Foundationsmethods

Graphing Biology Data: Choosing and Reading the Right Plot

Overview

A good graph reveals the biology; a bad one hides it. This lesson teaches which plot suits which data — bar charts for categories, line graphs for continuous trends, scatter plots for relationships, box plots for distributions — plus the honest-graph rules: labeled axes, truthful scales, and showing variability.

Learning objectives

  • Choose the appropriate graph type for a data type.
  • Explain when to use bar, line, scatter, and box plots.
  • Label axes correctly with units.
  • Identify misleading graph practices.
  • Interpret error bars and trend lines.

Match the plot to the question

Different data need different pictures. Comparing means across categories (drug doses, genotypes)? Bar chart with error bars. Showing change over a continuous variable (growth over time, rate vs temperature)? Line graph. Exploring a relationship between two measurements (gene expression vs growth)? Scatter plot. Comparing distributions (control vs treated spreads)? Box plot. The question picks the plot — not habit, not software defaults.

The four workhorses

Bar chart: categories. Line graph: continuous trends. Scatter plot: relationships. Box plot: distributions. Learn these four and you cover most biology.

Anatomy of an honest graph

Every graph needs: labeled axes with units, a descriptive title or caption, a legend if multiple series, appropriate scale, and visible variability (error bars, points, or boxes). The independent variable goes on the x-axis, the dependent on the y-axis — the cause-effect orientation readers expect. Captions should let the figure stand alone: what was measured, in what units, with what n.

Scales that lie

The fastest way to mislead is the axis: truncated y-axes exaggerate tiny differences (a bar chart starting at 90 makes 92 vs 98 look doubled), while stretched axes hide real ones. Three-dimensional effects, area-as-value distortions, and cherry-picked ranges all manipulate perception. Rule: bar charts start at zero (bars encode value by length); line graphs may zoom only if the zoom is explicit and justified.

Truncated bar charts

Bars encode quantity by length from zero — starting the axis at 90 turns a 6% difference into a visual doubling. It's the commonest graphical lie in science news.

Showing the noise

A mean without variability is a claim without evidence. Error bars should state what they are — standard deviation (spread of data), standard error (precision of the mean), or confidence interval — because each means something different. Better still: show the data points themselves (beeswarm or dot plots) so readers see distributions, outliers, and sample size directly. Transparency beats polish.

Reading graphs critically

Interrogate any figure: do the axes start honestly? Are error bars defined? Does the trend line reflect the data or wishful thinking? Is n stated? Would the conclusion survive a different scale? Graphs are arguments in visual form — read them with the same skepticism as text, and you'll catch the exaggerations that slip past words.

The dual-coding bonus

Well-made graphs exploit dual coding — visual plus verbal processing — which is why a clear figure often teaches faster than a paragraph.

Key points

  • Choose plot type by data: bars for categories, lines for trends, scatter for relationships, boxes for distributions.
  • Label axes with units; independent variable on x, dependent on y.
  • Bar charts start at zero; truncated axes exaggerate differences.
  • Always show variability — and state what error bars represent.
  • Read graphs skeptically: check scales, n, and whether the visual matches the claim.

Common misconceptions

✗ Myth: Any graph type works as long as the data are plotted.

✓ Truth: Wrong plot types hide patterns — categories need bars, trends need lines, relationships need scatter. The choice shapes what readers can see.

✗ Myth: Error bars are just decoration.

✓ Truth: They quantify uncertainty, and SD vs SE vs CI mean different things — an undefined error bar is meaningless decoration.

Practice questions

Q1. To show how enzyme activity changes continuously with temperature, the best choice is a:
  1. bar chart
  2. line graph
  3. pie chart
  4. box plot
Answer: B.

Temperature is continuous, so a line graph reveals the trend — including the optimum and denaturation decline — that bars would chop into pieces.

Q2. A bar chart with the y-axis starting at 80 instead of 0 most likely:
  1. is standard practice
  2. visually exaggerates small differences between groups
  3. hides large differences
  4. shows error bars better
Answer: B.

Truncation breaks the length-equals-value encoding of bars, making modest differences look dramatic.

Q3. Showing individual data points on a graph is valuable because it reveals:
  1. nothing extra
  2. distributions, outliers, and true sample size hidden by summary bars
  3. only the mean
  4. the p-value
Answer: B.

Raw points expose skew, clusters, and outliers that means and error bars can conceal — transparency that summary statistics lack.

Click a question to reveal the answer and explanation.

Frequently asked questions

When should you use a box plot instead of a bar chart?

When the distribution matters, not just the mean — box plots show median, quartiles, range, and outliers, revealing skew and spread that a bar chart's single mean hides.

Why must bar charts start at zero?

Bars encode value through length from a zero baseline. Truncating the axis visually exaggerates small differences, misleading readers about the effect's true size.

What is the difference between SD and SE error bars?

Standard deviation describes the spread of the data points; standard error describes the precision of the mean estimate. SE bars are always smaller — confusing them overstates certainty.

What should a figure caption include?

What was measured, in what units, the experimental groups, the sample size (n), what error bars represent, and the takeaway — enough for the figure to stand alone.