McCarthy's 1960 paper gives notation two different jobs, and section e is where you can watch them separate. Functions are first written as M-expressions, which is how the paper explains them to a reader. Then they get translated into S-expressions, so that a function can be handed to another function and operated on as data.
A description for explaining a function and a representation for an operation to consume don't have to be the same thing. In the paper they aren't. The M-expression side never became the notation programmers actually wrote, and the data representation ended up doing both jobs.
Calling all of it Lisp syntax flattens that history and hides a real choice, whether the notation you read and the structure a program manipulates should be one artifact or two. Lisp ended up with one. Scala keeps them apart, since a macro receives a typed tree rather than the text you wrote. I don't think either answer is obviously right, and the paper is a nice place to see the fork while it was still open.
