April 10, 2012

St Augustine rides again

Psychologists at the University of Pennsylvania claim to have discovered that babies at 6 to 9 months can understand certain nouns, according to a report on Science Daily:

In research focused on 6-to-9-month-old babies, University of Pennsylvania psychologists Elika Bergelson and Daniel Swingley demonstrated that the infants learned the meanings of words for foods and body parts through their daily experience with language.
---
These findings unseat a previously held consensus about infant learning. It was widely believed that infants between 6 and 9 months, while able to perceive and understand elements of the sounds of their native language, did not yet possess the ability to grasp the meanings conveyed though speech. Most psychologists believed word comprehension didn't emerge until closer to a child's first birthday.
---
To test this belief, Bergelson and Swingley recruited caregivers to bring their children to a lab to complete two different kinds of test. In the first, a child sat on the caregiver's lap facing a screen on which there were images of one food item and one body part.
The caregiver wore headphones and heard a statement such as, "Look at the apple," or, "Where's the apple?" and then repeated it to the child. The caregiver also wore a visor to avoid seeing the screen. An eye-tracking device, which can distinguish precisely where a child is looking and when, then followed the child's gaze.
The second kind of test had the same set-up, except that, instead of the screen displaying a food item and a body part, it displayed objects in natural contexts, such as a few foods laid out on a table, or a human figure. For both kinds of test, the question was whether hearing a word for something on the screen would lead children to look at that object more, indicating that they understood the word.
In total, Bergelson and Swingley tested 33 6-to-9-month olds. The researchers also had 50 children from 10 to 20 months complete the same tests to see how their abilities compared with the younger group.
---
... the 6- to 9-month-old babies fixed their gaze more on the picture that was named than on the other image or images, indicating that they understood that the word was associated with the appropriate object.
This is the first demonstration that children of this age can understand such words.
---
"I think it's surprising in the sense that the kids at this age aren't saying anything, they're not pointing, they're not walking," Bergelson said. "But actually, under the surface, they're trying to put together the things in the world with the words that go with them."
"I think this study presents a great message to parents: You can talk to your babies and they're going to understand a bit of what you're saying," Swingley said. "They're not going to give us back witty repartee, but they understand some of it. And the more they know, the more they can build on what they know."


Undoubtedly, Bergelson and Swingley's observations are interesting (though not, perhaps, as surprising as they suggest). They help make us aware of the enormous amount of learning that goes on in a child throughout its waking hours from the moment of birth, and precedes the use of language in communication. However, I would argue that their account of the experiment misconstrues the relation between the phenomena they've recorded and what it means to understand words. Linking a sound with a type of object is not what we mean by understanding a word.

In fact, Bergelson and Swingley's view of language learning and word comprehension comes close to that of St Augustine, as discussed by Wittgenstein in the opening remarks of the Philosophical Investigations, as well as that of empiricists like Locke or Hume.

Generally speaking, we understand a word when we're able to use it in speaking to others, and to respond appropriately to others' use of it. Nouns usually occur in utterances, where they may be part of a report of some event ("The ball fell out of the window") or state of affairs ("The ball isn't here"), an order ("Help me find the ball"), a question ("Where is the ball?"), a warning ("Look out, here comes the ball!"), or any number of other things. These utterances, in turn, occur in the context of human interaction. In some contexts, the words uttered call for looking in the direction of the ball (but doing other things as well), in other cases they don't. What constitutes understanding is - in broad terms - the ability to do certain things, not the tendency to have a certain reaction or undergo a certain experience.

It seems that St Augustine, Locke and Hume were tempted to think that we learn to understand utterances by first learning to understand words and then learning to understand their combinations. But how are to suppose that "The ball fell out of the window" comes out of a tendency to link the sounds in it to various objects (or events, etc)?

I think we should say: what constitutes understanding a word (in the sense in which speakers of a language are said to understand the words in the language) is logically dependent on what constitutes understanding utterances. And this depends in part on what the utterance consists of, in part on the situation in which it is made, the relation between the speaker and hearer, etc. In fact, what it means to understand a word is highly abstract: saying that someone understands a word is an indirect way of referring to the kinds of abilities we acquire in learning to use language in communication. Linguistic communication begins with learning to understand people and the ways in which their utterances enter into the situations in which we interact, not with the understanding of individual words.

The reason we fail to see the complexity of the notion of understanding a word, I would suggest, is that we are familiar with the case of "learning a new word", either in learning a foreign language or in learning about some specialized use of our own language. However, when we are told that the French for horse is "cheval" or that this tree is called "manchineel", the reason we know what to do with this piece of information is that we already have an impressive command of various underlying linguistic and non-linguistic skills.

The six-to-nine month old baby doesn't have such command - although he or she is fast acquiring it. Whether and how the infant's reaction to speech sounds contributes to learning to understand speech is not a matter that can be resolved a priori.

(Anyway, the researchers' advice to parents is sound: speak to your child, don't wait until you're spoken to.)