We link specific mechanisms of biogenous sediment mixing with the commonly used bioturbation coefficient (Db) that describes their bulk effects. Using an isotropic, stationary, unbiased random walk model we mechanistically decompose the particulate bioturbation coefficient into the fundamental dimensions of length and time. The result shows that Db depends directly on the square of the distance particles are moved (step length) and inversely on the elapsed time between movements (rest period). This new decomposition in terms of explicit mechanisms (i.e., animal activities), leads to scaling arguments that large, deposit feeding animals will in nearly all cases dominate biogenous mixing. Paradoxically, such animals often transport particles vertically in an advective fashion (e.g., conveyor-belt feeding), making the widespread fit of the diffusion equation to tracer profiles equivocal. Finite-difference simulations reveal that even in the complete absence of vertical diffusion, rapid diffusive horizontal mixing coupled with vertical advection can produce vertical profiles characteristic of diffusion. We suggest that near-surface horizontal mixing rates by animals far exceed vertical mixing rates in the same stratum and that this anisotropy may persist throughout the surface mixed layer. Thus, despite their apparently good kinematic fit, one-dimensional biodiffusion coefficients may not accurately describe the dynamics of sediment displacement, leading to errors in models of early diagenesis.