Measurements of formaldehyde (CH2O) from a tunable diode laser absorption spectrometer (TDLAS) were acquired onboard the NASA DC-8 aircraft during the summer 2004 INTEX-NA campaign to test our understanding of convection and CH2O production mechanisms in the upper troposphere (UT, 6-12 km) over continental North America and the North Atlantic Ocean. The present study utilizes these TDLAS measurements and results from a box model to (1) establish sets of conditions by which to distinguish "background" UT CH2O levels from those perturbed by convection and other causes; (2) quantify the CH2O precursor budgets for both air mass types; (3) quantify the fraction of time that the UT CH2O measurements over North America and North Atlantic are perturbed during the summer of 2004; (4) provide estimates for the fraction of time that such perturbed CH2O levels are caused by direct convection of boundary layer CH2O and/or convection of CH2O precursors; (5) assess the ability of box models to reproduce the CH2O measurements; and (6) examine CH2O and HO2 relationships in the presence of enhanced NO. Multiple tracers were used to arrive at a set of UT CH2O background and perturbed air mass periods, and 46% of the TDLAS measurements fell within the latter category. In general, production of CH2O from CH4 was found to be the dominant source term, even in perturbed air masses. This was followed by production from methyl hydroperoxide, methanol, PAN-type compounds, and ketones, in descending order of their contribution. At least 70% to 73% of the elevated UT observations were caused by enhanced production from CH2O precursors rather than direct transport of CH2O from the boundary layer. In the, presence of elevated NO, there was a definite trend in the CH2O measurement-model discrepancy, and this was highly correlated with HO2 measurement-model discrepancies in the UT.