BACKGROUND: Proof is lacking that pollen count is representative for allergen exposure, also because allergens were found in nonpollen-bearing fractions of ambient air. OBJECTIVE: We monitored simultaneously birch pollen and the major birch pollen allergen Bet v 1 in different size fractions of ambient air from 2004 till 2007 in Munich, Germany. METHODS: Air was sampled with a ChemVol high-volume cascade impactor equipped with stages for particulate matter (PM)>10 microm, 10 microm>PM>2.5 microm, and 2.5 microm>PM>0.12 microm. Allergen was determined with a Bet v 1-specific ELISA. Pollen count was assessed with a Burkard pollen trap. We also measured the development of allergen in pollen during ripening. RESULTS: About 93 +/- 3% of Bet v 1 was found in the PM > 10 microm fraction, the fraction containing birch pollen. We did not measure any Bet v 1 in 2.5 microm > PM > 0.12 microm. Either in Munich no allergen was in this fraction or the allergen was absorbed to diesel soot particles that also deposit in this fraction. Pollen released 115% more Bet v 1 in 2007 than in 2004. Also within 1 year, the release of allergen from the same amount of pollen varied more than 10-fold between different days. This difference was explained by a rapidly increasing expression of Bet v 1 in pollen in the week just before pollination. Depending on the day the pollen is released during ripening, its potency varies. CONCLUSION: In general, pollen count and allergen in ambient air follow the same temporal trends. However, because a 10-fold difference can exist in allergen potency of birch pollen, symptoms might be difficult to correlate with pollen counts, but perhaps better with allergen exposure.