Its all over the place. On the extreme end of things, Huckabee is at 23% in one and 36% in the next and Paul is at 10% and then 1%. How can this possibly be? Can a persons support jump 13% practically overnight? I would posit that it isn't possible without some fairly major event taking place, perhaps a disaster or attack or a candidate dropping out of the race. Therefor so-called "scientific" polling is dead or dying. It is no longer a reliable method of gauging public opinion and will need to be reworked in short order to remain valid.
Graphic is from RealClearPolitics.com specifically from the Iowa Republican race.