Only the bottom series of a stacked chart owns the common baseline

What it is

Common-baseline privilege describes a geometric property of conventional nonnegative stacked bars and areas: only the bottom series is measured from the same fixed zero line at every horizontal position. Every upper layer begins at the cumulative value of the layers below, so its origin moves. The outer boundary represents the total, but it is not a common baseline for the internal series. The statement does not mean that the bottom layer wins every possible task; it concerns direct comparison of component magnitude.

Why it happens

For layer i, the lower boundary is the sum of all preceding layers and the upper boundary adds the layer's own value. Reading the bottom layer requires locating one endpoint against zero. Reading an upper one requires tracking two locations and taking their difference. The component is encoded by vertical thickness, not by filled area and not by the height or slope of either boundary alone. Changes below translate both boundaries, so they may rise together while the component stays constant. Following only the upper edge therefore confounds a cumulative position with the component value.

Studying it

Experiments can render the same nonnegative data as grouped bars, target layers at different stack positions, and small multiples, then ask for values, differences, rankings, or change directions. Useful manipulations include target position, number of layers, thickness, volatility below the target, display size, gridlines, and interactive readouts. Outcomes include error, time, confidence, and gaze transitions. Shared-scale position often supports more direct magnitude judgment, but effect sizes depend on the task and stimulus; a percentage from one study is not a universal stacking penalty. Visual readings should also be compared with table and summary alternatives.

Where it stops holding

Upper segments in a stacked bar still expose segment length within one category; they are not devoid of information. The harder case is accurate comparison across categories or time. Diverging stacks place positive and negative values on opposite sides of zero, giving each side an axis-adjacent layer while requiring an explicit rule for cumulative signs. Conventional stacked areas usually assume additive nonnegative quantities. Missing observations must not silently become zero. A streamgraph with a moving, centred baseline deliberately gives up a fixed zero origin and is better suited to onset, duration, and overall flow than precise totals.

Applying it

• State whether the primary task concerns the total, one component, or composition. Use grouped bars, small multiples, or a selectable standalone view for precise component comparisons.
• If one series has priority, place it next to the fixed baseline and still provide values or a keyboard-accessible data table; do not rely on colour alone.
• Use an explicit diverging rule for negatives and show gaps or missing markers for absent observations rather than folding either case into an ordinary nonnegative stack.
• Test representative cross-position questions. Check whether readers confuse boundary direction with component direction, and validate visual, keyboard, and screen-reader routes separately.

Related

• Same group: U2.04.2 Upper series read only by band thickness, at markedly lower precision · U2.04.3 Stacking suits totals, not per-series trends · U2.04.4 Percent stacking discards the totals · U2.04.5 Stack order decides which series stay readable
• Nearby: U1.02.1 Position along a common scale is the most accurate channel · U2.15.1 Small multiples replace overlay with repeated like-for-like panels
• Search terms: common baseline · floating baseline · stacked chart · diverging stack