How to Study Figures and Tables in Research Papers

A figure is not decoration around a research paper's “real” text. In many papers, the table or graph contains the evidence that the abstract compresses into one sentence. If you only read the conclusion, you can miss the comparison, uncertainty, subgroup, or limitation that changes what the result means.

The goal is not to memorize every axis label. It is to reconstruct the chain from question → data → pattern → claim → limit, then test whether you can explain that chain without looking.

Start with one figure, not the whole paper

Pick the figure or table that seems most important to the paper's main claim. The George Washington University guide to scientific papers notes that figures and tables present selected data and should be understandable through their title and legend. That makes the caption a useful first checkpoint.

Read in this order:

1. Figure or table number
2. Caption or title
3. Axes, column headers, units, legend, and symbols
4. The methods needed to understand what was measured
5. The results paragraph that refers to the figure
6. The authors' interpretation and stated limitations

Do not start by asking “What does the graph prove?” A single figure normally supports a narrower claim than the whole paper.

Use a Figure Fact card

The Brandeis guide to interpreting scientific data recommends a structured “Figure Fact” approach. Use this reusable version in your notes:

Field	Your note
Paper and figure	Citation + figure/table number
Question	What specific question is this panel testing?
Data source	Who or what was measured?
Comparison	Which groups, conditions, or time points are compared?
Variables and units	What do the axes, rows, and columns measure?
Pattern	What changes, stays similar, or varies?
Uncertainty	What do error bars, intervals, sample sizes, or missing values show?
Author's claim	What does the results text say this pattern means?
Limit	What can this figure not establish?
Exam prompt	What question could require you to interpret or apply it?

If you cannot fill one field, mark it as a gap. Do not invent the missing method or label.

A teaching example

This is a fictional teaching example, not a real study or a Notoo product test.

Imagine Figure 2 compares the average quiz score of two student groups. Group A used a rereading worksheet. Group B used practice questions. The bars are 68 and 75, but the caption says the sample contained only 18 students and the error bars overlap.

A weak note says: “Practice questions are better.”

A stronger Figure Fact card says:

• Question: Did the two study conditions produce different quiz scores in this small class sample?
• Pattern: The plotted mean is higher for Group B.
• Uncertainty: The error bars overlap, and the sample is small.
• Claim: The figure shows an observed difference in this sample.
• Limit: The image alone does not establish a reliable effect, explain causation, or show whether the difference is statistically supported.

The important habit is separating what you can see from what you infer.

Decode common display types

Bar charts

Check what each bar summarizes, where the baseline starts, and whether the display shows individual observations or only an average. Look for error bars and sample size. A taller bar is a visual difference, not automatically a meaningful or causal effect.

Line graphs

Identify what changes along the horizontal axis and whether the same participants or samples were measured repeatedly. Check for unequal intervals, missing time points, and whether two lines use the same scale.

Scatterplots

Describe direction, shape, spread, and unusual points before interpreting a fitted line. Association does not by itself show that one variable caused the other.

Tables

Read the title, row population, column definitions, units, footnotes, and missing-value symbols. Compare the denominator as well as the numerator. A percentage can look different when the group sizes differ.

Multi-panel figures

Treat panels A, B, and C as linked but separate evidence. Write one sentence per panel, then one sentence explaining why the authors placed them together.

Check the figure against the paper

Use a three-way check:

• Caption check: Does your explanation match the variable, group, and condition named in the caption?
• Methods check: Do you know how the measurement was made and what the units mean?
• Results check: Does the paper claim more, less, or something different from your first visual impression?

Then find the limitation section. A figure can be correctly read and still support only a narrow conclusion because of sample selection, measurement choices, confounding, or study design.

Use AI as an organizer, not the final reader

Notoo's PDF-to-notes workflow can create a first-pass overview of a research paper's question, methods, and findings. Its current guidance also warns that scans and image-heavy pages can be harder to interpret and tells students to check figures, tables, equations, citations, and small text in the original.

A safe AI prompt is:

From the readable text of this paper, identify every place the authors refer to Figure [number]. Separate the stated method, observed result, author interpretation, and limitation. Do not infer values from the image. Mark anything that requires me to inspect the original figure.

This prompt uses AI to organize surrounding text. You still read the axes, legend, caption, footnotes, and visual data yourself.

Turn a figure into exam practice

After completing the card, close the paper and answer four prompts:

1. What question does the figure address?
2. What is the most important visible pattern?
3. What evidence would weaken your interpretation?
4. How would the conclusion change if one axis, group, or control changed?

For a quantitative course, redraw a simplified version with blank axes and labels. For a methods course, name the design and the strongest claim it permits. For a humanities or social-science course, explain how the table operationalizes a concept and what the categories exclude.

Advice for international students

Keep the paper's technical term next to a clear-English explanation. Build a small glossary for verbs such as increased, predicted, associated, mediated, and caused because they imply different relationships.

Do not translate axis labels without preserving the original wording and units. If a caption is dense, rewrite it as short clauses: who, condition, measurement, unit, and comparison. Then check your rewrite against the source.

A final verification checklist

Before you study from your Figure Fact card, confirm that you have:

• the correct figure or table number;
• every axis, column, unit, legend symbol, and footnote;
• the sample or data source;
• the comparison and control condition;
• uncertainty or missing-data information;
• a claim no stronger than the design supports;
• a clearly marked gap where the paper does not provide enough information;
• one closed-book question that requires interpretation.

Read evidence, not just summaries

Choose one figure from this week's reading and complete the card before rereading the discussion. If you want a structured first pass through the surrounding PDF, create notes with Notoo, then verify every visual claim in the original paper.