Fine root (<2 mm) cycling rates are important for understanding plant ecology and carbon fluxes in forests, but they are difficult to determine and remain uncertain. This paper synthesizes minirhizotron and isotopic data and a root model and concludes that (1) fine roots have a spectrum of turnover times ranging from months to many years and (2) the mean age of live root biomass (A) and the mean age of roots when they die (i.e., their turnover time (τ)) are not equal. We estimated A and τ of fine roots in three forests using the root model Radix. For short-lived roots, we constrained τ with existing minirhizotron data; for long-lived roots, we used new radiocarbon measurements of roots sampled by diameter size class and root branch order. Long-lived root pools had site mean τ of 8–13 y and 5–9 y when sampled by diameter and branch order, respectively. Mean turnover times across sites were in general not significantly different as a function of branch-order, size class, or depth. Our modeling results indicate that ∼20% of fine root biomass has turnover times of about a year, and ∼80% has decadal turnover times. This partitioning is reflected in our predicted mean ages of ∼9 y and turnover times of ∼3 y. We estimate that fine root mortality contributes between 38 and 104 g C m−2 y−1 to soil in these forests. These estimates are 20 to 80% of previous estimates in these and similar forests, in part because we explicitly account for the large portion of fine-root biomass with decadal cycling rates. Our work shows that both fast and slow cycling roots must be modeled jointly to account for the heterogeneous nature of fine-root dynamics.