Error bars and shaded interval bands are the most basic tools for showing a plausible range

What it is

Error bars (short vertical segments on bars or points showing the range of variation) and interval bands (a semi-transparent ribbon along a line chart showing the plausible range as x varies) are the most fundamental and widely used means of expressing uncertainty in visualization. They share one design logic: overlay a visual element encoding the range of variation on top of the existing encoding (position, length), letting the reader see both "where the point estimate is" and "how far it might deviate" without introducing a new chart type. They suit different graph forms—error bars for discrete bars or points, bands for continuous lines.

Why it happens

Their effectiveness comes from minimal intrusion into existing visual channels: the point's position encoding is unchanged (the estimate stays in place), with a segment or ribbon added to encode a second variable (uncertainty). Readers can ignore this added layer and continue reading the point estimate, or extract the interval when needed—this "ignorable but extractable" property means uncertainty display need not destroy baseline readability. Their limitations come from the same design: error bars carry low visual weight (a thin segment is easily missed across a full chart), and readers' default assumptions about "what the error bar represents" (standard deviation? standard error? 95% confidence interval? range?) are inconsistent—the same graphic may mean entirely different intervals to different readers and must be stated explicitly in the caption or tooltip. Interval bands hit occlusion problems on multi-series charts: once two bands overlap, the reader struggles to attribute the overlap to which series and which side.

Where it stops holding

Error bars and bands express symmetric, unimodal uncertainty well; for skewed distributions (right-skewed income data), bimodal, or multimodal distributions, a symmetric interval misleads—violin plots or box plots are better substitutes. With many categories, error bars turn visual "caterpillar" (too many vertical segments interfering with each other), where bands or gradient encodings work better. Another boundary is reader interpretation: research shows a substantial proportion of readers cannot correctly interpret error bars (taking the bar's end as a data point, or not knowing what interval it represents); the encoding itself cannot substitute for reader training or caption clarity.

Applying it

• Default to error bars on bar or scatter charts and interval bands on line charts when uncertainty must be shown; never mix the two in one chart.
• State in the caption or tooltip what the error bar/band represents (e.g., "95% confidence interval" or "±1 standard deviation").
• On multi-series bands, use semi-transparent fills with distinct hues per series so overlap regions remain attributable through color mixing.
• Verification: show the chart to a reader who has not seen the caption and ask "what do the error bars represent?" A mismatch with the actual meaning means the caption is insufficient.

Related

• Same group: U8.01.1 Point estimates hide interval information · U8.01.3 Omitting uncertainty claims certainty
• Nearby: U8.01.1 Point estimates hide interval information · U1.09.1 Each visual channel supports only a limited number of reliably distinguishable levels
• Search terms: error bars · confidence band · uncertainty encoding