The conversation is certainly moving fast and furiously. I would like to say a word about "correlation," though. Yesterday, mhaze pointed out that the time series had a weak correlation coefficient in the range of .22 to .44. This caused some controversy, culminating in TrueSkeptic saying "Would you consider this graph, showing long-term temps against CO2 displays strong correlation? Would you demand a numerical value?"
I reconstructed a similar graph from the Vostok data at can tell you that the correlation coefficient I came up with was .214. While I'm sure there are some flaws in how I put things together (I was going pretty fast), I think it was pretty close to what TrueSkeptic posted.
I think there is some confusion over what it means to calculate a correlation coefficient. If you have a "first A happens, then B happens" situation, your correlation may be low, even if A causes B. The correlation coefficient looks at the distances of x and y from their respective means at the same point in time. If there is a delay from the time A happens to the time B happens, this can lower your value.
For instance, imagine I have an alcoholic uncle who is always broke. Periodically, I give him money and he then drinks at a steady rate until the money is gone. If you plot the amount of money and his drunkness on a graph, you will see a peak in his cash, followed by a peak in his drunkness. I made such a graph and calculated a correlation coefficient of .108, a very weak association!
I would love to post links to my pretty pictures, but, being a new user, I can't.
But to summarize, a correlation coefficient around .22 between CO2 concentrations and temperature just tells us that these two variables tend to increase or decrease together, although the association is weak (explaining about 5% of the variation).
Hope this helps clear things up.