Why is the technique of sampling important in ensuring rigorous research results?
The technique of sampling requires that the researcher pick his or her subjects mostly at random, and that the sample size be sufficiently large so that the subjects will be roughly representative of the population as a whole. If a study uses subjects picked because of their proximity to the study, or because those subjects are already known to the researchers, the data produced from that study will be skewed. In other words, the findings will not be representative of the population group the researchers hope to study. Consequently, the study will have limited value.
For example, if I want to figure out the "political leanings of most Americans" and I decide to figure that out by asking that question of everyone I personally know, the results of my study will be almost worthless. The people I know are not a random sample because they are filtered, so to speak, by the fact that I already know them, and vast amounts of research show that most people associate with others who share their political beliefs. In other words, my study will come back looking like an affirmation or mirror of what my own political leanings are. It will be a study of the political leanings of my friends.
A similar problem exists if I ask the same question of five hundred people who live in a one mile radius of me: the sampling is geographically biased and not at all representative of the country as a whole. That is why social scientists devised the notion of random sampling, in which a reasonably large group of people, from the population that is supposed to be studied, is picked at random (to avoid unconscious bias).
To find out the political leanings of most Americans, our hypothetical study would have to contact people in all fifty states, making sure that the people contacted were in both cities and rural areas (again, so that the population of the study is representative of the country as a whole). The sampling method would also require that the study does not "oversample" from the East Coast or the West Coast, or from any other region, relative to the population as a whole.
Additionally, researchers would want to avoid talking primarily to one age group or socio-economic group, because such an over-representation will not be a good proxy for the population as a whole. The researchers will also want to make sure that their sample is roughly split between men and women, for the same reason as mentioned above.
Essentially, the technique of sampling is designed to ensure that the group being studied is actually representative of the larger group that the researchers are hoping to understand. Accomplishing this task is much harder than it might first appear.