Order judgements are slower and less accurate when reversed. That is, when participants see two events in a sequence (e.g., circle aEuro broken vertical bar square), they are quicker to report 'Before' statements (e.g., "Circle before Square") than 'After' statements ("Square after Circle"). The current study sought to determine whether a reversal effect will also occur when participants are instructed to produce a sequence of responses. Twenty participants were trained to criterion on simple 'Before' and 'After' instructions that specified sequences of two responses (e.g., "Circle before Square"). In a subsequent test, participants produced instructed sequences (e.g., circle aEuro broken vertical bar square) more quickly and more reliably when instructed to choose one stimulus before another than when they were requested to choose one stimulus after another. The implications of these findings for current theories of relational responding are considered.