The principal components analysis of the pilot correlation matrix and the correlations between the principal components and the experience measures is consistent with the findings of the demographic analysis. That is, the first principal component seems to encode 'discomfort' and it is negatively correlated with experience in automated airplanes. Since we know that the first principal component of the attitude probe correlation matrix splits the probes into positive and negative probes, it is not surprising that most of the variance in the pilot correlations is accounted for by a similar dimension.
The emergence of the ``magic'' dimension as the second principal component of the pilot correlation matrix was a surprise. Of the many aspects of automation that might have been made salient by the response patterns of the pilots, this was not expected. As mentioned above, the idea that automation is magic appears often in pilot explanation of automation. It is interesting, however, that this component also has a negative correlation with the automated experience measures. It may be that the ``magic'' interpretation is a conceptualization that serves pilots early in their experience with automated airplanes but is replaced later by a model that contains more insight into the internal logic of the system.
It came as something of a surprise that the third principal component which we had called 'by the book' did not correlate with the military/civilian training demographic variables. This is probably simply a case where our stereotypes lead us to expect a relationship that does not exist. Pilots with airforce training were higher on PC2(magic) than pilots with no military training were. It is difficult to interpret this, however, because pilots without military training have, on average, more automated experience than pilots with airforce training and PC2 is negatively correlated with automated experience.
It is interesting that the measure of total flight experience is not significantly correlated with any of the principal components of the pilot correlation matrix. This suggests that attitudes toward automation are formed in the use of automation and may not depend much on the flying experience that preceded the experience of automation.