Small-scale indentation and floe-splitting experiments were conducted on columnar ice floes of various sizes and at different speeds. During low-speed indentation (0.2–8 mm s−1 ), the ice floes always split apart, while at higher indentation speeds (> 100 mm s−1 ) they did not. The reason is attributed to differences in the process of deformation and failure. At low speed, a large zone of microcracked ice forms in front of the indentor. Development of compressive stresses in the microcracked ice zone leads to buildup of transverse forces that drive crack propagation. These zones of microcracked ice are not observed during high-speed indentation. Rather, the ice fails by continuous crushing. The theoretical effective pressure required to split an ice floe, as predicted by Bhat (1988), agrees to some extent with those measured during experiments.