But what is interesting is not anyones definition of pause or hiatus or whatever. It's the comparison between predictions of climate models and real temperatures. And the models are really, really bad at predicting reality:
http://www.drroyspencer.com/2014/02/95-of-climate-models-agr...
If this trend continues, real temperatures will fall below the error range of all models.
Roy picks a baseline of 1979-83, which is unusual (I've never seen any other anomaly graphs plotted using that baseline, and it's especially strange since the graph only begins in 1983). The reason he does that is that during those years there is a brief but significant deviation between the model runs and the surface temperature data. By picking that baseline the effect is to shift the model results upwards. Hopefully these two images will make it clear what has been done:
This is the full data for GISTemp, UAH and a mean of the CMIP5 model runs (which I believe is the dotted black line on Roy's graph) based on a longer 1981-2010 baseline. Also note the model runs actually started in 1979 not 1983:
http://i.imgur.com/c2adVtw.jpg
As you can see, there is a brief but strong spike in the temperature data above the model mean (the model takes a dip as well increasing the effect). Using a narrow baseline of 79-83 gets you a graph that looks like this:
http://i.imgur.com/Ys1ZMZF.jpg
With the data starting at 1979 it's a bit more obvious what has been done. Roy cropped his graph at 1983 so that it looks like the models start in 1983. Then used a baseline that shifted the model mean upwards to align with the 1983 temperature so that is looks like the model runs began with a starting point 1983 with the 1983 surface temperature and then immediately diverged. That is very misleading.
If you are wondering why some climate scientists don't take the skeptics seriously, then this kind of manipulation is one of the reasons why. Roy is supposedly one of the reputable skeptics, being an actual scientist.
Graphs taken from: http://blog.hotwhopper.com/2014/02/roy-spencers-latest-decei...
The fact that the CMIP5 mean and UAH have nearly the same anomaly value in 1983 when using a 1979-1983 baseline is just coincidental. They diverge quite a bit in several of the other years in the 1979-83 period. Roy crops his graph to hide that and make it look like they begin together. So he hasn't even done what you claim (aligned the trends to match at the very beginning of the satellite record) because his method puts the CMIP5 mean anomaly about 0.2 of a degree above UAH at the beginning of the satellite record. Plotting all the available data would show that. Hiding data just deceives the reader.
If he really wanted to just "match at the very beginning of the satellite record" as you suggest, why not just use a baseline of 1979? I suspect the answer to that is that it wouldn't support his argument to present the data that way.
Your idea of aligning to the beginning of the satellite record is ill conceived anyway. That would only make sense if GCMs took the temperature data at that time as their starting point. They don't.