Currently it is not possible to draw random samples from the general Internet population because there is no high quality sampling frame available. As a result most access panels consist of volunteer-opt-in samples. This can seriously hamper the results of surveys based on these samples. When under coverage of certain groups results in non-random missingness (non-MAR) the validity of the results could be seriously corrupted and selection bias will become apparent. It is not yet known, how to deal with selection bias in non- random samples, but since access panels are here to stay and are used more and more in research for policies and practices scientists have the obligation to study the phenomenon and search for solutions that will prevent against or correct for selection bias in access panels. Solutions can be methodological, i.e. aiming at the prevention of selection bias or statistical, i.e. aiming at adjustment techniques like weighting. This presentation will give a short overview of current streams of thinking on this problem, as an introduction to the presenters of the session on selection bias in panel research.