This is my own take on perhaps the most frequently misunderstood argument in the history of linguistics. What makes this argument amazing is that understanding it doesn’t require any knowledge of linguistics. Yet, I think it is one of the most important insights in the whole history of linguistics. Much of what is written here is not particularly new, and some great posts (and comments) about the argument can be found on the Faculty of Language blog. However, writing something yourself is always a great way to gain a deeper understanding of some problem.
Part 1 deals with the more or less classic version of the POS argument. In the upcoming part 2, we will look at what progress has actually been made in solving the issue that the argument addresses. I will try to be as concise as possible.
The logical problem of language acquisition
How do children acquire a language? As is often the case in science, it is necessary to first get the definitions right. And in this case, we must first define what it means to know a language. Here’s the simplest possible definition: to know a language means to be in a state where you can draw the difference between utterances that belong to your native language and utterances that don’t. To successfully acquire a language, then, means to reach that state. Consider the following English Yes/No questions:
(1) Is Mary at home? (Possible answer: Yes, Mary is at home.)
(2) Will John come to the party? (Possible answer: Yes, John will come to the party.)
It seems that the declarative sentences and Yes/No questions are related in a specific way: to make a Yes/No question, find an auxiliary in the corresponding declarative sentence and put it at the front. All (adult) native English speakers know this in the above sense. Obviously, native speakers do not have to explicitly follow the procedure of finding auxiliaries and moving them every time they want to ask a question. They just know that Yes/No questions that happen to be related to their declarative counterparts in this and only this way are valid English utterances. Take a moment and read this again. Speakers do not need to be linguists to understand that (1) and (2) are correct ways to ask a question in English. But they speak as if they obey the rule above.
Actually… that seems to be the wrong rule!
To see why, consider a couple more examples:
(3) Will you believe that I am a linguist?
(3’) Am you will believe that I a linguist?
These sentences have two auxiliaries. Since (3’) is a word salad and (3) is the only correct way to ask a Yes/No question, we can see that it is not enough to just find some auxiliary (AUX) and put it at the front. In fact, there are three possible options:
a. Move any AUX to the front
b. Move the leftmost AUX to the front
c. Move the main clause AUX to the front
Like I just pointed out, rule (a) appears to be wrong. However, both rules (b) and (c) appear correct. Which one do native speakers actually obey? Let us look at another example:
(4) The man who is tall will leave now.
(5) Will the man who is tall leave now?
(5’) Is the man who tall will leave now?
If English speakers obeyed the rule (b), then (5’) would be a correct way to turn (4) into a question, which it is not! On the other hand, rule (c) rules out both (3’) and (5’) while correctly predicting that (1), (2), (3) and (5) are all correct ways to form a question. Finally, we can say that adult English speakers know the rule (c).
Which, in turn, means that children too come to know the rule (c). Here’s the one million question: why do children prefer this rule and not the rule (a) or (b)? Obviously, to figure out which rule is the right one, children need to be exposed to some language data (which is called primary linguistic data, or PLD). So it is important that we turn our attention to the PLD and its characteristics.
First of all, PLD is finite. Children neither have access to all possible sentences of their native language, nor do they have access to the same data. This means that children may never hear some complex sentences like (3) and (5), which are required to even consider that something like the rule (c) exists, as we have seen above.
Second, since PLD is basically whatever people around say, it is by definition correct. Adults already speak as if they obey the rule (c) and never make errors and say anything like (3’) or (5’). The wrong utterances are usually called negative data. If there is no negative data in the PLD, it is impossible to conclude that the rule (c) is the only correct one among three alternatives.
To sum up, a growing child might not have an opportunity to even consider the rule (c) as a possible rule of English, and even if she did, she would not have reasons to prefer (c) over (a) or (b), which are both compatible with her observations. So why do literally all children still fix on the rule (c)? Here’s the answer:
Humans are built to never even consider options like (a) or (b).
This is the whole argument. Given the premises, the logic seems airtight. In the upcoming part 2, we will take a look at how good the premises are, given recent experimental research (spoiler: they’re good!).
P.S. What this argument is not about
It seems that even serious linguists (like Geoffrey Pullum here) sometimes misconceive the POS argument. Apparently, for Pullum, the argument refers to the obvious fact about rocks and kittens that Noam Chomsky occasionally mentions in his talks: even if kittens or rocks are exposed to the same language data as children, they do not begin to talk. So, not only do you need a brain to talk, but a brain that is language-ready. This, however, has little to do with poverty of stimulus. It is just a funny way to say that at least in some sense, human brains are special.