Axis truncation is the most common misleading device

What it is

Truncation sits permanently at the top of the misleading-technique list: without touching data or encoding, simply move the axis start from zero toward the data, and differences multiply before the eye. Its popularity comes from the low barrier — one default parameter does it, and the author can even plead "the chart is correct." This leaf owns the mechanics and the detection: how the amplification is computed, how readers self-check, how standards intercept it at the source; recognizing the manipulator's intent and the social consequences belongs to the ethics family.

Why it happens

Amplification lives in the denominator: the smaller the visible range, the larger the share of screen the same value difference occupies. Amplification factor ≈ data range / visible range — a 90-versus-100 difference occupies 10% under a zero baseline but 50% when the axis starts at 80, a fivefold boost. On bar charts it counterfeits proportion (readers read "nearly double" off the length ratio); on line charts it inflates fluctuation, rendering noise as drama. Detection is a fixed three steps: find the axis-start value, mentally compute the amplification factor, rescale the conclusion by that factor. Truncation also enables a statistical disguise: with chosen endpoints, any flat series can be tuned to read "surge" or "collapse."

Where it stops holding

Not every truncation misleads: legitimate truncation (previous leaf) — position encoding plus explicit marking — serves resolution and is transparent and checkable. Judgement rests on three probes: encoding type (length encoding + truncation = always wrong), marking level (a hidden start tick = suspect), and beneficiary direction (the truncation happens to amplify the author's thesis = high risk). Reverse truncation (axis starting below zero) exists too, merely rarer. Tolerance for "well-meaning truncation" should not extend to mass-audience settings — readers lack the checking habit, so the defence belongs on the chart-maker's side.

Applying it

• Fix the reader's self-check routine: locate the axis start first; for any length encoding with a non-zero start, divide the visible difference by the amplification factor before concluding.
• Intercept on the production side: add an "axis-start check" to the release pipeline (block non-zero starts on length encodings), and archive axis parameters with sensitive charts for later audit.
• Verification: show truncated and untruncated versions of the same data, record readers' differing conclusions; a significant difference aligned with the author's thesis empirically demonstrates the misleading effect.

Related

• Same group: U3.01.1 Length encodings must start at zero · U3.01.2 Position encodings tolerate truncation when clearly marked
• Nearby: U3.02.2 Across the break, equal distances no longer mean equal values · U3.04.2 Dual axes can fabricate correlation that does not exist
• Search terms: truncated axis · misleading visualization · distortion factor · axis manipulation