Of Minds and Language (38 page)

Read Of Minds and Language Online

Authors: Pello Juan; Salaburu Massimo; Uriagereka Piattelli-Palmarini

BOOK: Of Minds and Language
10.95Mb size Format: txt, pdf, ePub
Lila Gleitman

I would like to back up a little and point the conversation toward the case of the child learning the meaning of a word – a theme which came up in Noam Chomsky's discussion earlier in this conference, and also, in a very different way, in Wolfram Hinzen's talk about arguments and adjuncts.
17
Here's the problem. It's obvious that in deciding on the meaning of a new word, we rely at least in part on the extralinguistic situation, the context in which the word is being uttered. What's obvious, though, is only
that
this is so. What is not obvious and, rather, lies almost altogether beyond our current understanding is
how
this is so, or even how it
could be
so. The information that children – or any learners – get from the world about the meaning of a new word is often flimsy, certainly variable, and not infrequently downright misleading. This is perhaps most poignant in the case of verbs and their licensed argument structures. I got interested in this problem about thirty years ago when Barbara Landau and I studied language acquisition in a congenitally blind child (Landau and Gleitman 1985). We were very startled to discover that the first verb in this child's vocabulary, at two years old or maybe even slightly younger, was
see
, and her usage seemed much like our own from the start, referring to a perceptual accomplishment. That is, this child never seemed to have confused
look
or
see
with
touch
, even though, given her perceptual capacities, she herself necessarily touched as a condition for seeing. This case dramatizes the fact that while it is true that situational context commonly fits the intended interpretation, most of the explanatory burden for understanding learning rests on the infant's ability to represent that context “in the right way.” In this instance, the contexts of the teacher/speaker (the sighted adult community) and the learner aren't even the same ones. In this brief discussion I want to illustrate the issues by showing you some findings from Peter Gordon (2003) demonstrating prelinguistic infants' remarkable capacities and inclinations in regard to the meaningful interpretation of events.

In Gordon's experiments, infants of about 10 months of age (who as yet utter no words) are shown videos depicting what to adults would be
giving
or
hugging
events. In the former case, a boy and a girl are shown approaching each other; one hands a stuffed bear to the other, and then they part. In the latter video, the two approach each other, embrace, and then part. The clever part of this manipulation is that in the
hugging
scene as well as in the
giving
scene one of the two actors is holding the stuffed bear. So crucially there are three entities involved in a motion event, in both cases. The only difference between the two events is that only in the
give
scene is this toy transferred from one participant's
grasp to the other's. Gordon recycled these videos so that infants saw them again and again, leading to habituation (measured as the infant spending less and less time looking at the video at all, but rather turning away). Any individual baby in this experiment saw only the give scene or only the hug scene. Once babies were habituated, they viewed new scenes that were identical to the originals except that the toy was now absent.

Fig. 14.1. Habituation effects for argument versus adjunct: This figure graphs habituation in infants who are watching either a scene depicting
giving
or
hugging
(panel a). When a toy animal that one character is carrying is subsequently removed from the video, dishabituation is observed for the
giving
video but not for the
hugging
video (panel b).

Source
: Courtesy of P. Gordon, 2003

As you see in
Fig. 14.1
, babies dishabituated (started visually attending again) in response to the new (toyless)
give
scenes but not to the new (toyless)
hug
scenes. Gordon also tracked the babies' eye movements to various scene elements during the course of the events. What is shown in the next two Figures is the proportion of time that the babies visually attended to the three entities – the boy, the girl, the toy – as the event unfolded in time, specifically, before, during, and after the two actors interacted.

For the
give
scene (
Fig. 14.2
) visual attention is heavily attracted to the toy as the actors encounter each other; and when the toy is removed the infants persist in looking at the actors' hands – where the toy used to be – as though searching for it. In contrast, they did not seem to notice the toy very much when it was there in the
hug
scene, as
Fig. 14.3
shows.

No more did they seem to notice when it magically disappeared. That is, they hardly looked toward the hand of the hugger who previously had held it, nor provided other measurable signs that they were thinking, “Whatever happened
to that delightful stuffed animal?” Apparently, the babies' implicit supposition was that, even though stuffed bears are of great interest in everyday life,
hugging
events are not “relevantly” changed as a function of whether one of the huggers is holding one of them during the performance of this act. But an act of
giving
is demolished if the potential gift does not change hands. Bears are no more than adjuncts to
hugging
but they can be arguments of
giving
.

