Putting two curves on one chart makes readers assume one drives the other, with no words needed

What it is

Place two curves on the same chart (ad spend on one, sales on the other) and—even with no causal language anywhere—readers will automatically interpret "the former drives the latter." This is a rhetorical effect inherent to juxtaposition: shared axes, time alignment, color linkage, vertical stacking—every one of these design devices implies a direct relationship between the variables, while the data itself may be two independently rising time series. The causal implication of juxtaposition is not readers "over-interpreting"; it is the natural reading of visual narrative grammar. Whether the chart maker realizes this determines whether juxtaposition is honest evidence display or invisible causal insinuation.

Why it happens

Juxtaposition implies causation because human causal perception is automatic: people have a strong attribution tendency toward "one change following another" (of Hume's conditions for causality, temporal succession and spatial contiguity operate at the perceptual level), and juxtaposition manufactures that contiguity artificially—shared axes, shared time windows, visual binding. The classic experimental finding from contingency-judgment tasks is that even when told two variables are unrelated, juxtaposed presentation significantly raises causal judgments. Graphic design amplifies the effect: dual-axis alignment (making the lines' inflection points coincide, implying linkage), vertical stacking (the spatial relation "above affects below"), same-hue coloring (color binding implying one mechanism). The countermeasure is not abandoning juxtaposition (it is the fundamental tool for discovering covariation) but making the relationship's nature explicit: captions stating "covariation shown; causation not established," visually separating the two lines (independent axes for different units, avoiding color binding), and dedicated notation when expressing "correlated but unproven causation" ("observational data" beside the correlation coefficient). The ethical boundary is intent: juxtaposition during exploration is hypothesis generation; juxtaposition in communication without stating the relationship's nature presents a hypothesis as a conclusion.

Where it stops holding

The strength of juxtaposition's causal hint varies with design features: time-aligned dual-axis overlay is the strongest hint; two separate small charts side by side is the weakest. Reader expertise moderates it—statistically trained readers have some resistance to the hint but are not fully immune. The presentation of correlation coefficients carries hints too: displaying r=0.85 on the chart is honest information, but combined with a causal title ("X boosted Y"), the statistics endorse the causal insinuation. The substantive basis for judging "is this juxtaposition misleading" is the presence of third variables (confounders): when known confounders suffice to explain the covariation (ice cream sales and drowning deaths rising together in summer—temperature is the confound), juxtaposition without naming the confound misleads; when major confounders have been controlled or discussed before juxtaposition, it is legitimate evidence display.

Applying it

• When communication charts juxtapose two variables, the caption must state the relationship's nature: "covariation shown; causation not established," or list the major confounders.
• Use dual-axis charts with care: their independently scalable axes make fabricated linkage easiest; when necessary, label "two independent vertical axes."
• Distinguish exploratory juxtaposition (hunting covariation) from presentational juxtaposition (drawing conclusions): the former uses no causal wording; the latter must state the relationship's nature.
• Verification: after viewing a juxtaposed chart, ask readers to write down the relationship between the two variables; if more than half write causal statements while the caption claimed none, the juxtaposition constitutes suggestive misleading.

Related

• Same group: U10.03.2 Uncontrolled variables must be stated explicitly · U10.03.3 Narrative wording often exceeds what the data supports
• Nearby: U2.05.3 Correlation is not causation · U10.03.2 Uncontrolled variables must be stated explicitly
• Search terms: correlation vs causation · dual axis · implied causality