Those listening tests are must-views for anyone using 128kbit or lower for their encodings. I note there's not really any "high bitrate" comparisons.
BTW a quick note on reading those graphs: treat the result bar for each codec as an "error bar" representing the possible range of values; that range has a high probability of containing the result of a listening test with any representative sample. In other words, the fact that this group of listeners picked a particular rating for how good a codec is means less than the fact that any particular group of listeners would be likely to pick a value within some specific range. So if one codec's full range does not overlap another's at all, you can have confidence that those two codecs have distinctly different quality. If one codec is higher than another but their bars overlap, you CANNOT confidently say that the one is better than the other. Thus in the test with MPC, for example, MPC has the highest overall rating, but it should actually be viewed as "as good as" AAC and WMA overall due to the overlap in their bars. Note that these "overall" ratings don't mean some codecs won't win in particular tests; for example in that same listening test, there was one sample where WMA was rated very low, possibly indicating a case WMA doesn't (or didn't) handle well. Such cases can be found for many lossy formats.