Fig. 14.2. Visual attention to argument change: This figure shows eye-tracking records for infants to the toy animal in the
give
scene as the characters approach, contact each other, and depart (panel 1) and the persistence or enhancement of visual attention when the toy (that which is given) subsequently disappears (panel 2).

Source
: Courtesy of P. Gordon, 2003

Fig. 14.3. Visual attention to adjunct change: Visual attention is diffuse across the characters in the
hug
scene (panel 1) but shifts to the hugger (the boy) and huggee (the girl) when the toy disappears. The toy itself is largely ignored (panel 2).

Source
: Courtesy of P. Gordon, 2003

In one sense these charming findings are unsurprising. Of course it would have to be the case that infants could recognize these entities and represent their roles differently as a condition for acquiring
hug
and
give
. But we are very much lacking in any detailed knowledge of the conditions or procedures that underlie
evocation of these representations for the sake of word learning. How does an infant – or for that matter an adult – select relevant representations from those made available by inspection of the world that accompanies speech acts? I believe that many developmental psychologists breezily beg or at least trivialize the questions and puzzles here by suggesting that word learning is at bottom demystified merely by alluding to the reference world. Of course it is right that in significantly many cases there is plenty of information around. The issue that Noam Chomsky has sometimes termed the “poverty of the stimulus” problem isn't always, or perhaps even usually, that there isn't any potential information. On the contrary, the problem is usually that there's enough information to drown in – sometimes I have even called this the “richness of the stimulus” problem. To understand word learning at all we have to get a lot more specific about how the relevance problem in word learning is solved with such laser-like accuracy by mere babes. To return to the present example, how does one know enough to ignore a bear held aloft while hugging?
18

Fig. 14.4. A change-blindness manipulation: A stuffed cat turns into a dog as it is transferred from the man to the woman.

Source
: Trueswell et al., in progress

Some useful directions of research, inspired by Gordon's work, try to extend and generalize his procedures for older children and adults by using a change-blindness paradigm. Notice in
Fig. 14.4
, which shows three temporal points
within events, that the animal changes into another at the time of interaction. Pilot findings suggest that this change is more noticeable for
giving
than for
hugging
(Trueswell et al., in progress).

More generally, observation of the reference world, while informative for word learning, seems hardly ever to be sufficient unless the category encoded is of a basic-level object (cf. Rosch 1978). In other cases, a mosaic of conspiring cues – each of them inadequate or even obfuscating by itself – from the situation and from the surrounding speech event are exploited by learners young and old to converge almost errorlessly on the lexicon of the native tongue.

Language invariance and variation
Luigi Rizzi

In this short presentation, I would like to focus on how linguists deal with the problem of invariance and variation in natural language. If you describe and compare languages, you observe that some properties are constant and other properties vary across languages. Then the question is how we can express what is universal and what are the observed patterns of variation. The theoretical entities that are used to address this issue are the concepts of Universal Grammar and particular grammars. These concepts have undergone significant development in the last twenty-five years or so. Let us briefly go through these developments. The “traditional” approach for me, the one that I studied when I first entered the field, is the Extended Standard Theory of the early and mid-seventies. The approach is really focused on the concept of particular grammar. A particular grammar is a set of precise formal rules that are related to constructions. So the particular grammar of English, for example, is a set of rules about the form of, let's say, active sentences, passive sentences, questions, imperatives, relatives, and so on. This set of rules somehow represents, in an intrinsic manner, the knowledge of the language that the speaker has intuitively. In addition to particular grammars there is a general entity, Universal Grammar (UG), which in the framework of Extended Standard Theory would be considered a kind of grammar metatheory: if a particular grammar is a theory of a language, UG is a theory of the theory of the language. So UG specified, in this way of looking at things, the format of grammatical rules–that is, what the ingredients are that you may expect to find in the rules of specific languages. And then there were certain general conditions on rule application, like Chomsky's A-over-A Principle, principles expressing empirical generalizations like Island Constraints, and so forth.

Other books

Illyrian Summer by Iris Danbury
DreamKeeper by Storm Savage
The Baby Bargain by Dallas Schulze
Command Decision by Haines, William Wister
Figment by Elizabeth Woods
Predator - Incursion by Tim Lebbon
The Lisa Series by Charles Arnold
The Wall by Carpenter, Amanda
Scarlet Devices by Delphine Dryden