Clustering The US Senate

I created a hierarchical clustering graph based on the roll call votes for the 111th session of the US Senate using the same method as my last post. I have a few observations about this chart. The first is that Nelson(D-NE) clusters with the Republican half when cuttree=2. The second observation is that Reid (D-NV) closest associate is Sanders (I-VT). I had assumed that Reid was a middle-of-the-road voter, and that Sanders, a socialist, had more extreme votes. Perhaps I should graph their rhetorical distance rather than roll call voting pattern distance.
It is not a surprise that Bayh(D-IN) McCaskill(D-MO) break off into their own cluster when cuttree=4. I am more surprised to see Feingold(D-WI) and Specter(D-PA) to form their own cluster with cuttree=8.
Again, I have added the source code for auto-downloading and parsing the data, and drawing the chart. Group membership is here. Click image for a larger version.