People behave differently in polls than in actual elections. Polls have only signalling value, while votes have a definite outcome. When you vote you have to ‘play’ your hand, while when answering to the pollsters, you only signal what you ‘might’ do. That’s a crucial difference.
For example, many animals have signalling strategies they employ before fighting over a possible mate, and ‘cheating’ at the signalling stage — by signalling more strength than you actually are willing to employ in the fight — is a strategy that some employ successfully: the opponent might concede without a fight.
Scots have little to gain by telling pollsters they will vote ‘no’, even if their intention is to do exactly that on voting day: you may extract concessions, in this case from the rest of Britain. My view is simple: people know intuitively that this is a clever strategy, and are more likely to say ‘yes’ to independence to the pollsters than they are likely to vote ‘yes’ at the ballot box. Most polls you are reading don’t adjust for that bias, while they clearly should.
There are two other sources of bias, which will skew the result even more: (a) polls which are conducted by interviews have the effect that people feel their responses might be ‘judged’ by the interviewer, and they are less likely to go for the ‘unfashionable’ option than they are at the anonymous ballot box. Internet polls, which don’t have that problem, have led to a higher ‘no’ percentage on average than phone polls about the Scottish referendum. (b) polls are commissioned by either media outlets or one of the two sides of the divide, and all of them have an interest in making the result look more balanced than it actually is, in order to attract attention or to increase voter turnout.
So don’t be surprised if tomorrow’s vote on Scotland’s independence is a rather resounding ‘no’. While the pollsters say that about 52% +/- 3% will vote ‘no’, my prediction is that there is a 50:50 chance that the ‘no’ vote will be 60% or higher, and only the most minuscule probability that the ‘yes’ campaign will win the referendum.