Seriously though, most of the papers I've read require significant interpretation, filling in gaps and dealing with ambiguous terminology or use of symbols.
What I meant was that there is more recalling involved (more of step 1) when you want to say something than when you want to prove something. Of course you are not aware of it. The brain does it subconsciously.
1) Recall the grammar/rules of the output language (for both natural language and formal languages).
2) Form the idea of what it is you want to express in your head using your internal language (usually your first natural language).
3) Encode the idea you have in your mind using the formal rules from step 1.
Step 1 is usually internal and subconscious for natural languages that you can communicate in freely and to an extent this is even the case for other languages and even programming languages.
For your first natural language, step 1 is usually done behind the scenes by your brain automatically, or these rules are already present in your brain in a usable form, and step 3 is trivial because you are just expressing the idea which is already in this language in your mind.
There are two steps to conveying a thought in a language:
1. recall the grammar rules and vocabulary 2. form the sentence
The steps are fundamentally different in that the first one is an exercise in recollection whereas the second one is a "creative" process of some sort.
To prove a mathematical fact you go through the same two steps:
1. you recall all the past mathematical knowledge 2. you deduce the fact in a creative process of meshing together the facts you just recalled
So conveying a thought in a language is sort of analogous to proving a mathematical fact.
But clearly when doing math step 2 is given the spotlight whereas when you are talking in a foreign language step 1 is by far the hardest.