Language in humans Today: how do humans process language? - - PDF document

1

19/02/09 Susen Rabold 1

Language in Humans

Human Communication 1 Lecture 16

19/02/09 Susen Rabold 2

Language in humans

Today:

• how do humans process language?

We’ve looked above at syntactic processing. There are many other aspects of apparently similar complexity.

• The main questions are

– the overall organization of the processing, and – the extent to which we can observe modularity. knowledge have to interact.

19/02/09 Susen Rabold 3

Further points

We’ll see

• how we can use ambiguity to probe

certain aspects of the process and

• how different sources of knowledge

have to interact.

19/02/09 Susen Rabold 4

A very simple model

• of the processing of speech:

This is highly modular:

• process pressure waves to produce “sounds”
• process sounds to produce words
• process words to produce trees
• process trees to produce DRSs

We have already seen that it’s too simple in one respect: there are too many interpretations because ambiguities multiply.

2

19/02/09 Susen Rabold 5

An experiment (a)

• Do we retrieve all possible meanings

(and syntactic categories) for the words that we hear? CROSS-MODAL priming allows us to investigate this question. Present the following aurally:

• The gypsy read the man’s palm (1) (2)

for only a dollar because he was broke.

19/02/09 Susen Rabold 6

An experiment (b)

• Imagine you are conducting an

experiment and ask the participants whether they see a word of English - at times (1) and (2) present the words hand, tree and other unrelated words visually - and note their reaction time (RT).

19/02/09 Susen Rabold 7

An experiment (c)

• Your results would be:

(1) RT for related words (e.g. hand and tree ) is the same, and quicker than for unrelated words. (2) RT for hand is quicker than RT for tree and unrelated words. The same result is obtained for words of differing syntactic categories (e.g. tyres : wheels vs. wears out ).

19/02/09 Susen Rabold 8

Conclusions of experiment

• We process the sounds to compute all

possible syntactic categories and meanings of those words. These are then filtered (within a matter of tenths of seconds) according to what is syntactically and contextually

• appropriate. Note that this involves the

interaction of different kinds of information.

3

19/02/09 Susen Rabold 9

And the moral is

Morals:

• just because we can’t introspect and

“feel” the multiple interpretations, that doesn’t mean they’re not there.

• at least some part of the human

language processor operates in parallel.

19/02/09 Susen Rabold 10

Shadowing (a)

• We process language fast! Some can

listen to a spoken text and repeat it back with a delay of less than .5s.

• They correct errors (e.g. cigaresh for

“cigarette”) as they go. Try it yourself.

19/02/09 Susen Rabold 11

Shadowing (b)

Choices:

• opt for a serial syntactic model which always

makes a choice

• weaken modularity and allow some

interaction between processes. One version of the second is to allow semantic and discourse information to rule out

• analyses. This leaves us with a paradox. How

can we know what words to retrieve if we correct words as we go along?

19/02/09 Susen Rabold 12

Serial vs. parallel (a)

• We’ve seen some evidence of parallelism at

the level of words. What about other aspects

• f the process?
• Garden path sentences provide some

interesting evidence:

– The man who hunts ducks out on weekends – The cotton shirts are made from grows in Mississippi – The old train the young – The daughter of the king’s son loves himself

4

19/02/09 Susen Rabold 13

A definition of Garden Path

• Garden Path sentences = Sentences

that lead the human sentence processor (HSP) to construct an initial syntactic structure, which turns out to be incorrect and thus requires syntactic (and semantic) re-analysis.

19/02/09 Susen Rabold 14

Serial vs. parallel (b)

• Introspection suggests that we “get stuck”.
• We could model this by our first choice

above: only ever work on one analysis, and have a rule for deciding which syntactic rule to use.

• There is no or only limited possibility of

revision of a choice.

19/02/09 Susen Rabold 15

Serial vs. parallel (c)

• On the other hand, there are some

sentences in which we seem to be able to avoid having to make an early choice:

– Have the police . . . eaten their supper? – come in and look around. – taken out and shot.

19/02/09 Susen Rabold 16

Serial vs. parallel (d)

• We can replace the . . . with any

amount of material. But that doesn’t seem to induce the same kind of hiccup as garden path sentences.

• One possible conclusion from this

example is that humans can make some use of ambiguity in processing.

5

19/02/09 Susen Rabold 17

Looking at cards

• Experiment: ask people to move playing

cards around a table.

• For example: 4♠ 6♣ 6♥ 3♦ K♣
• Instruction: “move the six of clubs to

beneath the three of diamonds (1).”

– (1) is space where card is being moved to

19/02/09 Susen Rabold 18

Eye tracking your participants

• Use an eye tracker to work out where their

attention is.

• Results: by the time, (1) is reached, the

subject’s gaze alights on the 3♦.

• Conclusion: as with shadowing, people can

allow their knowledge of the context to allow early processing of semantic and discourse

• information. In this case, the uniqueness of

the 3 allows the subject to work out which card is being referred to.

19/02/09 Susen Rabold 19

Eye tracking Solitaire

• http://www.youtube.com/watch?v=XwM
• AqgikRM&feature=related

19/02/09 Susen Rabold 20

Summary

• The mechanisms by which humans process

speech are still the subject of great controversy. We can say:

– they involve some parallelism, certainly at the level

• f words, and perhaps at the level of syntactic

analyses – they don’t conform to the strictest version of modularity – and there can be relatively large-scale interaction between different sources of knowledge.