I chanced across the original paper [1] by Dr. Thomas Bak a few days ago through a BBC article [2]. The section how they measure this is interesting, for those interested.
Probably applicable to spoken languages as well as computer languages. I find that having deep dived into lisp out of interest, my functions in other languages have over time become almost purely functional, with the occasional side effect - I actually cringe a little bit when a function doesn't return anything. The result is that my code has become immediately unit testable. Perhaps the analogy is true?
[1] PDF: http://onlinelibrary.wiley.com/doi/10.1002/ana.24158/pdf