Charts That Overclaim: A Pre-Publication Review for Student Researchers

A scientific figure is rarely a neutral illustration. It is an argument about what the data show, how confident the authors are, and why the reader should care. In student labs, figure-making often receives less scrutiny than statistical analysis or manuscript prose. A plot is produced late in the project, adjusted until it looks clean, and inserted into a poster or paper. Yet the same plot can quietly overstate an effect, hide the number of observations, or make a result look more stable than it is. For undergraduate researchers preparing campus symposium posters and early-career researchers preparing first submissions in 2026, a pre-publication figure review is a form of science communication and a form of quality control.

How honest figures fail

Most misleading figures are not deliberate. They fail through three ordinary mechanisms. The first is encoding exaggeration: the visual channel makes a difference look larger than the underlying data warrant. The second is uncertainty omission: the figure shows a summary without the variability, sample size, or model assumptions needed to interpret it. The third is reproducibility loss: the figure cannot be regenerated from the data and code available to the team, so nobody can check whether a later version changed the message.

These failures matter because readers do not usually audit a figure the way they audit a methods section. They absorb the visual impression first. A bar chart with a truncated axis, an error bar with no definition, or a line drawn through categorical groups can create a conclusion that the underlying analysis does not support.

Five common failures in student labs

The first is the truncated bar chart. Bars encode magnitude through length, so a bar chart should normally begin at zero. Starting a bar axis at a high value can make a modest difference appear dramatic. Line charts and dot plots may use a truncated axis when the question concerns change rather than absolute magnitude, but the choice should be explicit and the axis should be labeled clearly.

The second is the undefined error bar. An error bar can represent a standard deviation, a standard error, a confidence interval, or a range. Readers cannot know which unless the caption says so. The choice also changes the visual message. A standard error can look much smaller than a standard deviation or a confidence interval. State the statistic, the sample size, and what the interval is intended to convey.

The third is aggregation that hides the data. A bar chart of group means can conceal small groups, outliers, and bimodal distributions. When possible, show individual observations alongside the summary. A dot plot or strip plot with a mean and interval often communicates more honestly than a bar chart alone.

The fourth is the dual-axis plot. Two vertical axes can make two unrelated series appear tightly linked because the author has chosen scales that create visual alignment. If the relationship matters, plot the variables against each other or use small multiples with shared axes. Do not ask readers to infer correlation from two independently scaled lines.

The fifth is decorative encoding. Three-dimensional bars, pies with many slices, rainbow colormaps, and color used without meaning all add perceptual noise. They can also create accessibility problems. Prefer position and length over angle and area, use colorblind-safe palettes, and make sure color is not the only way to distinguish categories.

A pre-publication figure checklist

Before a figure leaves the lab, the author and at least one reader should answer the following questions.

• Does the figure title state the finding rather than the topic?
• Is the sample size visible or stated in the caption?
• Is uncertainty shown, and is its meaning defined?
• Is the axis range appropriate for the chart type, with zero included for bars and area encodings?
• Are units, transformations, and excluded observations explained?
• Are colors accessible and used consistently across figures?
• Can the figure be understood from its caption without reading the main text?
• Does the caption identify the statistical test, the error bar definition, and the number of replicates?
• Does the figure match the claims in the text, including the direction and size of the effect?
• Would a different but defensible scale change the conclusion?
• Is the chart type the simplest one that supports the comparison?
• Can the figure be regenerated from a script and the archived data?

This checklist is not a substitute for statistical review. It is a communication review. Its purpose is to catch the visual choices that a reader will notice before they notice the methods.

Redesign principles for clearer plots

First, show distributions before summaries. A mean with an interval is useful, but it is more honest when readers can also see the observations that produced it. Second, choose encodings by data type. Position along a common scale is easier to compare than angle, area, or volume. Third, make uncertainty legible. Use intervals that answer the question the figure is asking, and define them in the caption.

Fourth, use small multiples instead of dual axes when comparing trends across groups. Fifth, annotate intentionally. A short label that identifies an outlier, a treatment window, or a threshold can prevent misreading, but annotations should not become a substitute for a clear caption. Sixth, treat the caption as part of the methods. It should state what was measured, how many observations are shown, what the error bars represent, and what software or code produced the figure.

Finally, build a reproducible figure pipeline. Keep the data and the plotting script together. Name figures by purpose rather than by draft number. Archive the version that appears in the manuscript. A reproducible figure is easier to revise, easier to audit, and easier for a collaborator to understand.

A culture of figure review

Student labs can make figure review routine without making it heavy. Add a five-minute figure check to a lab meeting before a conference deadline. Ask a peer who did not make the plot to read the caption aloud and describe what they think the figure shows. If their description differs from the author's intended claim, the figure needs revision. This practice is especially useful for undergraduate posters, where space is limited and a single misleading axis can dominate the viewer's impression.

Honest data visualization is not about making figures dull. It is about making the visual argument match the evidence. A clear figure can be elegant, memorable, and persuasive. It should also be reproducible, accessible, and candid about uncertainty. The goal is not to avoid persuasion; it is to ensure that the persuasion is earned by the data.