Ran across you (again) via your comments to Yulsman’s post – you are right on.

e.g.
http://www.cejournal.net/?p=1337#comment-1686

A being the average temperature from 1980 to 1999, plotted against the year 1990, and B being the temperature in year X.

However this gives me 0.11 degrees per decade for GISS for 2008, which is much lower than what you show, so the calculation must be something different?

[Yes, I agree that I should have made the calculation method clearer. Each trend is the difference between two averages (not a single point). So to take the GISS 2008 example, I take the difference in average between 2000-8 and 1980-99 baseline, and divide by the elapsed time centre-to-centre. The centre of the projection period is 2004 and that of the baseline 1989.5, for an effective elapsed time of 14.5 years. In this case, the difference is 0.26 deg/1.45 decade = 0.18 deg. decade.]