Light source selection using rough surface illumination is crucial when especially high accuracy roughness measurements are required. It is therefore important to establish an extensive relationship between the wavelength of the illuminations, speckle size and roughness. Therefore, in this paper, an experimental approach for surface roughness evaluation based on the autocorrelation analysis of the spectral properties of speckle patterns caused by milled metal surfaces is reported. The speckle patterns obtained by the illumination of three 633, 604 and 543 nm wavelengths of He Ne laser were also analyzed. It was found that wavelength depended speckle correlation analysis (WDSCA) is very sensitive to small variations in speckle sizes caused by spectral properties of speckle patterns such as increasing the wavelength lead to increased speckle sizes. The results are in good agreement with the results obtained from the mechanical stylus profilometer for the milled metal surfaces with roughness values between R-a; 0.36 and 1.98 mu m. The technique reported here has a great potential for precise and non-contact optical measurements of rough surfaces.