We address the issue of why different models may be getting different responses of the AO/NAO in climate change experiments. The results from part 1 (Rind et al., 2005) suggest that for substantive climate changes, the differences are likely to be found in the patterns of tropospheric climate change, rather than from the stratosphere. We assess the various tropospheric forcings through a variety of experiments. We first use extreme paleoclimate experiments (Ice Age, Paleocene) which feature large variations in the low level latitudinal temperature gradient; the results show that under these circumstances, changes in the eddy transport of sensible heat, and in situ high latitude forcing, dominate the AO response. We next test the effect of more modest SST temperature gradient changes in the current climate, and find a similar result with a model configuration that does not easily transport the low level temperature changes into the upper troposphere. We then reanalyze the results from different 2 × CO2 experiments with the GISS model and find that they can be understood by assessing: (1) the magnitude of tropical SST warming; (2) the translations of that warming into the upper troposphere; (3) the change in the extratropical low altitude temperature gradient; and (4) the change in the high latitude SST/sea ice response. We suggest that these features might explain the varying results among modeling groups, and that forecasts will not converge until these features do.