We performed this meta-analysis to estimate the associations of maternal exposure to PM2.5 and its chemical constituents with birth weight and to explore the sources of heterogeneity in regard to the findings of these associations. A total of 32 studies were identified by searching the MEDLINE, PUBMED, Embase, China Biological Medicine and Wanfang electronic databases before April 2015. We estimated the statistically significant associations of reduced birth weight (β = -15.9 g, 95% CI: -26.8, -5.0) and LBW (OR = 1.090, 95% CI: 1.032, 1.150) with PM2.5 exposure (per 10 μg/m(3) increment) during the entire pregnancy. Trimester-specific analyses showed negative associations between birth weight and PM2.5 exposure during the second (β = -12.6 g) and third (β = -10.0 g) trimesters. Other subgroup analyses indicated significantly different pooled-effect estimates of PM2.5 exposure on birth weight in studies with different exposure assessment methods, study designs and study settings. We further observed large differences in the pooled effect estimates of the PM2.5 chemical constituents for birth weight decrease and LBW. We concluded that PM2.5 exposure during pregnancy was associated with lower birth weight, and late pregnancy might be the critical window. Some specific PM2.5 constituents may have larger toxic effects on fetal weight. Exposure assessment methods, study designs and study settings might be important sources of the heterogeneity among the included studies.