Little is known about spatial variability of microbial activity, particularly at microscales. This is especially true for the fate and degradation of pesticides. The objective was to sample soil from micro to field scales and apply geostatistics on the potential mineralization of a widely used herbicide 2,4-dichlorophenoxy acetic acid (2,4-D; C8H6Cl2O3). Soil cores were sampled in the plow layer of a cultivated soil with a systematic sampling procedure. In a first experiment 2,4-D mineralization was measured on 39 crushed cores and we analyzed variability of mineralization at the field scale, from decameter to meter and from meter to decimeter scale. In a second experiment, 432 soil cubes (about 216 mm3) were used to study the variability of mineralization at the "microhabitat" scale from meter to millimeter. The spatial dependence of 2,4-D mineralization was first quantified by computing an empirical variogram function. Spatial independence was then tested by comparing the empirical variogram function to its individual confidence bounds at 95% level obtained under independence assumption by a Monte-Carlo Method. The potential for 2,4-D mineralization was spatially heterogeneous from field to microhabitat. Mineralization variability increased when the scale decreased from field to microhabitats. Specifically, the coefficient of variability (CV) of 2,4-D mineralization was of 18.5% at the field scale, 7–22% at the meter scale, 47.9% at the inter-cores scale, and 25–160% at the intra-core scale (microhabitat scale). 2,4-D mineralization was spatially structured only at the microhabitat scale (hot spots). This variability at fine scales should be considered when sampling soil processes involving microbial activities.