# Reading scientific graphs

Source: https://onco.cc/terms/graph-reading/  
OnCo record `graph-reading` (Term). Data CC BY-NC 4.0, attribute "Data from OnCo (onco.cc)"; commercial use needs a licence.

## TL;DR

Read a graph with its caption and methods: identify who was measured, what changed, the outcome, the comparison and the uncertainty. A visible pattern does not supply those missing details.

## Summary

A scientific graph needs its caption and methods to identify the population, independent measurement unit, intervention, outcome, comparison and uncertainty. Axis scale, baseline and time origin can change what a visible difference means. Biological units and repeated technical measurements must remain distinct, and different error bars describe different quantities. A plotted association, growth reduction or late survival plateau requires its own interpretation before any claim about causality or patient benefit.

## Fields

- Kind: Term
- Last checked: 2026-11-10
- Also known as: graph reading; figure reading; five-pass graph reading

## Notes

- Read a graph as an answer to a question before interpreting its shape. Establish who or what contributes independent observations, what intervention was given and when, which outcome and units the axes show, what baseline or control supplies the comparison, and what uncertainty or missing information remains. Entry criteria, dose, measurement site, time origin and follow-up can change the answer. A whole product and one component are different interventions; cell activity, tumour size, recurrence and survival are different outcomes. A mouse-volume graph cannot become a patient-survival result because the plotted separation is large. This five-pass reading habit is a teaching synthesis, not a score certifying a study; an explicitly unreported independent sample count remains useful information.
- Biological replicates are independent units relevant to the question, such as donors or animals. Technical repeats measure the same material again: three tubes or ten wells from one donor still represent one donor. Twelve measurements from three donors therefore require an analysis that accounts for repeated measurements, rather than treating all twelve as independent people. Error bars must be defined. Standard deviation describes observation variability, standard error concerns precision of an estimated mean, and a confidence interval follows a specified procedure and assumptions. On logarithmic axes equal distances represent equal ratios; on linear axes they represent equal differences. A truncated axis may emphasise a small difference, so read the numbers as well as the shape. Missing observations are not measured zeros, and a complete caption does not substitute for sound methods.
- In a fictional animal experiment, four independent animals contribute to each group. Both groups start with average tumour volume 100 cubic millimetres; at day 28, control averages 1,000 and treated 300. The treated average is 70% below control but three times its starting size: reduced growth relative to control, rather than shrinkage below baseline. Twenty wells from one animal would still represent one independent animal. For a Kaplan-Meier curve, the same reading habit requires event definition, starting clock, censoring and numbers still at risk. A flat late tail based on few observations cannot support a precise late estimate simply because it looks stable.

## Sources

- SuperPlots: Communicating reproducibility and variability in cell biology.: https://doi.org/10.1083/jcb.202001064
- NIST: graphical techniques: https://www.itl.nist.gov/div898/handbook/eda/section3/eda33.htm
- CONSORT 2025: reporting randomized trials: https://www.bmj.com/content/389/bmj-2024-081123

## Connected records

- terms: [Confidence interval](https://onco.cc/terms/confidence-interval/), [Hazard ratio (HR)](https://onco.cc/terms/hazard-ratio/), [Kaplan-Meier curve, censoring and proportional hazards](https://onco.cc/terms/kaplan-meier-curve/)
- fronts: [Drug Discovery Platforms](https://onco.cc/fronts/drug-discovery/)

---
JSON: https://onco.cc/api/v1/entities/graph-reading.json