SummaryThe paper begins by arguing that much contemporary quantitative analysis of mortality change in populations, such as plotting life expectancy against time, involves 'thinking in slices', analogous to trying to understand complex spatial variation through first plotting (say) mean height as a function of latitude, then as a function of longitude. By contrast Lexis surfaces, a visual method, encourage 'thinking in surfaces', in which the aim of exploratory data analysis is initially not on intelligent data reduction, but on feature recognition.The paper then illustrates the use of Lexis surfaces through two case studies. In both case studies, three types of Lexis surface visualisation are presented: Shaded Levelplots (SLPs), which facilitate within population comparison; Comparative Levelplots (CLPs), which facilitate comparisons between any two populations; and schematic representations of key features, which facilitate discussion and communication of findings about patterns identified in the first two types of plot.The paper concludes by discussing some of the substantive implications of features identified in these case studies, and by suggesting that using Lexis surfaces effectively for population data exploration involves iterating between informal and formal, and geometric and aetiological, models of reasoning about population data and the implications thereof.