Are surveys biased because we prefer to mark the left-hand side of questionnaires? (Via Marginal Revolution.)