A well-built scientific table presents data so clearly that a reader can grasp the key findings without reading the surrounding text. The structure follows consistent conventions across most scientific disciplines: a numbered title above the table, clearly labeled column headers, organized rows of data, and footnotes below for context. Getting each of these elements right is the difference between a table that communicates and one that confuses.
The Five Parts of a Scientific Table
Every scientific table has the same basic anatomy, regardless of field. Understanding these parts before you start building will save you from restructuring later.
The title (also called a caption) sits above the table, never below it. It should tell the reader what’s in the table at a high level, including any key groupings. “Table 2. Mean Blood Pressure by Treatment Group at 12 Weeks” is far more useful than “Table 2. Results.” In APA format, the table number appears in bold on its own line, with the title in italics on the next line.
The column headers label each vertical column of data. When you have related columns that fall under a broader category, you can use a “spanner” header that stretches across multiple columns to show that grouping. For example, a spanner labeled “Treatment Group” might sit above two sub-columns: “Drug A” and “Placebo.”
The stub column is the leftmost column. It lists the row categories, like variable names, time points, or participant characteristics. If your categories have a hierarchy (say, a main category with subcategories), use indentation to show the relationship rather than adding extra columns.
The body contains your actual data cells. This is where alignment, rounding, and formatting consistency matter most.
The footnotes go below the table. Use them for abbreviation definitions, explanations of symbols, statistical context, and source references. They keep the table body clean while giving readers everything they need to interpret the numbers.
Choosing What Goes in Rows and Columns
Place items you want the reader to compare side by side in adjacent columns. The eye moves across a row more naturally than it scans down, so comparisons across columns are easier to make than comparisons across rows. If your table compares outcomes between a treatment group and a control group, those two groups should be neighboring columns, not stacked in separate rows.
Think carefully about the order of your rows. Alphabetical order is the default when no other logic applies, but chronological order works better for time-based data, and ranking by magnitude (largest to smallest) helps when you want the reader to see patterns. For hierarchical categories, place the broadest category first and indent subcategories beneath it.
Aligning Numbers and Text
Alignment is one of the most common things people get wrong, and it directly affects readability. The rules are straightforward:
- Text columns: Left-align both the heading and the content.
- Number columns without decimals: Center the heading and right-justify the numbers so the ones, tens, and hundreds places line up visually.
- Number columns with decimals: Center the heading and align all values by the decimal point. This lets the reader instantly compare magnitudes without mentally adjusting for different digit counts.
- Stub column: Left-align the content, but center the heading above it.
Most word processors and spreadsheet tools support decimal alignment, though you may need to use a tab stop or a specialized table style to get it right. In LaTeX, the siunitx package handles decimal alignment automatically.
Keeping Numbers Consistent
Every value in a column should use the same number of decimal places. If you report one mean as 4.52, don’t report the next as 4.5 or 4.521. This consistency signals precision and makes the column easier to scan.
The number of decimal places you choose should reflect the precision of your measurement. Reporting a weight to four decimal places when your scale only measures to the nearest tenth is false precision. A good rule: match the decimal places to the least precise measurement in that column. For calculated values like averages, the result should carry the same number of significant figures as the input with the fewest significant figures. If you’re adding or subtracting values, limit your answer to the same number of decimal places as the term with the fewest.
When rounding, drop digits less than 5 without changing anything, and round up when the dropped digit is 5 or greater. So 2.9699 rounded to three significant figures becomes 2.97, and 4.00443 rounded to four significant figures stays at 4.004.
Lines, Borders, and Spacing
One of the most common mistakes in scientific tables is using too many lines. A gridded table with borders around every cell looks cluttered and actually makes data harder to read. The standard in scientific publishing is minimal lines.
You need a horizontal line at the top of the table, one beneath the column headers, and one at the bottom of the table. You can add a line above column spanners and above a summary row (like totals). That’s it. No vertical lines between columns. No grid. Use white space and consistent alignment to separate columns instead of borders.
Use the same font and size as the rest of your document. Leave a blank line between the table and any surrounding text so the table doesn’t feel cramped against your paragraphs.
Reporting Statistical Results
If your table includes statistical results, clarity matters more than convention. Report exact p-values (like p = 0.032) rather than just noting “p < 0.05” or marking cells with asterisks. While asterisks are still common, many statisticians and journals now discourage them because they reduce a continuous measure to a binary label of “significant” or “not significant.”
Whenever you include p-values, also report the numbers that give them context: means or medians, confidence intervals, and effect sizes. A p-value alone tells the reader very little about practical significance. State clearly whether your test was one-tailed or two-tailed, either in a footnote or in the table title.
For confidence intervals, a clean format is to place them in parentheses next to the point estimate, like “12.4 (10.1, 14.7).” Define the confidence level in a footnote if it’s anything other than 95%.
Writing Useful Footnotes
Footnotes serve three purposes: defining abbreviations, explaining symbols or formatting choices, and providing statistical context. Use superscript letters (a, b, c) rather than asterisks or numbers to mark footnotes, since asterisks can be confused with significance markers and numbers can be confused with data.
Define every abbreviation used in the table, even if you’ve already defined it in the text. Tables should be interpretable on their own. Common abbreviations in scientific tables include SD for standard deviation, CI for confidence interval, and n for sample size, but spell these out in the footnote the first time. If your table includes units of measurement, place them in the column header (e.g., “Height, cm”) rather than repeating them in every cell.
Handling Missing Data
Empty cells are ambiguous. A blank could mean zero, not applicable, not measured, or lost to follow-up. Pick a consistent convention and explain it in a footnote. Common approaches include using a dash to mean “not applicable,” “NA” for data not available, and an ellipsis for data not yet collected. Never leave a cell blank without explanation.
Making Tables Accessible
If your table will appear on a webpage or in a digital document, accessibility matters. Screen readers navigate tables cell by cell, reading the associated column and row headers aloud so visually impaired users can orient themselves. This only works if the table is built with proper structure.
In HTML, use <th> tags for header cells and <td> tags for data cells. Add a scope attribute to every header cell: scope="col" for column headers and scope="row" for row headers. This tells the screen reader which cells belong to which header. Without these attributes, a screen reader will just read a stream of disconnected numbers.
In Word or Google Docs, mark the header row as a repeating header row in the table properties. Avoid merging cells whenever possible, since merged cells often break screen reader navigation. If you use a spanner header, make sure each sub-column still has its own labeled header cell.
A Quick Construction Checklist
- Title above the table: Numbered, descriptive, and specific about what the data shows.
- Column headers: Short, with units included. Grouped under spanners where appropriate.
- Stub column: Left-aligned, with indentation for subcategories.
- Body data: Decimal-aligned for numbers, left-aligned for text. Consistent decimal places within each column.
- Minimal lines: Horizontal lines at the top, below headers, and at the bottom. No vertical lines, no cell borders.
- Footnotes below: Abbreviations defined, symbols explained, statistical methods noted.
- No blank cells: Every empty cell explained with a symbol defined in the footnotes.
Building a clean table takes longer than dumping numbers into a grid, but the payoff is significant. A reader who can understand your table without re-reading it three times is a reader who trusts your data.