Heat transfer in internal combustion engines is taking on greater importance as manufacturers strive to increase efficiency while keeping pollutant emissions low and maintaining adequate performance. Wall heat transfer is experimentally evaluated using temperature measurements both on and below the surface using a physical model of conduction in the wall. Three classes of model inversion are used to recover heat flux from surface temperature measurements: analytical methods, numerical methods and inverse heat conduction methods; the latter method has not been previously applied to engine data. This paper details the inherent assumptions behind, required steps for implementation of, and merits and weaknesses of these heat flux calculation methods. The analytical methods, which have been most commonly employed for engine data, are shown to suffer from sensitivity to measurement noise that requires a priori signal filtering. The numerical methods are computationally demanding and do not offer any advantage over the analytical methods unless the wall is a composite structure. The inverse methods are found to be relatively insensitive to noise because of the use of future-time information, the inclusion of a regularization term that relaxes the exact-matching constraint, and the fact that finite derivatives are not required to estimate the heat flux.