Direct deposition is the only method that can be utilised for the standardisation of radioactive solutions because there is no deposition loss on a support. The present study investigated how much the roughness of the support influences the emission rate of alpha particles for direct deposition sources from the practical viewpoint of use of the method. A number of samples were prepared by evaporating a 0.1-ml aliquot of a dilute nitric acid aqueous solution that contained (241)Am on stainless steel supports with four different surface conditions; untreated supports were either polished (using metal abrasives available commercially) or not polished, and buffed supports (grid size of #400) were either polished or not polished. Alpha spectrometry of the samples revealed that the detection efficiency was significantly different between the non-polished and polished supports; the former was lower by 3 % than the latter for both the untreated and buffed supports. Microscopic observations clarified that the counting loss was attributed to irregular flaws or polishing lines on the non-polished supports, most of which were found to be in the order of submicron in depth and were diminished on the polished supports. One may usually assume that a direct deposition source offers no counting loss if its entire energy spectrum is seen above a low-energy discrimination limit of the spectrometer. However, this should be experimentally confirmed using a solution with known activity. It was difficult to identify the counting loss for the buffed supports without polishing because their energy spectra showed little degradation.