Many statistical tests assume that data (or residuals) are sampled from a Gaussian
distribution. Normality tests are often used to test for evidence that the distribution
is not Gaussian, but these tests are less helpful than many scientists expect them to
be.
Many books (but not mine) and programs (but not GraphPad Prism yet) suggest
using a QQ normality plot to visually assess whether or not a distribution appears to
be sampled from a Gaussian distribution. The problem is that most books give very
complicated explanations of QQ plots, making it very hard to figure out what exactly
they mean. In fact, the basic idea of a QQ normality plot is pretty easy to understand.