In this paper, we examine 3 different models used to estimate turnover of soil organic carbon (SOC) fractions using radiocarbon measurements: one conventional carbon dating model and two bomb 14C models. One of the bomb 14C models uses an atmospheric 14C record for the period 22,050 BC to AD 2003 and is solved by numerical methods, while the other assumes a constant 14C content of the atmosphere and is solved analytically. The estimates of SOC turnover obtained by the conventional 14C dating model differed substantially from those obtained by the bomb 14C models, which we attribute to the simplifying assumption of the conventional 14C model that the whole SOC fraction is of the same age. The assumptions underlying the bomb 14C models are more applicable to SOC fractions; therefore, the calculated turnover times are considered to be more reliable. We used Monte Carlo simulations to estimate the uncertainties of the turnover times calculated with the numerically solved 14C model, accounting not only for measurement errors but also for uncertainties introduced from assumptions of constant input and uncertainties in the 14C content of the CO2 assimilated by plants. The resulting uncertainties depend on systematic deviations in the atmospheric 14C record for SOC fractions with a fast turnover. Therefore, the use of the bomb 14C models can be problematic when SOC fractions with a fast turnover are analyzed, whereas the relative uncertainty of the turnover estimates turned out to be smaller than 30% when the turnover time of the SOC fractions analyzed was longer than 30 yr, and smaller than 15% when the turnover time was longer than 100 yr.