Quartz/feldspar fragment (‘clast’) sizes were measured in thin sections of three types of fault zone rocks, e.g., melting-dominated pseudotachylyte (M-Pt), crushing-dominated pseudotachylyte (C-Pt) and cataclasite (Ct), from two well-studied Precambrian shear/fault zones in the Indian craton (e.g., the Gavilgarh–Tan Shear zone in central India and the Sarwar–Junia Fault zone in western India). Logarithmic plots of clast area vs. cumulative frequency in the pseudotachylytes demonstrate a fractal clast-size distribution (c.s.d.) for the intermediate size range, whereas the finer and coarser clast size fractions clearly deviate from the fractal trend. Under-representation of the finer size clasts in the pseudotachylyte samples may be attributed to their preferential melting and removal from the clast population. The relative paucity of coarse clasts, on the other hand, is possibly due to a sampling bias against coarse clasts. The c.s.d of the cataclastic rock shows a multi-fractal character with two different slopes (i.e., lower D-value for finer clast sizes) and absence of the left-hand (finer size) fall off. This suggests less efficient crushing in the finer clast size fraction. The proportion of clasts, compared to the matrix, is very small in M-Pt, increases in C-Pt and is highest in Ct, suggesting that melting of rock/mineral fragments is a dominant process in forming M-Pt, whereas it is less significant in C-Pt, and is absent in Ct, which corroborates the microscopic observations.