Patricia Kuhl: The linguistic genius of babies | Video on TED.com

Patricia Kuhl: The linguistic genius of babies | Video on TED.com

"At TEDxRainier, Patricia Kuhl shares astonishing findings about how babies learn one language over another -- by listening to the humans around them and "taking statistics" on the sounds they need to know. Clever lab experiments (and brain scans) show how 6-month-old babies use sophisticated reasoning to understand their world."

Can children learn abstract syntactic principles by using general cognitive capacities?

One of the most hotly debated issues in the study of language acquisition is whether the abstract syntactic principles of a language can be learned by children

1. by using domain-general capacities (such as pattern finding, analogy, statistical learning, categorization and generalization, etc.)

or whether they need

2. innately specified knowledge of language that enables them to form the right abstract syntactic categories that cannot be infered from the surface level of linguistic utterances (Chomsky's Poverty of Stimulus Argument)

In a new paper in the journal Cognition, Perfors et al. (2011) argue that domain-general capacities are sufficient for children to be able to learn abstract syntactic principles inherent in the linguistic input.

Here's the abstract:

Children acquiring language infer the correct form of syntactic constructions for which they appear to have little or no direct evidence, avoiding simple but incorrect generalizations that would be consistent with the data they receive. These generalizations must be guided by some inductive bias – some abstract knowledge – that leads them to prefer the correct hypotheses even in the absence of directly supporting evidence. What form do these inductive constraints take? It is often argued or assumed that they reflect innately specified knowledge of language. A classic example of such an argument moves from the phenomenon of auxiliary fronting in English interrogatives to the conclusion that children must innately know that syntactic rules are defined over hierarchical phrase structures rather than linear sequences of words (e.g., [Chomsky, 1965], [Chomsky, 1971],[Chomsky, 1980] and [Crain and Nakayama, 1987]). Here we use a Bayesian framework for grammar induction to address a version of this argument and show that, given typical child-directed speech and certain innate domain-general capacities, an ideal learner could recognize the hierarchical phrase structure of language without having this knowledge innately specified as part of the language faculty. We discuss the implications of this analysis for accounts of human language acquisition.

Social Cognition and Linguistic Innateness I

Normal 0 21 false false false MicrosoftInternetExplorer4

Laughing Man of Complex Adaptive Systems has asked an interesting question regarding my post announcing that Michael Tomasello had published a new book on “The Origins of Human Communication:”

“A point I don't get: How does this argument (Gricean maxims of cooperation evolved from general cooperative human structure) criticise Chomsky's innateness theory?”

He refers to the following passage from the book description:

“Challenging the Chomskian view that linguistic knowledge is innate, Tomasello proposes instead that the most fundamental aspects of uniquely human communication are biological adaptations for cooperative social interaction in general and that the purely linguistic dimensions of human communication are cultural conventions and constructions created by and passed along within particular cultural groups”

As Laughing Man rightly points out, the evolutionary priority of cooperative, social-communicative capacities to our capacity for linguistic communication is in principle compatible with an innatist position regarding language acquisition. A generativist more interested in interdisciplinary matters than Chomsky, as for example, Ray Jackendoff, would be sure to agree with the importance of social interaction for language acquisition.

However, Tomasello’s argument goes further than that. He claims that general skills of social interaction such as human intention-reading skills, along with primate pattern finding skills, are all what is needed for a child to be able to learn a language when placed in the right environment. I haven’t looked at Tomasello’s newest work, but in his previous books “The Cultural Origins of Human Cognition” (1999), and “Constructing A Language” (2003), he makes a strong case why we needn’t postulate a language-specific, innate “Universal Grammar” specifying the architecture and input conditions of a domain-specific language system.

In Contrast, here’s Ray Jackendoff’s opinion on the relationship between social cognition and an inbuilt specialized linguistic toolkit:

“Language acquisition of course rests on social skills such as theory of mind (rudimentary in chimpanzees) and understanding pointing (not found in chimpanzees), as well as on more general perceptual machinery such as attention. These provide scaffolding for language acquisition but do not themselves provide the material out of which knowledge of language is built. “

Language also relies on many other domain-general cognitive capacities, but

“what is special about language is the collection of mental structures it employs – phonology and syntax – and the interfaces between them and conceptual structure. The processing and acquisition of these structures may be accomplished by brain-general mechanisms of long-term memory, integration in working memory, learning (including statistically-based learning), and attention, and may rely as well on understanding of social interaction and theory of mind. But unless the specific unique building blocks for phonology, syntax, and their connection to concepts are present, language and language acquisition are not possible. It is these building blocks that constitute the narrow language faculty.” (Jackendoff 2007: 388f.)

