Reference strains of Streptococcus pyogenes and strains from recent epidemics and sporadic cases of scarlet fever were examined for their ability to produce erythrogenic toxin type A (ET A) by ELISA and double immunodiffusion (Ouchterlony) using an anti-ET A antibody purified by affinity chromatography. Of the reference strains (most of them isolated before 1945) 16/51 produced more or less ET A (Table 1). ET A synthesis is strain-specific, but not type-specific. Well-known toxin producers like the strains NY-5; 594 or "Smith" produce up to 16.000 micrograms/l under optimal culture conditions. Type 3 strains isolated from scarlet fever patients during the outbreak 1972/73 seem to belong to one clone as evidenced by the uniform SDS-PAGE pattern: They were found to produce 5-200 micrograms/l (mean 68 micrograms/l) ET A only. Type 3 strains from sporadic cases, isolated 10 years later, produced 0-138 micrograms/l (mean 30 micrograms/l). Strains of the type 1 clone, causing the epidemic in 1982/83 produced only 0.75-10 micrograms/l (mean 8 micrograms/l) ET A (Table 3). Only a few strains of S. pyogenes isolated 1984 or later synthesized ET A but they were found more often to produce ET B (proteinase precursor) in batch cultures. S. pyogenes strains seem to have lost their ability to produce large amounts of ET A during the last decades. Because this toxin must be considered as a pathogenicity factor the decrease in toxin production may be one reason for the present mild form of scarlet fever.