The size of atmospheric black carbon (BC) is a key determinant of its health and climate impacts. However, how BC sources vary with particle size remains poorly understood. We conducted radiocarbon (<sup>14</sup>C)-based source apportionment of atmospheric BC in Guangzhou, China, across six discrete particle-size fractions from < 0.49 μm to 7.2–10 μm during winter and summer. The dataset comprises 81 Δ<sup>14</sup>C measurements, resolving fossil-fuel and biomass-burning contributions across particle sizes. Additional δ<sup>13</sup>C measurements of four representative winter sample sets were incorporated into a Bayesian model to apportion fossil-derived BC between coal and petroleum sources. Fossil sources dominated BC, contributing 91 ± 3 % in winter and 73 ± 1 % in summer. In both seasons, the fossil contribution first increased and then decreased with increasing particle size,<span> </span>peaking in the 0.95–1.5 μm fraction (95 % in winter and 88 % in summer) and declining to its lowest level in the 7.2–10 μm fraction (82 % in winter and 50 % in summer). Coal and petroleum made comparable contributions to BC in the selected winter samples, with no consistent particle-size-dependent trends. Thus, the particle-size dependence of BC sources was driven mainly by an increasing biomass-burning contribution toward the coarse fraction, rather than by a systematic shift in the relative contributions of coal and petroleum. These findings show that bulk source fractions do not fully capture the source composition of BC across particle sizes, underscoring the need to account for source-specific size distributions when interpreting BC transport, removal, and associated climate and health effects.