Daraius Edibam

Dynamic language means (roughly) that you can change the program instantly (sometimes even when it's still running !).

If you've ever seen a Lisp-programmer working in the Lisp-environment (called the 'debugger'), then you would have seen that he typed the function-definition on a command line like this :
   (defun square (x) (* x y) )

He can then execute this function on the next command-line like this :
  (square 5) ==> error: the symbol Y has no global value

He sees the error, and he can instantly correct it by supplying the correct definition:
  (defun square (x) (* x x)) (square 5) ==> 25

You see ? It looks like just a programming environment (a stupid one in this example), but it's actually a build-in feature in the language.
This kind of languages are normally impossible to compile (ok, Dylan is an expection) because the interpreter can't known anything about the objects/symbols that the programmer uses. If he finds the 'square' symbol somewhere inside the definition of a function, then he can't compile the code to a function-call. He has to wait until the function is executed : each call can have a different meaning for 'square' (a function, a list, a keyword, a variable, ...)

The programmer can interactively define, debug, trace and execute the entire code. You can for example define the 'main-function' without specifying the definitions of the lower-level functions.