Appearance
Small multiples
Working notes.
Small multiples repeat the same visual form across many units — countries, institutions, time windows, variables — so that comparison happens through juxtaposition rather than overlay. The form assumes that the reader can hold a single encoding scheme in mind and scan across panels for pattern, divergence, and exception.
When they work
Small multiples work when the number of panels is manageable and the scale is consistent across all of them. The reader compares shape, not absolute value — a rising line in panel A against a flat line in panel B, a cluster of outliers in one corner of the grid against uniformity elsewhere. The form is strongest when the unit of comparison is homogeneous: the same metric, the same time range, the same axis bounds. Homogeneity is what makes juxtaposition informative rather than misleading.
They also work when the analytical question is distributional or typological rather than precise. How many units show this pattern? Which units are exceptions? Is there a visible gradient across the set? Small multiples answer these questions through visual search, not through reading individual values.
When they fail
The form fails when there are too many panels. Beyond a certain count — which depends on panel complexity and display size — the grid becomes a texture rather than a readable set. The reader cannot hold individual panels in attention; the form collapses into an impression of overall density or chaos without supporting specific comparison.
They fail when scales differ across panels without explicit signalling. If one panel's y-axis runs to 100 and another's to 10, shape comparison becomes misleading. The assumption of shared scale is load-bearing; breaking it silently breaks the form.
They fail when there is no anchor — no reference panel, no highlighted unit, no summary statistic that orients the reader before the scan begins. Without an anchor, small multiples present equivalence among units that may not be analytically equivalent. The form's neutrality is a feature and a risk.
What they reveal about visualisation assumptions
Small multiples make explicit something that other forms hide: the decision about what constitutes a comparable unit. Choosing to show one panel per country embeds the assumption that the country is the right grain of analysis. Choosing one panel per decade embeds a temporal assumption. The grid is not neutral; it is a statement about what should be held constant and what should vary.
They also reveal the trade-off between detail and overview. A single complex chart can show interrelationships among variables within one unit. Small multiples sacrifice that within-unit complexity for across-unit comparison. The choice between the two is not a technical one; it is a question about what the analysis needs to make visible.
These notes are incomplete. They record current thinking rather than settled method.
Working notes. Not a finished piece.