I find that often reading a scientist's dissertation will often tell you something about how they approach the rest of their research. Somewhere I actually have a paper copy of Feynman's Thesis under Wheeler, which covers exactly this material. He was clearly enamored of Lagrangian formulations quite early (I didn't know that it was due to a High School teacher).

You can find this actual thesis on-line at CERN: http://cds.cern.ch/record/101498/files/Thesis-1942-Feynman.p...

It's short at 74 book pages, and relatively readable.