For central banks, conducting policy in an environment of uncertainty is a daily fact of life. This uncertainty can take many forms, ranging from incomplete knowledge of the correct economic model and data to future economic and geopolitical events whose precise magnitudes and effects cannot be known with certainty. The objective of this paper is to summarize and compare the main results that have emerged in the literature on optimal monetary policy under uncertainty with actual central bank behaviour. To this end, three examples are studied in which uncertainty played a significant role in the Bank of Canada's policy decision, to see how closely they align with the predictions from the literature. Three principles emerge from this analysis. First, some circumstances - such as when the policy rate is at risk of being constrained by the effective lower bound - should lead the central bank to be more pre-emptive in moving interest rates, whereas others can rationalize more of a wait-and-see approach. In the latter case, the key challenge is finding the right balance between waiting for additional information and not falling behind the curve. Second, the starting-point level of inflation can matter for how accommodative or restrictive policy is relative to the same situation without uncertainty, if there are thresholds in the central bank's preferences associated with specific ranges for the target variable, such as the risk of inflation falling outside of the inflation control range. Third, policy decisions should be disciplined, where possible, by formal modelling and simulation exercises in order to support robustness and consistency in decision making over time. The paper concludes with a set of suggested areas for future research.