> hiding the "real" language in a higher-level encoding
Exactly. In the middle of the article they discuss coming up with an expression to mean "car". Of the many word combinations one might choose (they go with "tomo tawa"), one of them would have to emerge as the standard way to say "car", otherwise effective communication would not really be possible. But at that point you've just created a new word. "Tomo tawa" no longer means anything that combination of words might possibly mean - it means car.
I'd suggest there's a certain irreducible vocabulary, and its size is going to be the same no matter how many base words are used to compose it.
Basically, exactly what happens with compounds like <black> + <bird> => <blackbird> -- blackbirds are a specific sub-category of "black birds", not just any bird with black feathers.
And it's completely analogous to every other language where we come up with words by combining "auto" and "mobile". It would be challenging to determine what an automobile was from just the word and it would be almost impossible to recreate the word had you just been given the object and asked to come up with one. Might as well say that English is a language with 26 words.
Though I imagine this makes spelling and pronunciation easier to remember.
Exactly. In the middle of the article they discuss coming up with an expression to mean "car". Of the many word combinations one might choose (they go with "tomo tawa"), one of them would have to emerge as the standard way to say "car", otherwise effective communication would not really be possible. But at that point you've just created a new word. "Tomo tawa" no longer means anything that combination of words might possibly mean - it means car.
I'd suggest there's a certain irreducible vocabulary, and its size is going to be the same no matter how many base words are used to compose it.