The Photochemistry Experiment in BERLIOZ (PHOEBE) was conducted in July/August 1998 at a rural site located near the small village Pabstthum, about 50 km NW of downtown Berlin. More than 60 nonmethane hydrocarbons (NMHC) in the range of C2--C10 were measured using two in situ gas chromatography (GC) systems. The first (GC1) was capable of measuring C2--C10 hydrocarbons with a relatively high separation efficiency but low time resolution (80--90 min), while GC2 provided quasi-continuous measurements of C5--C10 hydrocarbons with a time resolution of 20 min but with a poorer separation efficiency than GC1. The advantages of both systems were joined by interpolation between two data points of GC1 with the pattern given by GC2. For compounds that could not be reliably measured with GC2, patterns of compounds with similar reactivity were used. Air masses with the lowest photochemical age as estimated from the toluene/benzene ratio and the highest hydrocarbon mixing ratios were observed on 20 and 21 July when air was advected from the direction of Berlin. Alkanes were the most abundant hydrocarbons (~60%) on a molecular basis, followed by alkenes and aromatics. The reactivity of the hydrocarbons toward OH was dominated by the alkenes (>60%), with isoprene and α-pinene constituting the major part. The hydrocarbon data were used together with the other trace gases measured at Pabstthum to simulate OH, HO2, and RO2 concentrations with the condensed chemical box model RACM. Relatively good agreement of the simulated radical concentrations with the spectroscopic measurements made at Pabstthum is observed for NOx mixing ratios >5 ppb, whereas the model overestimates OH and HO2 by 100% and 40%, respectively, at low NOx. The discrepancy between measured and modeled OH does not correlate with the concentration of particles. The RO2 concentrations are in good agreement with the measurements over the entire range of NOx. Sensitivity studies show that peroxyacetyl nitrate (PAN) is an important radical source and that missing volatile organic compound (VOC) reactivity is an unlikely explanation for the overestimation of HOx: By doubling of the VOC reactivity, OH and HO2 can be brought into agreement. However, the model then overestimates the organic RO2 concentrations by almost a factor of 2. Another important finding is that RACM overestimates the measured NO/NO2 ratio by 25%. This and the overestimation of HO2 lead to an overprediction of the local ozone formation rate by about 40% at low NOx mixing ratios.