A population consists of all the individuals of a single species that live in a limited geographic region, such as all the stickleback fish in a
lake. Biologists study populations because evolutionary change occurs over many generations at the population—not individual—level. In nature, a
population of any given species is typically too large to count. Thus, researchers estimate properties and infer conclusions about the population
as a whole by studying randomly selected subsets of individuals—or random samples.
Without adequate random sampling, measurements based on samples may depart somewhat from measurements that would have been taken from the entire
population. These differences can simply be due to chance. For example, a lake may have two types of stickleback, those with pelvic spines and
those without. If you only collect five individuals, they might all have pelvic spines, and you may conclude that all stickleback in this
population have pelvic spines. This type of error is called sampling error. Generally, larger sample sizes are less affected by chance. If you
collected 20 stickleback fish or more, you would have a much better chance of capturing the mix of stickleback types and thereby reducing sampling
error.
Another source of error is called bias, which occurs when samples are not taken quite at random. For example, if you only collect individuals from
one small area because it is the easiest spot to place traps, you may miss individuals with very different traits that live in different parts of
the lake.