OBJECTIVE Collagen is one of the major endogenous skin fluorophores. Alteration in the structure of collagen due to chronic ultraviolet radiation (UVR) exposure may influence the intensity of the autofluorescence. The aim of this study was to investigate the relation between collagen-linked autofluorescence and sun exposure to clarify whether the skin can be used as a biological UVR dosimeter. METHODS We conducted an in vivo study with 131 healthy volunteers. Fluorescence was measured from sun-exposed (dorsal forearm, forehead and shoulder) and sun-protected (buttock) skin and corrected for the impact of pigmentation and redness. The excitation wavelengths (Ex) and emission wavelengths (Em) were: Ex330:Em370, Ex330:Em455 and Ex370:Em455 nm. Individual UVR exposure data were collected both retrospectively and prospectively using questionnaires and electronic personal UVR dosimeters for a summer period. RESULTS Age, but not sex, skin type or smoking habits correlated significantly positively with skin autofluorescence at Ex370:Em455 at all body sites (P<0.001, r(2)=0.08-0.26), and at Ex330:Em455 only at the buttock (P=0.001, r(2)=0.08), whereas age was not correlated with Ex330:Em370. Sun-protected buttock skin had significantly higher autofluorescence than sun-exposed skin (P-values<0.0001). Because of great between-subject differences in autofluorescence at different body sites, and because the autofluorescence at the unexposed buttock represents the baseline value, individual correction of skin autofluorescence measurement with that of the buttock was performed. Different measures of individual chronic cumulative UVR doses correlated significantly negatively with the skin autofluorescence ratio (F(ratio)), but the correlations were poor (r(2)=0.03-0.10). CONCLUSIONS The results indicate that the collagen-linked skin F(ratio) might be best to use as a measure of individual photodamage, a UVR dose effect, and that it is also a better marker of individual cumulative UVR dose than the used UVR exposure measurements. The methods used to obtain UVR exposure data might not be sensitive and specific enough.