scientific-skills/scientific-writing/references/figures_tables.md
Figures and tables are essential components of scientific papers, serving to display data patterns, summarize results, and provide evidence for conclusions. Effective visual displays enhance comprehension and can sustain reader interest while illustrating trends, patterns, and relationships not easily conveyed through text alone.
A recent Nature Cell Biology checklist (2025) emphasizes that creating clear and engaging scientific figures is crucial for communicating complex data with clarity, accessibility, and design excellence.
Example use cases:
Example use cases:
Can the information be conveyed in 1-2 sentences of text?
Each figure or table must stand alone without requiring the main text.
Essential elements:
Example of self-explanatory caption:
Figure 1. Mean systolic blood pressure (SBP) over 12 weeks in intervention and control groups.
Error bars represent standard error of the mean (SEM). Asterisks indicate significant
differences between groups at each time point (*p < 0.05, **p < 0.01, ***p < 0.001,
two-tailed t-tests). n = 48 per group. BP = blood pressure; SEM = standard error of mean.
Do not duplicate information between text, tables, and figures.
Bad practice:
"Mean age was 45.2 years in Group A and 47.8 years in Group B. Mean BMI was 26.3 in
Group A and 28.1 in Group B. Mean systolic blood pressure was 132 mmHg in Group A..."
[Also shown in Table 1]
Good practice:
"Baseline characteristics were similar between groups (Table 1), with no significant
differences in age, BMI, or blood pressure (all p > 0.15)."
[Details in Table 1]
Key principle: Text should highlight key findings from tables/figures, not repeat all data.
Maintain uniform formatting across all display items:
Example of inconsistency to avoid:
Follow the "one display item per 1000 words" guideline.
Typical manuscript:
Quality over quantity: A few well-designed, information-rich displays are better than many redundant or poorly designed ones.
Avoid cluttered or overly complex displays:
Best for:
Design guidelines:
Common mistakes:
Example applications:
Best for:
Design guidelines:
Common mistakes:
Example applications:
Best for:
Design guidelines:
Common mistakes:
Example applications:
Best for:
Design guidelines:
Common mistakes:
Example applications:
Best for:
Design guidelines:
Common mistakes:
Example applications:
Best for:
Design guidelines:
Common mistakes:
Example applications:
Best for:
Design guidelines:
Example applications:
Best for:
Design guidelines:
Basic anatomy:
Column headers:
Data presentation:
Statistical annotations:
Footnotes:
Table 1. Baseline Characteristics of Study Participants
Characteristic Intervention (n=50) Control (n=48) p-value
─────────────────────────────────────────────────────────────────────────
Age, years 45.3 ± 8.2 47.1 ± 9.1 0.28
Male sex, n (%) 28 (56) 25 (52) 0.71
BMI, kg/m² 26.3 ± 3.8 27.1 ± 4.2 0.32
Current smoker, n (%) 12 (24) 15 (31) 0.42
Systolic BP, mmHg 132 ± 15 134 ± 18 0.54
─────────────────────────────────────────────────────────────────────────
Data presented as mean ± SD or n (%). p-values from independent t-tests for
continuous variables and χ² tests for categorical variables. BMI = body mass
index; BP = blood pressure; SD = standard deviation.
For each comparison, report:
Choose the appropriate measure:
| Measure | Meaning | When to Use |
|---|---|---|
| SD (Standard Deviation) | Variability in the data | Showing data spread |
| SEM (Standard Error of Mean) | Precision of mean estimate | Showing measurement precision |
| 95% CI (Confidence Interval) | Range likely to contain true mean | Showing statistical significance |
Key rule: Always state which measure is shown.
Example caption:
"Error bars represent 95% confidence intervals."
NOT: "Error bars represent standard error."
Recommendation: 95% CI preferred because non-overlapping CIs indicate significant differences.
Common notation:
* p < 0.05
** p < 0.01
*** p < 0.001
n.s. or NS = not significant
Alternative: Show exact p-values in table or caption
Best practice: Define significance indicators in every figure caption or table footnote.
