The case in Haskell is obvious: 2, "foo", (Data.Set.singleton 3); none of these are functions. Functional programming isn't about "everything is a function", it's about (among other things) making functions first-class, and using higher-order functions to build powerful abstractions. (There are also a lot of other things associated with FP, incl. but not limited to: immutability, purity, powerful static type systems, focus on "plain old data" over OO hierarchies, laziness, etc). If you want a language in which everything is a function, you want the untyped lambda calculus.
The case in Lisp is a little less obvious; I've been meaning to write a blog post about this. But consider:
(let ((x 0)) x)
Is `((x 0))` an expression? No. It's a list of binding-pairs.Just by being somehow basically-reductionist, you'll see that the Universe does a lot with just electrons, protons and photons, yes?
Well... if you for example adhere to String Theory then you will be able to model a lot with just strings http://en.wikipedia.org/wiki/String_theory
"Everything is a String" => you have something Paradigmatic. A "space" or "knowledge body" that can explain a lot using little.
That's what I mean by being paradigmatic.
If you want more reference on this principle and you take the Quantum way, one thing you'll find is that you can explain a lot with little too. There is actually a theorem that deducts Newton's second law using only Schrödinger's Equation as foundation.
But since I've talked about this here I've won 3 downvotes, then you can rest for sure that the Universe is broken.
When "Everything is a.." you're forced to do a lot with little and that's really powerful. And creates a paradigm.
Some examples of it:
LISP: "Everything is an Expression" Haskell and Miranda: "Everything is a Function" Self and Smalltalk: "Everything is an Object"