What depth of learning can policy appraisal stimulate? How we can account for the survival policies that are known to pose significant countervailing risks? While heralded as a panacea to the inherent ambiguity of the political world, the proposition pursued is that policy appraisal processes intended to help decision-makers learn may actually be counterproductive. Rather than simulating policy-oriented learning, appraisals may reduce policy actors’ capacity to think clearly about the policy at hand. By encouraging a variety of epistemic inputs from a plurality of sources and shoehorning knowledge development into a specified timeframe, policy appraisal may leave decision-makers overloaded with conflicting information and evidence which dates rapidly. In such circumstances, they to fall back on institutionalised ways of thinking even when confronted with evidence of significant mismatches between policy objectives and the consequences of the planned course of action. Here learning is ‘single-loop’ rather than ‘double-loop’—focussed on adjustments in policy strategy rather than re-thinking the underlying policy goals. Using insights into new institutional economics, the paper explores how the results of policy appraisals in technically complex issues are mediated by institutionalised ‘rules of the game’ which feed back positively around initial policy frames and early interpretations of what constitutes policy success. Empirical evidence from UK biofuels policy appraisal confirms the usefulness of accounts that attend to the temporal tensions that exist between policy and knowledge development. Adopting an institutional approach that emphasises path dependence does not however preclude the possibility that the depth of decision-makers’ learning might change. Rather, the biofuels case suggests that moves towards deeper learning may be affected by reviews of appraisal evidence led by actors beyond immediate organizational context with Chief Scientific Advisers within government emerging as potentially powerful catalysts in this acquisition of learning capabilities.