Use of cumulative frequency graphs for the visual comparison of tool assemblages is a common device among Palaeolithic archaeologists; however, there are problems in their interpretation which are frequently ignored. All archaeological material is subject to unavoidable sample bias, due to excavational procedure, and to the possibility of errors in typological identification; these points are touched on only briefly being largely of archaeological interest. More seriously, statistical problems are involved in the application of analyses suitable only for ordinal data to data which are inherently nominal. Also, the use of percentages of tool types entails the problem of the inter-dependence of percentile variables and obscures sample bias by utilizing relative frequencies of types, which may suggest non-existent parallels between types. Finally, perceptual difficulties arise in the comparison of curves, both in the distinction of null categories from small percentiles, often a critical point, and in the presentation of apparently similar assemblages which may contain more dissimilarity than is visually apparent. Graphs for several classic assemblages of differing cultural attribution are presented both in usual typological order and re-ordered on the criterion on maximal difference between type frequencies to demonstrate the change in visual appreciation based on greatest visible separation.