Mapping words to concepts

Press the right key for the next slide (or swipe left)

also ...

Press the left key to go backwards (or swipe right)

Press n to toggle whether notes are shown (no equivalent if you don't have a keyboard)

Press m or double tap to see a menu of slides

fresh start (the shipwreck survivor)

Our question is, How do humans first come to communicate with words?

Let's make a fresh start and consider another approach.

‘children learn words through the exercise of reason’ (\citealp[p.\ 1103]{Bloom:2001ka}; see \citealp{Bloom:2000qz})

Bloom 2001, p. 1103

‘much of what goes on in word learning is establishing a correspondence between the symbols of a natural language and concepts that exist prior to, and independently of, the acquisition of that language’

Bloom 2000, p. 242

\citep[p.\ 242]{Bloom:2000qz}

‘to know the meaning of a word is to have:

1. a certain mental representation or concept

2. that is associated with a certain form’

Bloom 2000, p. 17

\citep[p.\ 17]{Bloom:2000qz}

‘Augustine describes the learning of human language as if the child came into a strange country and did not understand the language of the country; that is, as if it already had a language, only not this one. Or again: as if the child could already think, only not yet speak.’

\citep[15--16, §32]{Wittgenstein:1953mm}

(Wittgenstein 1953, p. 15--16, §32)

Does the view Wittgenstein is attacking sound like a mere caricature? Bloom explicitly endorses it, noting that

‘Augustine’s proposal is no longer seen as the goofy idea that it once was’ \citep[p.\ 61]{Bloom:2000qz}.

This view on language acquisition is not new.

Eve Clark quotes a book from 1958 which apparently suggests that children learn words by formulating and testing hypotheses about their meanings:

'The tutor names things in accordance with the semantic customs of the community. The player forms hypotheses about the categorical nature of the things named. He tests his hypotheses by trying to name new things correctly. The tutor compares the player's utterances with his own anticipations of such utterances and, in this way, checks the accuracy of fit between his own categories and those of the player. He improves the fit by correction.'

Brown (1958, p. 194) as quoted by Clark (1993, p. 19)

(Brown 1958, p. 194 as quoted by \citep[p.\ 19]{Clark:1993bv})

One consequence of this view ...

Assumption:

If someone is in a position to learn a word, she already has the corresponding concept

Consequence:

Learning words cannot ever be a route to acquiring concepts.

Case study:

Kowalski and Zimilies on colour terms and concepts.

The problem is ...

This is not enough for us to reject the Assumption. But it shows at least that the advocates of the Assumption have to explain how this case is possible and why they think there are no other such cases.

Recall this idea:

‘much of what goes on in word learning is establishing a correspondence between the symbols of a natural language and concepts that exist prior to, and independently of, the acquisition of that language’

Bloom 2000, p. 242

Incidentally, this isn't only Bloom's view, it's incredibly widespread.

Here's another example:

‘One of the first problems children take on is the MAPPING of meanings onto forms … They must identify possible meanings, isolate possible forms, and then map the meanings onto the relevant forms.’

Clark 1993, p. 14

\citep[p.\ 14]{Clark:1993bv}

I want to suggest that this idea is wrong because acquiring language is in significant part a creative process, as Eve Clark herself emphasises \citep[in][]{Clark:1993bv} .

Let me show you a particular case ...

‘puttaputta’

June, age 1;3.0

Is mapping illuminating?

Thanks to Roy Higginson's CHILDES data (1985) we can trace June's use of the novel word puttaputta over six months in fourteen conversations with her mother.

June's first recorded use of puttaputta occurs when she was fifteen months (1;3.0).

At first, her mother mistakenly takes “puttaputta” to mean Peter Piper and uses it as a noun:

JUNE: puttaputta.

MOTHER: puttaputta … ok.

MOTHER: this puttaputta?

MOTHER: Peter Piper picked a peck of pickle peppers […]

[…]

JUNE: puttaputta.

