Drafting results without overclaiming
Lesson 2 of 6 · Manuscripts and reviewer responses
1 · Learn the move · Constraints + negative instructions
The results section reports; the discussion argues. Keeping them apart is mostly a matter of banned verbs and mandatory attachments. Banned where unearned: prove, confirm, demonstrate, establish, show conclusively. Required on every claim: the n, the test, and the effect size with its interval - 'reduced by 40% (95% CI 28-51, n = 6, paired t-test)' survives review; 'dramatically reduced' does not. 'Consistent with' and 'we observed' are not weakness, they are accuracy. And one hard rule from the wet side: never a number from an image - a blot or gel picture never yields a quantity in your text; numbers come from the quantified data file. The model drafts and checks against these constraints; concluding is the discussion's job, and only partly even there.
Draft results paragraphs from my analysis output: [paste numbers with n, tests, intervals]. Constraints: (1) banned verbs unless I explicitly authorize: prove, confirm, demonstrate, establish, (2) every claim carries n, test, and effect size with interval, (3) 'consistent with' over causal verbs wherever the design is not causal, (4) no quantity that traces to an image - quantified exports only; anything without a number in my paste is [NEEDS: value from data file], (5) report, do not conclude.
2 · Your turn. You write the prompt
Your first draft of the results says the knockdown 'demonstrates that the pathway controls migration' - written at 1 a.m. after the third replicate finally cooperated. Your PI circles sentences like that. The actual numbers are solid enough not to need the inflation, and the section is due for group meeting Thursday.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.