There is nothing there corresponding to the words of natural languages.
Let us now make a generalization that will make our task clearer: Simple expressive speech acts, even when performed intentionally, are not \'linguistic\' in the sense we are trying to make explicit, and the corresponding words of actual languages are not \'words\' in our sense. Ouch! Damn! Yuck! Wow!
We need to distinguish between the conventional meaning of words, sentences and other symbols, and the speaker meaning which the speaker expresses in making an intentional utterance.
Now we are getting much closer to language, because the first phenomenon is essential to the performance of speech acts, and the second phenomenon, the repeatable devices, consist typically of words and sentences of a language.
That is another way of saying that the hominids need to evolve elements that correspond to our words and morphemes and to ways of combining these into sentences in a compositional manner, in a way that enables the participants to figure out the meaning of the sentences from the meanings of the elements and their arrangement in the sentence.
There must be repeatable devices each of which can function as a possible communication unit (sentence) and these must be composed of elements (words) which are such that the communicative content of the whole is determined by the elements and by the principles of their combination in the sentence.
How do we introduce these features—words and sentences----where the sentences are systematically built out of the words?
We will add rules or procedures for arranging those devices (words) into the complex resultant structures (sentences).
The paradox is: how do we achieve the unity of the sentence (and hence the unity of the expressed proposition) when the sentence is entirely composed of discrete entities, the string of words and morphemes that constitute it?
They are parts of sentences, and thus correspond to words, but they are not yet whole sentences.