Fatigue in materials is generally associated with the production of heat, leading to the "self-heating" of the tested material. The associated heat power density, named mechanical dissipation or intrinsic dissipation, can be deduced from the temperature changes captured on the tested specimen's surface by infrared thermography. When mechanical dissipation is spatially homogeneous in the tested specimen, the processing can be performed using a macroscopic approach, also named zero-dimensional (0D) approach. The latter uses an averaged temperature over the whole specimen's measurement zone. The present study aims to analyze the error generated by the 0D approach in the assessment of mechanical dissipation. This error was measured with respect to a one-dimensional (1D) approach, which is applicable for longitudinal specimens subjected to uniaxial loading. Experimental tests were performed on pure copper and acrylic glass. A model was also developed to analyze the influence of the material and of the heat exchanges with the specimen's environment. The results obtained show that the error generated by the 0D approach in mechanical dissipation measurement may not be negligible, and that attention should be paid to the choice of approach for fatigue analysis.