Bias introduced by non-random selection of observations, such that the sample is not representative of the underlying population.

The Wikipedia page lists different flavors and examples of this bias in action.