When your text and tables disagree: checking claims against results
How sentences drift away from the results they describe, and how to check each claim against its table.
The results section is usually written twice: once when the first numbers come in, and again, partly, when they change. The tables get regenerated from the latest run. The sentences describing them get edited by hand, and some get missed. The result is a paper whose prose says one thing and whose tables say another.
Reviewers check exactly this. They read the claim, look at the table, and if the two disagree they trust neither.
Four patterns we see again and again
Universal claims the table doesn't support
“All”, “every”, “consistently”: one exception in the table is enough to make the sentence false.
From a computer-vision paper
The text says the method “significantly outperforms” the baseline “in every category”. In the table it cites, the baseline wins one category, 0.91 to 0.85.
Fix: say what the table shows: better in 9 of 10 categories.
Old numbers next to a new table
From a genomics paper
The results text says two language models reach 0.74 and beat an expert baseline at 0.72. In the table on the same page, the baseline is 0.74 and the two models score 0.78 and 0.79.
Fix: update the sentence to the numbers in the table.
Method settings the results didn't use
The method section describes the plan; the experiments section describes what happened. They are not always the same.
From a computer-vision paper
The method section says a key parameter “is set to t = 0 by default”. The ablation five pages later finds the best performance at t = 200, and the main results use t = 200.
Fix: state the value the reported results actually use.
Captions that describe something else
Captions are written once and rarely revisited when a figure is swapped or relabelled.
From an astrophysics paper
A figure caption says inset (a) shows “HH 201”. The label on the figure and the text that discusses it both say HH205, a different object.
Fix: change the caption to HH205.
How to check claims against results
- Find every sentence that cites a table or figure (“Table 2 shows”, “as seen in Fig. 4”), and every sentence that states a result without citing anything.
- Read each one with its table open beside it. Check the numbers, the direction of every comparison, and every case the sentence covers.
- Treat strong words as claims to verify. “Every” needs every row. “Significantly” needs a statistical test, not just a bigger number. “Consistently” needs no exceptions.
- Compare the method section with the experiments. For every parameter given a value in the method, check that the results used that value or say why not.
- Read each caption against its figure: labels, panel letters, legend entries and units.
- Update the prose last. Once the final tables exist, re-read every sentence that describes them, even ones you didn't touch.
Why it's hard to see in your own paper
You know what the results are supposed to say, so you read the sentence and agree with it without looking at the table. A co-author new to the section, or a tool that compares every claim with the table it cites, doesn't have that problem. DeskReady does the second: it reads the typeset pages, including tables and figures, and reports each disagreement with both places highlighted.
More in the pre-submission checklist and in numbers that don't add up.
DeskReady runs these checks on your PDF in a few minutes and shows every finding highlighted on your own page. Your first check is free and needs no account.
Check your paper