In my opinion, that's all you can really say about them.
The idea is that in a higher-order language, a function is just another kind of value. The same way you can have (3+4) without an identifier in C, you can have (lambda (x) +(3 x))) without an identifier in scheme.
The key here isn't that there's anything special about anonymous functions. It's only the restrictions of other languages that require functions be treated differently from any other value. The special definitions are for procedural languages which don't allow them, not for functional languages with do.