References:

Jackendoff, Ray (2007): Linguistics in Cognitive Science: The State of the Art, The Linguistic Review 24, 347-401.

Tomasello, Michael (1999): The Cultural Origins of Human Cognition. Cambridge, Massachusetts; London, England: Harvard University Press.

Tomasello, Michael (2003): Constructing A Language. A Usage-Based Approach. Cambridge, Massachusetts; London, England: Harvard University Press.

Innateness, Creoles, Birdsong, and Blue Brains

Phew. I’ve got so much stuff to do, and things to read that I can hardly find any time to blog at the moment. But here are some interesting things you should check out:

Over at Language Log, Mark Liberman has posted very interesting two-part discussion of “creole birdsong” and its implications for human language acquisition, especially regarding the question of innateness of language.

Creoles, i.e. stabilized and grammaticalized languages emerging from and incomplete linguistic input, crop up when a group of children is exposed only to impoverished pidgin input. The fact that children are able to construct a new, full-fledged language from poor data, is often attributed to a view of language acquisition as

“an interaction between environmental exposure and innate abilities.” (Senghas & Coppola 2001)

A particularly interesting example is Nicaraguan Sign Language,

“a signed language spontaneously developed by deaf children in a number of schools in western Nicaragua in the 1970s and 1980s.”

The children had no exposure to any other kind of language and beforehand only communicated in homesign.

But.

"sequential cohorts of learners systematized the grammar of this new sign language.” (Senghas & Coppola 2001).

Interestingly, the learners which systematized the language were all under the age of 10, an age when the proposed linguistic biases are supposedly more pronounced than in adulthood.

Liberman reports that something similar happened in Ofer Tchernichovski’s lab colony of zebra finches. The initial founder of the colony grew up isolate from others and didn’t learn to sing properly, because there was no one to imitate. However,

“As each succeeding generation learns songs from the preceeding one, the effects of biases in the learning process accumulate, so that after a few generations, normal zebra finch songs have re-emerged.”

Liberman links this to the discussion of Derek Bickerton’s language bioprogram hypothesis, which proposes that children use their innate language capacities, some kind of a mental template for language, when learning and creating language, which has come under critique in the last two ecades.

Liberman argues that, when social learning is involved,

“perhaps it's normal for the phenotype to emerge over multiple generations, and to involve complex relationships among genetic and social structures.”

This also sits well with the proposals of Simon Kirby and his colleagues from the Language Evolution and Computation Research Unit, who propose that language acquisition can be seen as the interaction of a small set of genetic biases (a set of “Universal Biases” instead of a Universal Grammar, so to speak) where

“cultural transmission can magnify weak biases into strong linguistic universals” (Kirby et al. 2007) and language is the “result of nontrivial interactions between three complex adaptive systems: learning, culture, and evolution“ (Kirby et al. 2007).

Thus,

“a population whose speakers are linguistically biased — for whatever reason — may, over many generations, transform its language in ways that reflect the preponderance of individual biases among language acquirers“ (Ladd et al. 2007, see also this previous post on the work of Dan Dediu and Robert Ladd)

(As I just see, Robert Ladd has also drawn attention to this recent paper of his in the comment section of Language Log)

Kirby and his colleagues also have published a paper on the evolution of birdsongs, but do not refer to Tchernikovski's work, and the paper in general is way to advanced and technical for me to understand it.

In his second post, Mark Liberman, extends his thoughts on the possibility of a “multi-generational language bioprogram”, and gives some examples of simple learning algorithms which embody

“a bias towards certain outcomes” that lead to “coherent shared patterns,” like a shared vocabulary, or, more generally, a shared language

The comment sections of the two posts are also very interesting, as prominent figures such as Derek Bickerton himself, and Shimon Edelman, professor of psychology at Cornell university weigh in on the question.

-------

UPDATE:

There is a guest post on Language Log by Derek Bickerton, in which he adresses the issue of a multi-generational bioprogram and rejects it in favor of a single generation bioprogram, which as far as I cunderstand it, doesn't have anything to do with a specific Chomskyan "language organ," but with infant brains being pre-equipped for learning symbols/words andputting them together in fashion of neural sequential movement planning. And
"The way the brain does it causes the child automatically to put words together in certain ways."

Uhm. Er well. So the brain does it. Cool. problem solved.

On a related note, Talking Robots, a very cool podcast on Robotics and AI, features an interview with Henry Markram of the Blue Brain Project, which is

“the first comprehensive attempt to reverse-engineer the mammalian brain, in order to understand brain function and dysfunction through detailed simulations.”