The spatial distribution, depths and diameters of roots in soil are difficult to quantify but important to know when reinforcement of a rooted slope or the stability of a plant is to be assessed. Previous work has shown that roots can be detected from the depth–resistance trace measured using a penetrometer with an adapted blade-shaped tip. Theoretical models exist to predict both forces and root displacements associated with root failure in either bending or tension. However, these studies were performed in dry sand under laboratory conditions, using acrylonitrile butadiene styrene root analogues rather than real roots. In this paper blade penetrometer field testing on two forested field sites, with Sitka spruce and pedunculate oak in sandy silt and clayey silt, respectively, is used to evaluate models under field conditions. Root breakages could be detected from blade penetrometer depth–resistance traces and using complementary acoustic measurements. Predictions of additional penetrometer resistance at root failure were more accurate than the displacement predictions. An analytical cable model, assuming roots are flexible and fail in tension, provided the best predictions for Sitka roots, whereas thick oak roots were better predicted assuming bending failure. These matched the modes of failure observed in three-point bending tests of the root material in each case. The presence of significant amounts of gravel made it sometimes difficult to distinguish between hitting a root or a stone. The root diameter could be predicted when root strength and stiffness, and soil penetrometer resistance were known and the right interpretative model was selected. Estimates based on peak force were more accurate than those based on root displacement. This measurement procedure is therefore a potentially valuable tool to quantify the spatial distribution of roots and their reinforcement potential in the field.