Technical Abstract: Woody encroachment has dramatically changed land cover patterns in arid and semiarid systems (drylands) worldwide over the past 150 years. This change is known to influence bulk soil carbon (C) pools, but the implications for dynamics and stability of these pools are not well understood. Working in a Chihuahuan Desert C4 grassland encroached by C3 creosote bush (Larrea tridentata), we used two density fractionation techniques (2 and 7 pool density fractionations) and isotopic analysis to quantify changes in C pools and dynamics among vegetation microsites typical of an encroachment scenario (remnant intact grassland, shrub subcanopies, and in shrub intercanopy spaces within a shrubencroached area). The C concentration of bulk soils varied with microsite, with almost twice the C in shrub subcanopies as in intercanopy spaces or remnant grasslands. Estimated SOC accumulation rates from Larrea encroachment (4.79 g C m-2 year-1 under canopies and 1.75 g C m-2 year-1 when intercanopy losses were taken into account) were lower than reported for higher productivity Prosopis systems, but still represent a potentially large regional C sink. The composition of soil C varied among microsites, with the shrub subcanopy C composed of proportionally more light fraction C (\1.85 g cm-3) and C that was soluble in sodium polytungstate. Grassland soils contained very little carbonate C compared to shrub subcanopies or shrub intercanopy spaces. Stable isotope analyses indicate that inputs from C3 shrubs were incorporated into all density fractions, even in heavy fractions in which shrub inputs did not change overall C concentration. The seven density fractionation yielded unexpected d13C patterns, where the two heaviest fractions were strongly depleted in 13C, indicating strong fractionation following organic matter inputs. These results suggest that the utility of isotope mixing models for determining input sources may be limited in systems with similar fractionation patterns. We propose a five pool model for dryland soil C that includes a relatively dynamic light fraction, aggregate and heavy fractions that are stable in size but that reflect dynamic inputs and outputs, a potentially large and seasonally dynamic pool of soluble C, and a large pool of carbonate C. Combined, these results suggest that dryland soil C pools may be more dynamic than previously recognized.