The theory is actually kind of complex -- there are a lot of publications in queue theory about essentially one question (what is the fastest way to board a one-aisle plane?).

See a graphic example of the way the studies typically progress at http://leeds-faculty.colorado.edu/va...g/boarding.htm , or read one exemplary version of such a study at http://arxiv.org/abs/0802.0733 .

Simulations often show that doing only back-to-front is among the worst options. However, I think the method here approximates what, at the link above, is illustrated graphically as "reverse pyramid" (outside-in combined with back-to-front), which is still a pretty good choice and may be a better one -- I guess we'll see what the data show!