Domain-specific beats domain-general: anxiety in N≈800 high-achieving Russian adolescents
- **Single anxiety factor, but not the useful one.** Across eight measures (trait, state, GAD-7, math anxiety, spatial anxiety, social anxiety, worry, anxiety sensitivity), a single general factor explained **51% of variance** in N≈800 schoolchildren (M_age = 15.26) selected for top achievement in STEM, Arts, Sports, or Literature.
- **Domain-general anxiety did not predict grades.** Trait anxiety and GAD symptoms accounted for negligible variance in 13 academic subjects after adjustment. The clinical screener everyone uses (GAD-7) tells you nothing about how this kid is performing in school.
- **Domain-specific anxiety did the work.** Math anxiety predicted Algebra and Geometry performance specifically — not general grades. Effect size: small-to-medium. Domain-matching mattered more than overall anxiety load.
- **Worry correlated positively with performance** after controlling for the other measures — pointing to an arousal/motivational signal embedded inside the worry construct that standard scales conflate with pathology.
This is an HSE Saint Petersburg/Goldsmiths/Beijing Normal collaboration — a quietly competent psychometric study from a Russian developmental lab that tells us something most clinicians get wrong in intake. We screen with one general anxiety scale (GAD-7, Beck, STAI), get a number, and use it to set treatment intensity. The data here suggest that for high-functioning adolescents, the general number is decoupled from real-world impairment in the very domain they care about — academics.
What the data shows
The sample is unusual: ~800 Russian schoolchildren pre-selected for excellence in one of four tracks. That stratification is the point — it lets the authors test whether anxiety still bites in adolescents who, by definition, are outperforming peers. Answer: yes, but selectively. The single-factor solution is statistically clean (51% variance), which means general anxiety is a real construct. But when you regress school grades on it, the predictive power collapses. Specificity returns the signal: math anxiety hits math grades. Spatial anxiety, separately measured, does its own narrow work. The worry subscale's positive coefficient is the most clinically interesting finding — it captures something that looks like cognitive engagement, not distress.
For your practice
Two operational shifts. First, when an adolescent presents with academic decline alongside anxiety symptoms, do not stop at GAD-7. Ask domain-specific questions: "What happens when you open a math problem?" vs "What happens before a literature exam?" The pattern of where anxiety lands tells you where the intervention goes — graded exposure to math problem-solving is a different protocol than social-evaluation exposure before oral presentation. Second, do not pathologise worry uniformly. Pre-test worry that drives review is functionally different from post-test rumination. The standard "reduce worry" framing of CBT for GAD can flatten a useful arousal pattern if applied without specificity to a high-performing teenager.
A normal GAD-7 in a high-achieving adolescent does not mean the anxiety is not there — it means you measured the wrong domain.
Cross-sectional, self-report, single-country sample selected for high achievement — generalisability to clinical populations and average-achieving adolescents is limited. The "worry helps performance" finding needs replication before clinical use.