To better understand the physical links among hysteresis properties, defect distributions, and grain size, Barkhausen jumps have been studied in individual platelets of natural hematite from Elba, Italy during hysteresis. Both Bitter patterns and hysteresis curves have been investigated in two very different groups of platelets: large, ~1-mm-sized platelets with coercive forces (Hc) in the range of ~20--50 Oe (2--5 mT), and much smaller, ~100 μm-sized platelets, with Hc in the 100--140 Oe range (10--14 mT). Single platelets were cycled through both major and minor hysteresis, in order to determine (1) the changes of magnetic moment caused by Barkhausen jumps and (2) how these changes may depend on grain size and the critical field Hcrit required to unpin a wall from a defect. Results of these experiments lead to the following conclusions. First, the steepest part of the major hysteresis loop is dominated by large-scale wall motion through that part of the grain where the combination of wall nucleation and wall propulsion across defects requires the lowest fields anywhere in the particle. In the large platelets, this soft region (e.g., where Hcrit≤~50 Oe, or 5 mT) can amount to as much as 25% of the grain's volume; within this region, a wall can be driven well over 100 μm by fields commensurate with Hc. In the large platelets, the defects within this soft region appear to have quite variable volume densities. By contrast, in the small platelets, defects with comparably soft wall-pinning strengths appear to be distributed much more uniformly. Second, in both platelet groups, most of the relatively hard defects (e.g., Hcrit≥~80 Oe) which have the greatest impact on wall pinning during hysteresis generally appear to be concentrated within a localized portion of the particle, rather than being distributed randomly throughout the volume. This hard region can amount to several tens of percent of a grain's volume, but within it, the hard defects appear to be distributed rather homogeneously. It is through this hard region that the wall is forced, as fields stronger than Hc drive the platelet toward saturation. Consequently, these results lead to a very different picture of defect distributions than envisioned by previous models, based on random defects. We propose that this nonrandom spatial distribution of hard versus soft defects could originate from internal strain generated when these platelets cleaved naturally from their much larger, parent crystals. ¿ 1998 American Geophysical Union