Some statistical comments re. "average" penis size(s) & size distributions
A fatal problem with the vast majority of "surveys" of penis size one finds on-line and in the popular literature (magazines, e-zines, etc.) is that they are not based on random samples of the male population. Rather, they are based on self-selected samples.
In the case of the Kinsey reports, for example, his measurements of penis size and other determinations were based on people, in various subgroups of the population, who voluntarily stepped forward at his request for his measurements and interviews. Thus, his samples were self-selected by definition. Kinsey's books were criticized for that reason. He was devastated by this criticism that his statistical methods were invalid (or only partially valid), because his books were his life's work (after he became interested in human sexuality early in his scientific career). [Despite this criticism, Kinsey's work yielded fundamental advances in knowledge of human sexuality, and American's... as well as others, worldwide... are indebted to him for pressing on and carrying out his studies despite howls of outrage from "moralists."]
In any event, one should suspect (in my opinion) that the determinations of "average" penis size one finds online and elsewhere give an overly large "average," because men who are proud of their somewhat or very large penises are more likely to step forward and have them measured than men who feel embarrassed or even ashamed because their penises are smaller than average. This effect would yield "average" penis sizes that are biased in the over-large direction. Also, it's well-known that men who measure their own penises (then submit the results to, for instance, an online survey) conclude their penises are larger than when the penises of the population sample are measured by a clinician. This too would bias the "average" penis sizes one finds online and elsewhere in the over-large direction.
[Those interested (if any) can find a good description of how to identify a population sample for sex research that's truly random near the beginning, probably in Chapt. 2, of:
Michael, R.T., J.H. Gagnon, E.O. Laumann, and G. Kolata, 1994, Sex in America: A Definitive Survey, Little, Brown, and Company, 300 pp. (Hardcover).
The methodology required for statistical validity, which is well-known to sociologists (as the authors point out), includes an in-depth personal interview with each individual in the random sample. In the case of penis size, a statistically-valid survey would have to include in-person measurement by a qualified clinician of each individual in the (not self-selected but rather suitably-selected) random sample.]