The first twenty pages of Carnie make the contention that there is some innate understanding of the rules of grammar and syntax. This is illustrated in three ways: first, the postulation that any method of teaching or testing grammar would need to be infinite; second, that many rules seem hard to articulate and unknown but easy for people to call out (ex. "that sounds wrong"); and third, that parents do not actually teach their children grammar, yet they learn the rules around it. I find all three arguments lacking.
On the first issue, of infinity, Carnie gives the example of a machine that would need to scale infinitely. This might be true if our machines were boolean logic operators, but the concept of a finite-state automaton, with different accept/reject states depending on words it is fed, solves the majority of these problems. FSAs can happily interpret (near-)infinitely long strings of words, and determine whether they end validly or not; they can also tolerate nuances, such as having subroutines for subclauses. Alternately, a context-free grammar can be used to articulate the way people think (ex. sentence -> noun verb noun, noun -> adjective noun, verb -> adverb verb, and then a list of nouns, verbs, etc. that can be used as substitute words; in fact, this very structure takes up ~pp70-85). It is not therefore not at all clear that the 'problem of infinity' is a problem for the way we interpret language: we can 'learn' a machine with different states of acceptance and rejection and use that to make our judgments.
On the second and third issues, Carnie suggests that many concepts in language are hard to articulate (ex. when "that" is acceptable and incorrect) and yet that people can call them out; further, parents themselves never tell their kids what is correct and what is not, rather judging them on the basis of content. The latter contention borders on the absurd, since no children at age 6 who say "I writed" could tell you whether "which" or "that" is the appropriate relative pronoun to use. Rather, children have teachers who tell them what is and is not acceptable (a fact Carnie strangely ignores), at which point they begin to identify grammatical correctness – yet they also make mistakes! This itself suggests that Carnie's belief of intrinsic knowledge is flawed, for if people need to 'explore' their grammatical knowledge to 'get a feel' for what is and is not right, it is unclear what component of said knowledge was intrinsic in the first place.
The alternate hypothesis, hinted at in problem sets and seemingly quite reasonable, suggests that humans pick up linguistic rules (components of the context-free grammar or FSA, perhaps) as they go along, and revise them based on the data. Carnie's response to this argument is that we can never know whether our conception of grammar is correct, since another data point might 'disrupt' our hypothesis. Yet the 'disrupting hypothesis' model of learning is precisely how we observe children pick up language and syntax – they make mistakes, are corrected, and then revise their mental model of the syntactical world. Every time people err, they use that to revise their internal framework.
Carnie's only plausible response to this argument is that sometimes, people know whether something is right or wrong because it sounds wrong, not because they know the rules. Yet this is consistent with our alternate hypothesis as well. Every time we are corrected, our internal grammatical model is revised, but there is no obligation for us to be cognizant of precisely how the mind's syntactical rules are structured, just as the individual person cannot 'feel' the rearrangement of their neurons and cannot 'stimulate their language brain', but rather feel their mind as one coherent organ.
Finally, if Carnie truly believes humans are consistently good syntactical checkers, the fact that about three quarters of Americans spot mistakes in each other's communication might be cause for alarm. (There is a serious point here: certain grammatical features and the very construction of sentences do change over time, suggesting that the universal grammar is not quite as universal as Carnie would like.)
This comment has been removed by the author.
ReplyDeleteReally great post--I definitely agree with your contention against Carnie's second and third points. I think people's "intuitive" understanding of what is/is not great is the product of years of reinforced learning from different mentors growing up. In my own experience, I had a filipino nanny growing up who always used to say "helicofter" instead of "helicopter." Consequently, I would incorrectly use the word helicofter until my mom came in and pretty much was like, "what the hell?" and taught me the correct word. Up until that point, however, I thought I was using the right word. I think this is an accurate exemplar of your line, "the 'disrupting hypothesis' model of learning is precisely how we observe children pick up language and syntax – they make mistakes, are corrected, and then revise their mental model of the syntactical world. Every time people err, they use that to revise their internal framework." I think over time these pronunciations and grammatical rules become so embedded in our understanding of language that we mistakenly call them intuitive. However, it's almost akin to the way in which we immediately skip the computational process of 2+2 and just say 4 when asked to make this calculation. No one comes out of the womb with that knowledge embedded in them; at some point we had a teacher who taught us how to add and over time we developed an ability to just have the word four pop into our head whenever someone asks us what 2+2 is.
Delete