What did Taylor and Rohrer (2010) find with 4th-graders and prism problems?

Interleaved practice scored 77% vs 38% for blocked on a next-day test (d ≈ 1.21) — and the entire gain came from eliminating discrimination errors.

Taylor and Rohrer had fourth-graders practice four types of prism problem either blocked or interleaved. A day later the interleavers more than doubled the blockers' score (77% vs 38%, an effect size of roughly d = 1.21 — very large). The revealing detail was why the blockers failed: not because they'd forgotten how to execute each procedure, but because they applied the wrong procedure — a discrimination error. Interleaving trained them to first ask "which kind is this?", the exact skill blocking skips.