This is a cool topic. I need to revise myself yet again (if this were the Everything Else forum I would make a joke about a particular candidate I supported last year) and say that Doswell and Johns do NOT suggest statistical significance to the cluster in their plot. As Glen wrote, the plot serves their paper only to illustrate the range of CAPE and SRH under which strong tornadoes occured in the cases they studied. However, using the same data in another plot, they draw a line that suggests the lower limit "of combined CAPE / low level shear values that would support development of strong and violent tornadoes." [Broken External Image]:http://www.cycloneroad.com/images/chase2004/figure18.JPG Here they are not simply documenting the scarcity of low CAPE / low shear environments but proposing that those environments are unsupportive. They do not, however, go so far as what I first proposed (admittedly without having looked at the data or even having read their paper!), which was that the cluster suggested a significance to that intermediate range beyond the frequency of its occurence. Ulitmately, this becomes more of a Stats novelty since we all know that two variables don't get you very far. Everything we've learned since 1992 demonstrates that supportive tornado environments come in all shapes and sizes and that the lowest levels are far more important, for both stability and shear, than was imagined fifteen years ago. Great Stormtrack topic!