Weekly values of the net O3 production efficiency (OPE), defined as the net number of O3 molecules produced per molecule of NOx (NO+NO2) consumed, are estimated from a 1990--1994 record of O3, NOx, NOy, CO, and C2H2 concentrations at Harvard Forest, Massachusetts. The OPE is inferred from the slope ΔO3/Δ(NOy--NOx) of the linear regression between O3 and NOy--NOx concentrations (NOy is the sum of NOx and its oxidation products); and alternatively from the slopes ΔO3/ΔCO and ΔO3/ΔC2H2 multiplied by regional estimates of the CO/NOx and C2H2/NOx emission ratios. The mean OPE values inferred from ΔO3/Δ(NOy-NOx) are 3--5 times higher than those inferred from ΔO3/ΔCO or ΔO3/ΔC2H2; the discrepancy may be due to the effects of HNO3 and O3 deposition and also to uncertainties in the CO/NOx and C2H2/NOx emission ratios. The relative seasonal trends of the OPE derived from ΔO3/Δ(NOy--NOx), ΔO3/ΔCO, and ΔO3/ΔC2H2 are, however, similar. Thus ΔO3/Δ(NOy--NOx) increases from about 4 mol/mol in May to 8 mol/mol in June--July, and gradually decreases back to 4 mol/mol by early October. The sharp rise of the OPE from May to June is attributed to onset of emission of the biogenic hydrocarbon isoprene. The decline from July to October is attributed to decreases in isoprene emission and in solar radiation. The O3 background at Harvard Forest, defined by the y intercept of the O3 versus NOy-NOx regression line, decreases from 40 ppbv in May to 25 ppbv in September, consistent with observations at remote sites in northern midlatitudes. The seasonal trend in the background explains why mean O3 concentrations at Harvard Forest peak in May--June even though the OPE peaks in June--July. ¿ American Geophysical Union 1996