Formaldehyde (CH2O) measurements from two independent instruments are compared with photochemical box model calculations. The measurements were made on the National Oceanic and Atmospheric Administration P-3 aircraft as part of the 1997 North Atlantic Regional Experiment (NARE 97). The data set considered here consists of air masses sampled between 0 and 8 km over the North Atlantic Ocean which do not show recent influence from emissions or transport. These air masses therefore should be in photochemical steady state with respect to CH2O when constrained by the other P-3 measurements, and methane oxidation was expected to be the predominant source of CH2O in these air masses. For this data set both instruments measured identical CH2O concentrations to within 40 parts per trillion by volume (pptv) on average over the 0--800 pptv range, although differences larger than the combined 2σ total uncertainty estimates were observed between the two instruments in 11% of the data. Both instruments produced higher CH2O concentrations than the model in more than 90% of this data set, with a median measured-modeled 2O> difference of 0.13 or 0.18 ppbv (depending on the instrument), or about a factor of 2. Such large differences cannot be accounted for by varying model input parameters within their respective uncertainty ranges. After examining the possible reasons for the model-measurement discrepancy, we conclude that there are probably one or more additional unknown sources of CH2O in the North Atlantic troposphere.