Research & PolicyFree
Research Figure Review and Chart Redesign
Send the figures and the claim each one is meant to support, get the mismatches named and the redesign specified.
River asks one question per figure before looking at it: what claim is this meant to support. Then it checks whether the comparison that claim requires is the easiest thing on the chart or the hardest. Most failures are that mismatch rather than bad taste. A gap that narrows is plotted as two series the reader has to subtract; a three-point difference is plotted on a truncated axis that draws it as a doubling. Each figure comes back with the mismatch named and the redesign specified.
Search this and page one is chart-type galleries and colour-palette advice. Pick the right chart for your data, avoid pie charts, use a colourblind-safe palette. All true, all downstream of the actual problem, and none of it asks what the figure is arguing. A figure can obey every rule on those pages and still bury its own finding, which is why reviewers write comments about figures that authors read as matters of taste.
Built for the author preparing a submission, the one answering a reviewer who criticised a figure, and anyone who inherited a co-author's chart and cannot tell what it argues. It reads the numbers behind each figure, so it works from a cleaning log that records every decision and the output block that a results section is written from. Where a figure carries a small subgroup, the feasibility check on that subgroup decides whether it can be shown at all.
Three of nine figures did the job
A manuscript with nine figures, sorted by whether the claimed comparison is the easy read or the hard one. Three make it the easiest thing on the chart. Four make it the hardest available. One supports a different claim than its own caption states, and one carries no claim at all. Three of nine work as intended, which is 33.3 per cent, and the remaining six are not badly drawn. They are drawn for a different argument than the one in the text.
Figure 2 is the instructive one. Two group means, 71.2 and 74.3 on a 0 to 100 scale, so an absolute difference of 3.1 points and a relative difference of 4.35 per cent. The y axis runs 68 to 76. That puts the bars at 40.0 and 78.8 per cent of the plot height, a ratio of 1.97, so the taller bar is drawn 96.9 per cent taller than the shorter one. The picture overstates the difference by 22.2 times. On a 0 to 100 axis the same bar is 4.4 per cent taller.
Figure 5 claims the gap between arms narrows across six visits, and plots both arms as grouped bars. To see the claim the reader performs six mental subtractions, remembers all six answers, then compares them. Plotting the difference itself as one series replaces that with a single line whose slope is the claim. Separately, both arms are distinguished only by fill colour, which fails the Level A criterion that colour is not used as the only visual means of conveying information.
How it works
Name the claim
One sentence per figure stating what it argues, taken from the text rather than the caption.
Find the comparison
Which visual comparison the claim requires, and how hard the current figure makes it.
Check the encoding
Axis range, aspect ratio, category order, colour dependence and contrast, all against the numbers underneath.
Specify the fix
The revised figure, drawn, plus the caption that states the claim it now supports.
What you get
- The claim each figure is meant to support, written out before the figure is judged
- Whether that comparison is the easiest read on the chart or the hardest available
- Axis truncation quantified as the factor by which the picture overstates the difference
- The mental arithmetic a reader has to perform, counted, and the plot that removes it
- Series distinguished by more than colour, at the contrast ratio the standard requires
- A merge or cut recommendation when the venue caps the figure count
Common questions
A reviewer said a figure was unclear without saying why. What do I do?
Start from what that figure was meant to prove and check whether the reader can see it without doing arithmetic. Nearly every vague complaint about a figure resolves to the claimed comparison being available but not salient. The review names which comparison the reviewer was reaching for and redraws so that comparison is the first thing visible.
Is truncating a y axis always wrong?
No, and a rule against it would be worse than the problem. On a scale with no meaningful zero, or a series whose variation is small relative to its level, truncation is honest. What the review does is quantify it: the drawn ratio against the ratio in the data, the same way the sensitivity grid in a power note replaces a judgement with a number you can argue about.
Will it actually redraw the figures or just criticise them?
It draws them. The revision is generated from the numbers you supply, so the redesign is a chart you can look at rather than a description of one. Where a fix needs data you have not supplied, such as the individual observations behind a mean, it says which figure needs what and draws everything it can.
The journal caps me at six figures and I have nine.
Then the claim inventory does the cutting. A figure carrying no claim goes first, a figure whose claim a sentence can carry goes second, and two figures making the same comparison merge into panels. What survives is chosen on argument rather than on effort already spent, which is the part authors find hardest to do to their own work.
Does this cover accessibility, or is that separate?
It is part of the same check, because a series only a sighted reader with good colour vision can separate is a series the figure has not communicated. Graphical elements needed to understand a chart require a contrast ratio of at least three to one under the Web Content Accessibility Guidelines, and colour cannot be the only cue. Both get checked and both get fixed in the redraw.
What about figures that are not charts?
Diagrams, flow figures, images and photographic panels get the same first question and a different second one. Instead of asking which comparison the claim needs, it asks whether the labelling and the order of elements carry the argument, and whether the caption states something the figure shows. A schematic with no claim is decoration, and the review says so.
What comes back?
A Doc with one entry per figure: the claim, the comparison it requires, where the current version fails, the fix, and a caption that states the claim rather than describing the axes. Charts for every revision the supplied data supports. A Sheet holding the figure inventory with each claim, verdict and disposition for the cover letter.
Research Figure Review and Chart Redesign
Fill in the form and your workspace opens with the work already underway.