MOTHER: puttaputta?

MOTHER: where's puttaputta?

MOTHER: can you show me puttaputta?

June turns the page.

[…]

JUNE: puttaputta.

MOTHER: that's not puttaputta.

Note that the mother corrects June because at this point she mistakenly takes “puttaputta” to mean Peter Piper.

But that isn't what June has in mind, and she persists in using the term differently.

June persists in using “puttaputta” to mean something like tell me about this or read me this one and her mum quickly gets the hang of it (within the same conversation):

Roy Higginson's CHILDES data (1985)

JUNE: puttaputta.

MOTHER: puttaputta?

JUNE: puttaputta.

MOTHER: ok … Doctor .

June takes the book, looks at it and the hands it back to her mother.

MOTHER: Foster went to Gloucester in a shower of rain … he stepped into a puddle right up to his middle and never went there again.

JUNE: puttaputta.

MOTHER: ok … the late Madame Fry wore shoes a mile high and when she walked by me I thought I should die.

In this first conversation, June uses “puttaputta” 20 times (3 of these are just “putta”). Most the other 15 words she uses appear only once and none appear more than twice.

So not only do infants coin new words, they will also persist in using them despite initial misunderstandings (even despite being 'corrected' by an adult) and they may rely heavily on their own words.

Roy Higginson's CHILDES data (1985)

In a later session (June is now seventeen months (1;5.0)) June continues to use “puttaputta” and to be understood as she intends:

JUNE: putta .

JUNE: puttaputta .

MOTHER: am I supposed to read that ?

MOTHER: you have to come over here then .

JUNE: puttaputta .

MOTHER: what do you want me to puttaputta ?

MOTHER: what's this ?

JUNE: car ?

In this same session, June's mum uses “puttaputta” as a verb herself.

June continues to use “puttaputta” frequently in until around eighteen months (“puttaputta” occurs 45 times in a conversation recorded when June was 1;6.0) and then drops it abruptly (“puttaputta” doesn't appear in any of the seven conversations recorded over the next three months).

Roy Higginson's CHILDES data (1985)

Why am I telling you about June's use of puttaputta.

One thing to note, of course, is that June isn't learning to map a concept to a word.

If she is doing anything with a word-concept mapping, she is inventing and teaching it rather than learning it.

This is one ammendment to the claim that acquiring language depends on coming to know word-concept mappings.

But there is a second point I want to draw from 'puttaputta'.

June uses puttaputta purposively, to a particular end, to get others to read or interact with her in certain contexts.

But I'm not sure that this involves mapping the word to a concept.

Is mapping the word to the concept is the key to understanding what June is doing?

I'm not saying it's not; I'm just saying that there's a challenge to Bloom here.

‘children learn words through the exercise of reason’

Bloom 2001, p. 1103

‘much of what goes on in word learning is establishing a correspondence between the symbols of a natural language and concepts that exist prior to, and independently of, the acquisition of that language’

Bloom 2000, p. 242

‘to know the meaning of a word is to have:

1. a certain mental representation or concept

2. that is associated with a certain form’

Bloom 2000, p. 17

So what am I saying?

I've been considering the mapping idea, the idea that children have to establish a correspondence between words and concepts.

I agree that this the right way to think about some language learning; in particular, learning colour words like 'red' is well modelled in this way.

But I've offered two qualifications.

First, in at least some of these cases it may be that the concept comes after the word.

Second, while it's appropriate to think of abilities to use language as requiring word-concept mappings in some cases, like the case of 'red',

I used the example of 'puttaputta' to suggest that there are at least some cases of communiation with words which don't seem to involve learning or inventing word-concept mappings.

I'm not claiming to have shown that the mapping idea is wrong.

I merely want to offer you an open challenge: can we make sense of the mapping idea and do we need it?

Note that the mapping idea depends on a philosophical claim about what it is to know the meaning of a word.

This is something we should question, I think. (But not now.)