When safety signs are in service they can be damaged beyond the point of being useful by sudden damage or due to gradual natural conditions, causing physical marks, defects, fading, discolouration or blurring of the sign. The damaged signs may consequently be much less effective in providing timely information about safety threats or risks. There has been limited research on the effects of damaged safety signs on user performance. To partially fill this gap, this study investigated recognition performance for damaged safety signs with different levels of color deterioration. Fifteen safety signs were chosen for study and damage was simulated, with different levels of color deterioration by using bandpass filters (Photoshop CS® software) to create 15 different levels for the ratio of white pixels to total pixels. Fifty Hong Kong Chinese males (21 - 45 years old) familiarized themselves with all the test sign referents first, and then the signs were presented in random order with each sign shown progressively from the most filtered to the complete version. Participants consistently waited to accumulate sufficient perceptual evidence before making an affirmative decision about the sign meaning. Accurate identification decisions mainly occurred around and between the seventh level pixel ratio (77.14%) and the ninth level (82.86%). The grand mean image level at which safety signs were correctly identified was 8.23 with an identification threshold (pixel ratio) of 80.68%. This implied that at a pixel ratio lower than this identification threshold, a sign may not be identified correctly and should be restored or replaced as soon as possible. The lifespan of a safety sign in future might also be determined through consideration of the estimated prescribed level of identification threshold for damaged signs. Overall, the findings of this study should help further develop and assist implementation of safety sign maintenance programmes and management systems from the perspective of human factors and ergonomics.