To avoid biases inherent in the study of referral patients seen at a single unit, we investigated a group of 1500 patients (males 71%), members of the UK National Ankylosing Spondylitis Society (NASS). Patients were divided into 15 cohorts by year of onset (mean size 91). The median delay in diagnosis for these cohorts was greater than 6 years until 1974. Thereafter, there was a marked improvement with the most recent cohort waiting a median of 2 years for diagnosis (e.g., 1969/72 cohorts vs 1979/82 cohorts; mean delay 7.4 years vs 2.3 years, p less than 0.0001). Until 1974 the median delay for women was 9 years compared to 6 years for men (e.g., 1956/62 cohorts 11 and 7 years, p less than 0.0015, falling to 3 and 2.5 years, respectively by 1979/82, p = NS). Major differences were seen when the 14 health regions were taken separately; the median delay ranged from 9 years for the 3 "worst" regions to 4 years for the 3 "best" (p = 0.022). This was reflected in the percentage with delay in diagnosis longer than 2 years (80 and 58%, respectively; chi 2 = 11.31, p less than 0.01). Strikingly, the median age of onset of symptoms steadily increased from 18 years in 1930/40 to 28 years in 1981/82 (e.g., 1956/62 vs 1969/72; mean ages 21.4 vs 23.5 years, p less than 0.004 and 1969/72 vs 1979/82; 23.5 years vs 28.7 years p less than 0.001).(ABSTRACT TRUNCATED AT 250 WORDS)