Viewers of a graph will readily interpret its contents, even when given no explicit instructions regarding what information to extract. However, little is known about the strategies that subjects adopt when making such spontaneous interpretations. In the present experiments, subjects studied single-function line graphs for self-determined periods. They provided written interpretations immediately following examination of each graph. The structural complexity of stimulus graphs was varied by eliminating symmetry, and by adding data points, departures from linearity, and trend reversals. Across two experiments, number of trend reversals was the main determinant of comprehension difficulty as measured by study times. An increased number of reversals also resulted in more local, detail-oriented content in interpretations. By contrast, the presence of such emergent features as symmetry and linearity led to increases in the amount of integrative, global content in interpretations, usually at the expense of local detail. Surprisingly, increases in the number of data points led to similar increases in the grain of subjects' interpretations. The last finding may reflect a shift from point-by-point to integrative study strategies necessitated by capacity limitations in working memory.Viewers of a graph will generally conclude that they understand the display's message, even when given no specific instructions about what information to extract. Advocates of graphical communication have argued that this readiness to make spontaneous interpretations is a key reason for favouring graphs over other forms of data reduction and representation: graphs can provide answers to specific questions, while also suggesting new questions to answer (e.g. Cleveland, 1985;Tukey, 1977). Yet little is known about the strategies people use in making spontaneous interpretations, or about the properties of graphical stimuli that guide these strategies. Even more fundamentally, little is known about how consistently various types of interpretive statements are made. What commonalities exist among individuals' criteria for graphical comprehension?