Samuelson’s Dictum argues that aggregate stock markets do not convincingly reflect information on fundamentals, such as dividends or earnings, and are, thus, inefficient in setting prices. By contrast, firm-level stock prices share a much closer connection with fundamentals and are, therefore, deemed relatively efficient. This paper presents an alternative explanation: Knightian uncertainty stemming from historically unique events produces differential impacts on investor forecasting bounds at the aggregate versus firm level. Several uncertainty proxies are employed from millions of unscheduled events identified in the universe of Dow Jones & Co. financial news reports over the last twenty years. Uncertainty based on count and proportion of unscheduled events shares a significant inverse relationship with the range of forecasting bounds on returns and earnings at both the aggregate and firm level suggesting informational effects may dominate. However, when uncertainty is measured based on the diversity of unscheduled event groups, the forecasting bounds for broad stock market returns widen, while those at the firm level narrow, implying model ambiguity dominates at the aggregate level. These findings are robust for the majority of Dow Jones Industrial Average 30 firms and for alternative categories of unscheduled events.