Universal Grammar

Alright all you lurking cog sci majors (psychology, neuroscience, linguistics, philosophy, AI), what are your thoughts on UG? Is it the single best explanation for language acquisition, or would a connectionist model be better? What is the state of research on connectionist models of language acquisition?

Other urls found in this thread:

scientificamerican.com/article/evidence-rebuts-chomsky-s-theory-of-language-learning/
twitter.com/SFWRedditGifs

UG is more a prerequisite for explanation than an explanation itself. In other words, you have to presuppose UG for your explanation to make any sense. Once you learn some basic syntax it becomes very hard to entertain non-generative hypotheses.

Chomsky is a self-hating structuralist.

But you shouldn't equate generative grammar with UG, despite the fact that both are bound. Models which argue for the emergence of syntactic knowledge are often still generative, but reject the idea that grammatical competence is built-in.

It's a complete "cart-before-the-horse" situation. We are apes who have adapted to be able to speak, so-called "universal grammar" is nothing of the sort, rather it's the innate predispositions towards forming language we inherited from our pre-linguistic ancestors.

Chimpsky merely refuses to accept that humans have any innate human nature whatsoever, so instead insists that grammar-- an abstraction created by humans to organise our languages-- is somehow "more fundamental" than our physical realities. He's a modern-age Platonist, and should be dismissed / ignored like we do other Platonists.

I strongly suggest you read Chomsky, because none of this is an accurate description of his ideas.

>Models which argue for the emergence of syntactic knowledge are often still generative, but reject the idea that grammatical competence is built-in.
Such as? In my experience even non-transformational approaches accept UG.

No, Chomsky would say that UG is part of human nature, and that it's really just an abstraction for some properties of the human brain.

Perhaps this is a loose definition of 'generativity', but construction grammar certainly rejects any notion of UG. You could however say that construction grammar is not generative, although it nevertheless produces the infinite set of utterances within a language using a finite set of atomic expressions. In that sense, it is generative.

>You could however say that construction grammar is not generative
I was thinking of construction grammar too. The reason I would equate generative grammar with UG is partly because the construction grammars used by people who accept UG are often called generative grammar, and the ones used by people who reject UG are not. For instance, most construction grammars are not usually called generative grammar, but Jackendoff's parallel architecture is considered to be a non-mainstream generative grammar, while cognitive linguistics is considered to be non-generative. Even though the word "generative" was originally used (rightly) in the sense you're using it, I think it has acquired an additional dimension related to UG.

noam chomsky is so fucking retarded.

From a language point of view, i'm more attracted to generative lexicon.

Is this little bio more or less accurate?

Noam Chomsky is a professor emeritus of linguistics and philosophy at MIT. In his early career he garnered fame for instigating a major paradigm shift in linguistics and cognitive science. Later in his career, he became interested in various theories of quartz crystals and other occult and esoteric fields of study. In 1974, Chomsky was introduced to quartz crystals and spent the rest of the 1970s investigating their tendency to also act as a vessel for storing, magnifying, and converting mental energies. Chomsky claimed to be able to duplicate the "Backster effect" (the idea that plants "feel pain" and have extrasensory perception (ESP), which was widely reported in the media but was rejected by the scientific community) using plants as transducers for bio-energetic fields from the human mind, showing that they respond to human thought. He claimed his findings had the same effect irrespective of distance and suggested that "inverse square law does not apply to thought." According to Chomsky, he had scientifically proven the existence of a compound that he described as a form of energy in the body that was the physical manifestation of the libido, building up in the body until it was successfully discharged through an orgasm.

Without some form of universal between people, the study of language is non-science, morw akin to just Bible study, exactly like what Chomsky got his start doing, going over old texts with the intention of finding better translations, or treasure hunting like Indiana Jones.
If this is your idea of interesting, you're on the wrong board.

That's a fair point. Does Jackendoff assume UG?

I agree, and I think everybody linguistics assumes that there is an upper-bound and a lower-bound to what constitutes a language. However, UG makes the claim that the structure of syntax is innate, and learning a language boils down to adjusting parameters. This means that language has its own module in the brain. Competing models will instead claim that language learning emerges not from a specific language module, but from a combination of other cognitive abilities.

Nope.

yes

Correct me if I'm wrong, but I don't think LFG assumes UG (I honestly know very little about LFG).

I don't know much about it either. I know HPSG usually assumes UG, and LFG shares a lot with HPSG.

Then what does that make you?

Is there anyone here who will defend opposing UG?

I guess we're all in programs where we are told UG is correct and are not exposed to other theories. Here's a good article arguing for the other view.
scientificamerican.com/article/evidence-rebuts-chomsky-s-theory-of-language-learning/