Bar Graph
A bar graph uses bars of different heights to compare measured values across separate categories or groups in an experiment.
Definition
A bar graph uses tall and short bars to compare different groups, with taller bars meaning bigger numbers. It compares the dependent variable across discrete categories of the independent variable, one bar per category with height equal to the measured value (usually a mean of trials), bars that do not touch, and a y-axis that starts at zero. In lab reports each bar should carry an error bar showing the range or standard deviation so readers can judge whether differences between conditions exceed the variability within them; grouped bars compare two factors at once and stacked bars show composition. Because bar height encodes magnitude from zero, a truncated axis is misleading.
Example
After growing bean plants in three soils, one bar each for sand ($4\ \text{cm}$), clay ($6\ \text{cm}$), and potting soil ($9\ \text{cm}$) makes the potting-soil advantage obvious. Comparing how far a cart rolls on wood ($79\ \text{cm}$), felt ($62\ \text{cm}$), and sandpaper ($41\ \text{cm}$) at one ramp angle, three bars give the ranking at a glance, and since surfaces are categories no line should connect them. Mean seedling heights for fertilizer doses of $0$, $1$, $2$, and $4\ \text{g/L}$ plotted at $10.9$, $13.2$, $14.6$, and $14.8\ \text{cm}$ with error bars of about $\pm 1\ \text{cm}$ show the $0$ and $1\ \text{g/L}$ bars clearly separated while the $2$ and $4\ \text{g/L}$ bars overlap.
Key Insight
Use bars when the groups are separate categories like soil types or brands, and a line graph when the bottom axis is a number that changes smoothly. A bar graph that does not start at zero exaggerates differences, so check the y-axis before judging which bar is "much" taller. Two bars of different heights are not automatically different results; if the error bars overlap substantially, the honest reading is that the experiment could not separate the conditions.