Recommendations:
Color-blind safe palettes:
Ensure readability:
Design for both media:
Vector formats (preferred for graphs and diagrams):
Raster formats (for photos and images):
Avoid:
Minimum standards:
Best practice: Create figures at final size and resolution.
Check journal requirements:
Recommendation: Design figures to fit single column when possible.
Allowed:
NOT allowed:
Ethical requirement: Report all image adjustments in Methods section.
Figures:
Tables:
Format:
"Results are shown in Figure 1."
"Participant characteristics are presented in Table 2."
"Multiple analyses confirmed this finding (Figures 3-5)."
NOT:
"Figure 1 below shows..." (avoid "above" or "below" - pagination may change)
"The figure shows..." (always use specific number)
For figures:
Figure 1. [One-sentence title]. [Additional description sentences providing context,
defining abbreviations, explaining panels, describing statistical tests, and noting
sample sizes].
For tables:
Table 1. [Descriptive Title]
[Table contents]
[Footnotes defining abbreviations, statistical methods, and providing additional context]
Essential information:
Example comprehensive caption:
Figure 3. Cognitive performance improves with treatment over 12 weeks. (A) Mean Mini-Mental
State Examination (MMSE) scores at baseline, 6 weeks, and 12 weeks for treatment (blue) and
placebo (gray) groups. (B) Individual participant trajectories for treatment group. Error bars
represent 95% confidence intervals. Asterisks indicate significant between-group differences
(*p < 0.05, **p < 0.01, ***p < 0.001; repeated measures ANOVA with Bonferroni correction).
n = 42 treatment, n = 40 placebo. MMSE scores range from 0-30, with higher scores indicating
better cognitive function.
Check journal guidelines for:
Prepare checklist:
Different journals have vastly different requirements for figures and tables. Before creating display items, always consult your target journal's author guidelines for specific requirements.
| Aspect | Variation by Journal | Example Journals |
|---|---|---|
| Number allowed | 4-10 display items for research articles | Nature (4-6), PLOS ONE (unlimited), Science (4-5) |
| File format | TIFF, EPS, PDF, AI, or specific formats | Nature (EPS/PDF for line art), Cell (TIFF preferred) |
| Resolution | 300-1000 dpi depending on type | JAMA (300-600 dpi), Nature (300+ dpi) |
| Color | RGB vs. CMYK | Print journals: CMYK; Online: RGB |
| Dimensions | Single vs. double column widths | Nature (89mm or 183mm), Science (specific templates) |
| Figure legends | Length limits, specific format | Some journals: 150 word max per legend |
| Table format | Editable vs. image | Most prefer editable tables, not images |
| Venue Type | Display Limit | Format | Resolution | Key Features |
|---|---|---|---|---|
| Nature/Science | 4-6 main | EPS/PDF/TIFF | 300+ dpi | Extended data allowed; multi-panel figures |
| Medical journals | 3-5 | TIFF/EPS | 300-600 dpi | CONSORT diagrams; conservative design |
| PLOS ONE | Unlimited | TIFF/EPS/PDF | 300+ dpi | Must work in grayscale |
| ML conferences | 4-6 in 8-page limit | PDF (vector preferred) | Print quality | Compact design; info-dense figures |
ML Conference Figure Requirements:
NeurIPS/ICML/ICLR:
Computer Vision (CVPR/ICCV/ECCV):
Key ML conference figure practices:
What reviewers check:
Common rejection reasons:
ML conference specific evaluation:
| Venue Type | Style | Example Features |
|---|---|---|
| Nature/Science | Concise | Brief; *P<0.05; minimal methods |
| Medical | Formal | Title case; 95% CIs; statistical tests spelled out |
| PLOS/BMC | Detailed | Complete sentences; all abbreviations defined |
| ML conferences | Technical | Architecture details; hyperparameters; dataset info |
ML conference caption example:
Figure 1. Architecture of proposed model. (a) Encoder with 12 transformer layers.
(b) Attention visualization. (c) Performance vs. baseline on ImageNet (error bars:
95% CI over 3 runs).
When changing venues:
Technical (all venues):
ML conferences additional:
For every figure:
For every table:
Overall: