I'd have to agree with the above poster. I am a physics grad student who picked up Fortran 77 earlier this week to finish some work for my professor. In fact I'm on lunch right now, with my fortran book in front of me and my inherited programs minimized on the screen.
I learned C, C++, Java in HS through classes, and then to a much greater degree on my own, although thinking back, Fortran would have done just as well. 95% of the programming I use for physics are script-like data processing. There is a few pages of mathematical work, which I convert into simple programs to process large amounts of data, or run simulations based on a random initial parameter space (Monte Carlo calcs).
The runtime is short, the math is generally rather simple, and most importantly, the turnover of a grad student is 3-4 years, and the turnover of undergrad interns is a fraction of that. A lot of the new students know the math and the physics, and may have taken some sort of programming class, or may not have. But its a lot easier passing down a Fortran program to a new student than C++ or something higher (I cant speak for Python or anything else, never had a reason to learn it)