total of 0–63. Its item content leans deliberately towards the physiological arousal side of anxiety - trembling, palpitations, dizziness, fear of losing control - rather than the cognitive symptoms of worry.
That somatic emphasis was a design choice: the BAI was built to measure anxiety with minimal contamination from depression, with which anxiety scales of the 1980s overlapped heavily. The result is an instrument that tracks panic-type arousal well, but captures the persistent worry of generalised anxiety less directly than instruments such as the GAD-7.
02 - Origin & purpose
Where it comes from.
The BAI was developed by Aaron T. Beck and colleagues at the Center for Cognitive Therapy, University of Pennsylvania, and published in 1988 (Beck, Epstein, Brown & Steer, Journal of Consulting and Clinical Psychology) 1. The stated aim was a brief anxiety measure that discriminated anxiety from depression, since existing scales correlated so highly with depression measures that they arguably measured general distress.
Items were derived from a pool of earlier Beck instruments and refined psychometrically in psychiatric outpatient samples. It became one of the most widely used anxiety severity measures in clinical practice and research, alongside its sibling the BDI-II. Unlike the free PHQ/GAD family, it remained a commercial instrument, now distributed by Pearson.
03 - Scoring & cutoffs
How scoring works.
Each of the 21 items is scored 0–3 and summed to a total of 0–63, reflecting symptom severity over the past week. The manual's severity bands (Beck & Steer, 1993) are: 0–7 minimal, 8–15 mild, 16–25 moderate, 26–63 severe 2. These bands were set from clinical experience rather than tested against a diagnostic gold standard, so they are best read as broad severity guides rather than diagnostic thresholds.
A total of 16 or above is commonly treated as the point at which clinically significant anxiety is likely. Because roughly two-thirds of the items describe bodily sensations, scores can be inflated in people with cardiac, respiratory, vestibular or other medical conditions - interpret elevated scores in medically unwell patients with that in mind.
Score
Severity
Interpretation
0–7
Minimal anxiety
Symptoms at or near the level seen in non-clinical samples.
8–15
Mild anxiety
Symptoms present but limited in intensity.
16–25
Moderate anxiety
Clinically significant anxiety is likely; assess and consider treatment.
26–63
Severe anxiety
High symptom burden; assess panic, avoidance and treatment intensity.
Severity bands are clinical conventions from the BAI manual, not diagnostically validated thresholds. Somatic items mean medical illness can raise scores independently of anxiety.
04 - Validation evidence
How well it performs.
The original 1988 validation in 160 psychiatric outpatients showed high internal consistency and adequate one-week stability, and - as intended - only moderate correlation with depression measures 1. Subsequent studies replicated the strong internal consistency across clinical, student and older-adult samples 34. Formal diagnostic accuracy data are thinner than for screening instruments such as the GAD-7, because the BAI was built as a severity measure, not a case-finder; one Korean clinical validation study reported good discrimination of panic disorder at a much lower threshold than the manual's severity bands 5.
α = 0.92
Internal consistency
In the original psychiatric sample (Beck et al., 1988); up to 0.94 in anxiety-disorder outpatients (Fydrich et al., 1992).
r = 0.75
Test–retest reliability
At one week (Beck et al., 1988), falling to r = 0.62 at seven weeks (Creamer et al., 1995).
Moderate
Discriminant validity
Designed to separate anxiety from depression; correlates only moderately with depression measures, its key advantage over older anxiety scales.
0.89 / 0.97
Case-finding (sens/spec)
In one clinical ROC study a cutoff of 8.5 identified panic disorder with sensitivity 0.89 and specificity 0.97 (Oh et al., 2018); the BAI is not primarily a screening tool.
05 - How it compares
How it compares to the alternatives.
Instrument
Items
Time
When to reach for it
BAI
21
5–10 min
Licensed (Pearson). Reach for it when a somatically weighted, self-report anxiety severity measure is specifically wanted and licensing is already in place.
-Settings that already hold a Pearson licence and have BAI baselines to compare against.
-Research continuity with the large historical BAI literature.
-When separating anxiety from depressive symptomatology matters.
Reach for something else when
-Routine screening in new services: the free GAD-7 covers this without licensing cost.
-Generalised anxiety where worry, not arousal, is the presenting feature.
-Medically unwell or older patients, where somatic items inflate scores.
-Any web or paper reproduction of items without a licence: the items may not be republished.
07 - Confidence & precision
Reading the score with care.
The BAI's internal consistency (α ≈ 0.92–0.94) is excellent, but one-week test–retest of 0.75 means individual scores move noticeably on repetition even without true change. No widely accepted minimal clinically important difference has been established for the BAI; the manual's severity bands are clinical conventions rather than empirically derived change thresholds.
In practice, small shifts within a band should not be over-interpreted, and change judgements are better anchored to band shifts plus clinical assessment. Note also that the severity cut points were set from clinical experience, not ROC analysis against diagnosis.
08 - Limitations
What it cannot tell you.
Copyrighted and fee-per-use (Pearson); items may not be reproduced, which limits use in low-resource and digital settings.
Heavy somatic weighting inflates scores in medical illness and older adults, and underrepresents worry - it is a weaker match for GAD.
A severity measure, not a diagnostic or screening instrument; severity bands were not validated against diagnosis.
One-week reporting window makes it a state measure; scores fluctuate with recent events.
Self-report; subject to the usual reporting biases.
See how Aisel removes friction where it costs most. A 20-minute walkthrough tailored to your clinic.
We value your privacy
We use cookies to analyse site usage and improve your experience. Analytics and embedded media (e.g. YouTube) only load if you accept. Read our cookie policy.