Alignment effects on learning multiple, use-relevant classification systems