Background: Health effects of exposure to particulate matter (PM) on male reproductive health remain unclear. Only a limited number of studies have investigated the effects of PM2.5 or PM10 exposure on semen quality, and the results were largely inconsistent. Objectives: To quantitatively assess the exposure-response association between PM exposure and semen quality in Chinese men who were exposed to a wide concentration range of PM. Methods: We investigated 1759 men from Wuhan, China, who were partners of women undergoing assisted reproductive technology procedures, and had semen examined at least once between 2013 and 2015. Individual PM2.5 and PM10 exposures during 0-90, 0-9, 10-14 and 70-90 days before each semen examination (corresponding to the entire and three key periods of sperm development, respectively) were retrospectively estimated by inverse distance weighting interpolation. Linear mixed models were used to assess exposure-response relations of PM exposure with sperm concentration, count and motility. Results: PM2.5 exposure during 0-90 lag days ranged from 27.3 to 172.4 mu g/m3. It was linearly and inversely associated with sperm concentration (beta: -0.20; 95% CI: -0.34, -0.07) and count (-0.22; -0.35, -0.08). For the three key exposure periods, only PM2.5 exposure during the 70-90 lag days was significantly associated with sperm concentration (-0.12; -0.22, -0.03) and count (-0.12; -0.21, -0.02). Sensitivity analyses for a subgroup (n = 1146) excluding subjects with abnormal sperm concentration, count or motility yielded similar results. Compared with PM2.5, we found generally similar associations for PM10 exposure in relation to sperm concentration and count, except that the associations appeared to be nonlinear with inverted J-shaped relationships. Neither PM2.5 nor PM10 exposure was significantly associated with sperm motility (all p > 0.05). Conclusions: Our results suggest that ambient PM exposure during sperm development adversely affects semen quality, in particular sperm concentration and count. (C) 2016 Elsevier Ltd. All rights reserved.