We are developing practical methodologies to characterize pool sizes and residence times for fractions of soil organic matter (SOM) using radiocarbon, with a particular focus on SOM in New Zealand pasture soils that responds to global change on decadal timescales. As single mean residence times for the entire SOM pool can be misleading or uninterpretable, we focus on the use of samples collected about 7 and 40 yr after the bomb14C spike to separate SOM into at least 2 pools. These results from a box model methodology yield sensible estimates of the proportion of “passive” SOM, and the residence time of the dominant pool with approximately decadal residence times. These results are supported by chemical analysis. Approximately 45-yr residence times of light-fraction SOM in a relatively infertile soil contrast with ∼16-yr residence times in a more fertile soil, and correspond to large differences in the proportion of lignin- and polysaccharide-derived SOM in these soils measured using pyrolysis-GC/MS. To achieve greater detail and assess the degree to which “active” SOM with annual turnover rates may bias results from the simple model, we use density as a means of isolating SOM with different degrees of mineral association. Initial results from grazed pasture soils sampled in 2003–4 emphasize that isolating non-mineral-associated light fractions can improve understanding, but may be less important than identifying fractions associated with unique mineralogy. In this soil, a fraction with density ≥2.55 g/mL shows much larger proportions of passive SOM than other fractions.