I concur. I've been working on http://fexl.com for some time, and when all is said and done, for my embedded scripting applications I prefer to use simple line- and token-oriented languages, where everything looks like this:

  word word word
  word word
  word word word word
  ... etc.

The semantics of these embedded languages tend to be domain-specific, or in the cases where I need Turing-equivalence, revolve around the notion of a general "whiteboard" where the program can scribble key-value pairs, and do basic conditions and looping. I find that non-programmers can follow this sort of logic very easily.

Nice thing is, I can write these interpreters in any language, such as C or Perl or whatever, and "sandbox" their semantics in an airtight way for security.