The goal in real-time scheduling is to satisfy the timing requirements of application jobs which often have hard deadlines. There are two aspects to Ada's scheduling policies which are detrimental to achieving this goal. First, Ada's constraints on the language's implementation limit the definition of priority and the task scheduling algorithm to preclude the use of the best algorithms for scheduling jobs with hard deadlines. Second, information about task priority is not used when selecting a task from an entry queue or when choosing among branches of a selective wait statement. Instead, FIFO and arbitrary disciplines are used, respectively, which can unnecessarily lead to missed deadlines, even for very low levels of processor utilization. We suggest some areas for change to make the language more suitable for building real-time systems.