An observationally constrained box model has been used to investigate the chemistry of the marine boundary layer at the Mace Head Atmospheric Research Station, a remote site on the west coast of Ireland. The model aims to simulate concentrations of the hydroxyl (OH) and hydroperoxy (HO2) radicals measured by an in situ fluorescence assay by gas expansion instrument, and the sum of peroxy radicals ([HO2] + S[RO2]) as determined by a peroxy radical chemical amplification instrument. The model has been constructed based on observed concentrations of nonmethane hydrocarbons, measured in situ during the campaign by gas chromatography. The chemical mechanism for the model is a subset of a comprehensive master chemical mechanism. This paper details comparisons of the concentrations of modeled and measured radical species from a field campaign held at Mace Head during spring 1997. The air masses arriving at the site have been split into three categories depending on their origin and chemical characteristics and model-measurement comparisons carried out for each air mass. The average model-measurement ratios are 2.4 for [OH], 3.6 for [HO2], and 0.9 for ([HO2] + S[RO2]), respectively, between 1100 and 1500 hours: the level of agreement is better for all three sets of radicals in the cleanest air mass. Possible reasons for the observed discrepancies are discussed. In addition, a rate of production analysis is used to identify key OH and HO2 reactions in the three air masses. The rate of OH production from HO2 with NO exceeds that from ozone photolysis by factors of 2-6 in the polluted air masses studied. In cleaner air from the northern polar region, primary production from ozone photolysis exceeds that from HO2 + NO by a factor of 2.5. HO2 and CH3O2 dominate the peroxy radical composition in all air masses, but peroxy radicals derived from the oxidation of nonmethane hydrocarbons are more important in polluted air masses.