Sorry to be so slow in responding to this - I tried to find some good references but got nowhere definite. My best impression is that the concept first came to light in the mid '60s ... probably inspired by the structure of the Algol 60 programming language. Likely first appeared in "research" operating systems in continental europe (people like Dijkstra and Wirth were much more keen on recursion than most UK or US researchers) and then appeared in better-known systems like Multics and George 3. Some systems had a limited number of levels of hierarchy - typically system/user and sometimes one more level below. Often this was done by "cheating" with what was really a single level directory. I'm not sure how Atlas and Titan did it. I'm also not convinced that the early versions of TOPS 10 had anything better than this ... similarly for the GE time sharing system that eventually evolved into GCOS. Andy