Knowing the complete results of all conducted studies on a question of interest is important to avoid publication and selective outcome reporting biases1 and to obtain a reliable picture of the evidence. However, in many types of research, the problem of selectively missing information does not pertain only to studies that have already been performed. A greater challenge is to understand how many potential studies do not exist, when they could readily have been conducted. Is it possible to know the unknowable? It is important to consider why studies have not been conducted and what could be done about them. Related to this phenomenon is the continuing problem of unregistered, exploratory observational studies often derived from large data sets.