Large amounts of data on the concentration of peroxides have been collected in vertical profiles over the North Atlantic Ocean by a Hercules aircraft. The measured peroxide concentrations have been compared with concentrations calculated by a simple algorithm derived assuming that the standing peroxide concentration is in equilibrium with its products and loss processes. In clean air where the peroxide and ozone concentrations are anticorrelated throughout the profile measured and calculated peroxide concentrations coincide, whereas in layers of polluted air within the profile, as determined from positive ozone peroxide correlations, calculated peroxide concentrations are greatly in excess of measured values. Using the degree of correlation between measured and calculated peroxide concentrations as a diagnostic, it is possible to show that many aspects of the seasonal cycle of ozone are caused by in situ tropospheric chemistry. Thus the summer minimum in the seasonal cycle of ozone, observed at clean marine ground-based sites such as Mace Head, is due to photochemical destruction, and elevated levels of ozone are associated with the transport of polluted air, on occasion over thousands of kilometers. Of particular interest if our analysis is correct, the broad maximum of ozone occurring between March and May at ground-based sites has a large contribution from ozone formed by tropospheric as well as stratospheric chemistry.