Position encodings tolerate truncation when clearly marked

What it is

Dot plots and line charts — position encodings — do not depend on an origin: what readers judge is "where the mark falls on the scale," not "how far it extends from zero." So they may legitimately truncate the axis: tighten the range to the interval where the data lives, and local resolution rises, revealing detail in the fluctuations. The price is declaring the truncation explicitly: the starting tick prominently visible, the endpoint values labelled. The formula for legitimate truncation is "position encoding + clear marking"; drop either condition and legitimacy collapses into the next leaf's misleading.

Why it happens

Position judgements are local: comparing the relative height of two lines or reading where a point falls references nearby ticks, not the origin, so cutting away the far reaches of the scale damages nothing and magnifies effective resolution (each pixel spans less value). But "position needs no zero" carries a reader-side hazard: the proportional habit trained by bar charts migrates over, and some readers still read differences in line height as length ratios — hence the marking duty is heavier than intuition suggests: the axis start must carry its tick value in a visible position, with a note if zero lies off-screen; leave only an unlabelled axis line and the truncation becomes a pit.

Where it stops holding

The exemption covers mark position only; once the chart fills (area fill between line and axis, band fills joining points), the filled region reverts to area encoding and the zero-baseline duty returns — area fills on line charts need care. Turning bars into "lollipops" does not dodge the rule: a thin stem plus a dot is still read as length, and readers still interpret proportionally. Interactive settings can discharge the marking duty with zoom affordances (fold marks at axis ends, range labels); static media rely on ticks and words alone.

Applying it

• When tightening the range on a line or dot chart, show the axis-start tick with its value made conspicuous (larger type or an end-of-axis range note).
• Add reference lines (mean, target) to anchor judgement in the zoomed view, easing "where am I" disorientation.
• Verification: ask readers "where does the axis start?" and "by what magnitude do the two lines differ?"; a blank on the first means under-marking, a proportional mis-answer on the second means residual length habit — fix both.

Related

• Same group: U3.01.1 Length encodings must start at zero · U3.01.3 Axis truncation is the most common misleading device
• Nearby: U2.03.1 Line charts suit trends over ordered variables · U3.05.3 Direct labelling can replace gridlines and legends
• Search terms: axis truncation · position encoding · value-range zoom · scale breaks