MIT Scheme has the ability to do this. (If you really want, comment on this and I'll give you the code. I had to find some undocumented functions to make it happen.)
However, it's not in the Scheme language definition, so implementations don't have to allow it. The reason for this is that in order to make functions faster, a good Scheme implementation will modify the functions. This means both rewriting them in a different language (either machine code or something fairly low-level) and taking out any bits you don't need - for instance, the +
function must in general check whether its arguments are numbers, and if so, what sort of numbers, but if your function is a loop which calls +
, you can just check once at the beginning, and make the function a lot faster.
Of course, you could still keep the lists around without too much trouble, even with all of these things. But if you tried to modify the lists, how would it work?
(Again, you could make it work. It would just be more work for implementers, and since it's not usually used in programs, most people probably just don't want to bother.)