Importance is strongest when a reader can see exactly where the evidence ends.
Use claim verbs deliberately
- Results show or estimate: “Error was 3.4 percentage points lower in this test set.”
- Interpretations suggest or support: “The pattern supports the hypothesis that added tactile input helps under low visibility.”
- Potential impacts may or could: “If the effect generalizes, the method could reduce failures in similar low-cost grippers.”
- Demonstrated impacts did: “In a six-week classroom deployment, setup time fell by 18 minutes per session.”
Avoid “proves,” “solves,” “revolutionizes,” and “will save” unless the study design and real-world evidence genuinely support those words.
Replace scale inflation with specificity
Avoid: “This will revolutionize medicine.”
Prefer: “In this retrospective dataset, the image classifier reduced false negatives from 12% to 9% relative to logistic regression. The result needs external and prospective validation before clinical use.”
Avoid: “Our biodegradable material will eliminate plastic pollution.”
Prefer: “The prototype lost 31% of its mass under the tested composting condition while retaining the required strength for seven days. Cost, toxicity, manufacturing scale, and behavior outside this condition were not tested.”
Numbers do not make an overclaim safe. The population, comparison, time period, and conditions must travel with the number.
Put limitations next to the claim
Useful limitations identify how the result could change:
- sample size or selection;
- measurement precision;
- untested populations or environments;
- confounding variables;
- simulated rather than real operation;
- short observation period;
- baseline quality;
- model or material assumptions.
Do not write “more research is needed” alone. State the next test that addresses the limitation.
Primary sources
Go deeper
Use these first-party references to check rules, study the method further, or adapt this guide to your field.