Step 1: In a factorial experiment the number of treatment combinations can be large, and it may not be practical to keep all of them in one homogeneous block.
Step 2: Confounding is the technique of deliberately mixing up certain higher order interaction effects with the differences between blocks, so that those interactions cannot be estimated separately from block effects.
Step 3: By sacrificing information on the confounded interaction(s), the treatment combinations can now be split into smaller, more manageable and homogeneous blocks.
Step 4: So the direct purpose of confounding is to allow a reduction in block size while still keeping the design in a single replicate or a small number of replicates.
Final Answer: (C) Blocks.