Well, there are some seldom-discussed limits. The one that concerns me today is the information limit: how likely is it that people will assimilate statistically unlikely statements and act on them?
It is possible that educational level will affect this. In this experiment, the average educational level, and level of economic success, is above the median in society. We're drawing from middle class, upper middle class, and upper class neighborhoods. We present information both by behavioral example, and visually, with a T-shirt or hat containing a slogan summing up the information.
In order to have maximum information content, the phrase should be unlikely, but, because I want to test its adoption and not get other people in trouble, in strict accord with the state law: in California, in places with no sidewalk (such as bike paths) pedestrians are expected to walk on the side facing oncoming traffic (the left), so that they see what is developing and coming toward them. It's a safety issue. But culturally, people follow the rules for vehicles, and stay to the right. The phrase: "Bike Right, Walk Left" -- short, emphasized by graphics, and citing the California Vehicle Code 21956.
What was the result of two weeks' trial on the bike path on the beach in Santa Monica? Perhaps two people read the sign, and one followed it. Observation over the period encountered the same people many times, none of whom bothered to read it, several of whom were unhappy that I was on the "wrong" side (and they didn't read it either.)
Conclusion: people do not assimilate information that they don't expect, or don't already know, from casual reading. Since casual reading is most of the scanning that goes on in weblogs, newspapers, and information sites, I expect that the ability of the Internet to teach people new information they were not already seeking is quite limited.
Conclusion slightly altered due to excellent comment. See below.