In trying to understand how the polls were so wrong in the New Hampshire primary, some people have considered the Bradley effect. Many of those looking at this only to dismiss it seem to have an incorrect understanding of exactly what the Bradley effect is, for example here.
According to the Pew Research Center, the Bradley effect is defined as:
White candidates in most of these races generally did better on Election Day than they were doing in the polls, while their black opponents tended to end up with about the same level of support as the polls indicated they had.
Using the standard estimate from Pollster.com and the rolling average from Real Clear Politics, I made a chart of how the polls compared to the actual results for Obama and Clinton.
Click for a larger, clearer view
You can judge for yourself whether or not that matches the description. This doesn't mean that it was the Bradley effect, but it clearly does not rule it out either.
Added: The Guardian gets a prize for actually understanding the Bradley effect correctly:
The Bradley effect suggests that white voters disguise their intentions from pollsters when a black candidate is in an election. In the case of New Hampshire the unusually high proportion of "undecided" voters - as much as 20% - may have concealed some of those hostile to Obama.
"I think it's very naive to dismiss the racial factors in this," said Larry Sabato, professor of politics at the University of Virginia.
Later: The president of the Pew Research Center writes in the New York Times that he believes it was the Bradley effect.