X-ray diffraction (XRD)-based modified Warren–Averbach (MWA) analysis, in comparison with the Williamson–Hall (WH) analysis, was applied to 0.3 mass% carbon martensitic steels, as-quenched and subsequently tempered at various temperatures, to give their dislocation densities. For the as-quenched martensite, the WH method gives a value of around 2.0×1016 m−2, which could be overestimated. Meanwhile, the MWA method gives a value of around 6.3×1015 m−2, which is below the possible upper limit of dislocation density, 1016 m−2. The MWA-derived value for the as-quenched steel seems to be 1.6–4.8 times higher than those expected from the precedent results derived by transmission electron microscope (TEM) observations. However, considering that the TEM-derived value gives the microscopically local average while the XRD-derived value gives the macroscopic average, such discrepancy between the TEM-derived value and MWA-derived value is tolerable. For the steels tempered at 723 K and 923 K, the MWA and WH methods give comparable values ranging in 1014 m−2, where the rearrangement of dislocation structure is observed by TEM. However, in these steels where the XRD peaks are narrower and the instrumental width of the present XRD system could be significant, care should be taken over the peak width correction.