Across networks of cold weather sites, eddy covariance measurements of ecosystem fluxes commonly indicate carbon uptake throughout winter, a problem that has been linked to open-path infrared gas analyzer (IRGA) self-heating. Though there is no definitive consensus about this phenomenon, the de facto correction for the prevalent LI-7500 is the methodology of Burba et al. (2008) (B08), which predicts self-heating from IRGA surface temperatures and boundary layers. In this paper we examine the validity of this model compared to an updated correction by conducting field studies with the LI-7500 and an IRGA designed by the National Oceanic and Atmospheric Administration (NOAA). Both types of sensors were instrumented to measure sensible heat flux within the instrument path and temperatures of key instrument surfaces. We test hypotheses that (1) considerable IRGA self-heating occurs at a cold, windy, high elevation, mountainous field site, (2) unwarranted boundary layer adjustment terms defined in B08 reduce the bottom and top surface heat fluxes by an order of magnitude, and (3) the weighted sum of the surface heat fluxes should be proportional to their surface area contribution to the IRGA measurement volume. We detected more sensible heat flux within the LI7500 path than outside of it (statistical slopes between 0.95 to 1.3 and offsets between 0 and 35 W m−2) with the largest discrepancies occurring during the day and in winter. The NOAA IRGA has a similar structure to the LI-7500, but with flat surfaces that do not warrant the boundary layer adjustment terms in B08. Bayesian analyses indicated the bottom and top surface heat flux weightings from B08 are improbable and an order of magnitude too large. For the LI-7500, when the boundary layer adjustment terms were omitted the B08 weightings were also improbable; by including these terms and using the new weightings, self-heating estimates were reasonable only because the bottom and top sensible heat fluxes became negligible. We conclude that it is likely the B08 self-heating correction has two errors that roughly cancel out at our field site; this should not generally be expected. The new formulation emphasizes the role of the spar in self-heating, which should be taken into consideration for all cold-weather installations regardless of how the IRGA is mounted.