a continuation on the discussion of Hockett’s features of language.
Consider a Martian exposed for the first time to a human language, thus having no preconceived notions or intuitions about how they are structured. The Martian must simply take the data they have and form hypotheses, testing it against more data. They only receive positive and negative feedback (presumably from some friendly humans).
The Martian analyses the construction of interrogatives from assertions.
‘The boy is late.’ -> ‘Is the boy late?’
Maybe we just move the third word to the front?
‘The boy in the red shirt is late.’ -> ‘In the boy the red shirt is late?’
Negative feedback received, that didn’t work, maybe move ‘is’?
-> ‘Is the boy in the red shirt late?’
Positive feedback, there is greater confidence in the hypotheses.
‘The boy who is wearing a red shirt is my brother’
-> ‘Is the boy who wearing a red shirt is my brother?’
Negative feedback, why didn’t this work?
This method of ‘linear’ scanning each word and applying hypotheses quickly breaks down for larger and more complex sentences. In order to properly appreciate the meaning of these sentences, and how we can manipulate them, we have to recognise the “units” comprising the language. We have to pay attention not to the number of words, but the structure. It’s as if there is some sort of ‘invisible’ structure in the sentences, and we already know how the individual words in the sentence relate to each other and the sentence as a whole.
“(This knowledge is) part of the child’s biological endowment, part of the structure of the language faculty.”
~ Noam Chomsky (1988)
According to Chomsky, language has to be organised on two levels: the surface structure, and the deep structure. He argued that there is a deep structure with some intended meaning, to which we apply the transformational rules of our internalized grammar, resulting in what we say or write, the surface structure.
Two sentences with distinct deep structures may not necessarily result in distinct surface structures. When this happens, we have what is known as an ambiguous sentence. We cannot reliably disambiguate the surface structure to get what we know to be the one intended deep structure. For example:
“I’m not going to buy her flowers.” ‘I’ will not buy flowers for her, or will ‘I’ not buy flowers from her?
“In order to catch the train George drove furiously to the station”
“The reporters assigned to George drove furiously to the station”
Garrett took these two sentences, and spliced recordings of a person speaking them such that the second part of the sentence, “George drove…” would be identical for both, leaving no auditory clue other than the sentence itself. Listeners were asked to mark where there should be a break in the sentence, before or after ‘George’.
The test subjects consistently identified that the break should be before ‘George’ in the first sentence, and after ‘George’ in the second sentence. The phrasing and the structure of the two sentences is perceived to be different, even though large pieces of the two are identical.
There is also a concept of recursion in such parsing of sentences, the idea that a NP, for example, can be embedded within another, or a sentence in another sentence, and so on.