shows the importance of selection. The listener constructs meaning from a specific set of auditory signals —
and ignores the rest.
Psychologists have performed experiments to explore the nature of selective listening. They have done
this by designing SHADOWING experiments where subjects repeat instantaneously a taped utterance, word
for word, as they listen to it. Meanwhile, the subjects listen simultaneously, in one ear, to another message
which they are told to ignore. The experiment is set up in such a way that the two messages that are listened
to are equally clear and have the same tempo and volume.
In these experiments people easily manage to focus on the utterance that they are shadowing and to
ignore the utterance that they are not shadowing. They might not even notice that the utterance which they are
not shadowing is in a foreign language, or whether the speaker was a man or a woman (Cherry 1953).
11.2.3
Exploiting syntactic and semantic clues
Do you have a friend or acquaintance who has that most infuriating habit of finishing your sentences for
you? Both the syntactic and semantic aspects of the utterance may be so predictable that the listener knows
what you are going to say even before the words come out:
[11.12]
A.
Do you know what? She has threatened to…
B.
…sue the police.
The syntactic semantic context plays a role in speech comprehension. Listeners use frames provided by
the syntax to retrieve a word with the appropriate meaning from the mental lexicon. This is easily seen in
the handling of ANAPHORIC EXPRESSIONS (i.e. grammatical elements that refer back to something
already mentioned) which are only interpretable in a specific syntactic context. To interpret anaphoric
expressions the hearer must be able to identify the element that has already been mentioned which is being
referred back to. Imagine reading the following sentence in a recipe book:
[11.13]
Add cream to
the meat casserole and leave it in the oven for 10 minutes.
How do you work out what
it refers to? In this sentence
it could conceivably refer to the
meat or the
cream
. But you do not even consider the
cream. This sentence is not ambiguous. Our knowledge of the
world rules out the
cream. It would be crazy to leave the cream cooking in the oven. But it is reasonable to
leave
meat cooking in a casserole in the oven.
A similar point can be made with regard to STRUCTURAL AMBIGUITY (i.e. situations where a string
of words has different interpretations depending on how we group together the words). Take this example:
[11.14]
I bought some new shirts and jumpers.
THE MENTAL LEXICON 167
The sentence can be paraphrased as ‘I bought some shirts and jumpers all of which were new’, in which
case words are bracketed as in [11.15a] or, ‘I bought some new shirts and some jumpers which may not be
new’. This latter interpretation is reflected in the parsing:
[11.15]
a.
[some new shirts and jumpers]
NP
b.
[some new shirts]
NP
and [jumpers]
NP
Understanding speech is impossible without the listener working out which words go together. Correct
parsing is a prerequisite to comprehension.
Sometimes, parsing is complicated. The listener makes a provisional analysis which has to be revised
immediately as more information becomes available. A key feature of the speech comprehension process is
the way in which listeners constantly update and revise their putative analyses. Nowhere is this clearer than
in the analysis of so-called GARDEN PATH SENTENCES. These sentences metaphorically lead you down
the garden path. You start doing a parse that looks plausible, but it turns out to be flawed when you get
more information:
[11.16]
a.
After
taking the right turn
at the lights, we rejoined the
highway.
b.
After taking the
right turn at the lights, we rejoined the
highway, but soon we realised
that it was the wrong
turn.
The sentence in [11.16a] is ambiguous.
Right turn may mean ‘right-hand turn’ or ‘correct turn’. You need
more information to resolve the ambiguity. If the sentence continued as in [11.16b], you would realise that
right
is to be interpreted as ‘right-hand’ rather than ‘correct’.
How much of a sentence does a listener need to hear in order to be able to make intelligent guesses? Not
a lot. I suspect that as soon as you hear
‘but’ you immediately suspect that
right does not mean ‘correct’
here. Everything that comes after that confirms your guess.
The point is driven home by this famous example of a garden path:
[11.17]
The horse raced past the barn fell.
This sentence is as clumsy as it is bizarre. Working out its meaning requires a number of attempts at
syntactic analysis and semantic interpretation.
The horse raced past the barn would be a problem-free
sentence. This is the interpretation that initially springs to mind. But the verb
fell at the end jolts us. It looks
as if it should go with
the barn fell. But that would make a nonsense of the whole sentence. So, we go back
to the drawing board. This time we interpret
The horse raced past the barn as a reduced relative clause
which we paraphrase as
The horse that was raced past the barn fell. And it works— which goes to show
that syntactic parsing is done and updated continually by the listener in the light of inferences that make
sense. Speech comprehension is not simply a matter of associating sounds with the meanings of words.
These speech comprehension findings are not easy to reconcile with the position adopted in (11.1.3),
which otherwise seems justified, that we wheel out the morphological parser quite frequently to deal with
168 ENGLISH WORDS
complex words and that they are usually not stored as pre-packaged units for the simple reason that there is
insufficient space in the mind to store all the word-forms we know. But the speed of processing argues for
listing in the mind pre-packaged units which are retrieved as soon as clues to their identity are obtained.
Although my inclination is to go for morphological parsing rather than listing, it not absolutely clear that it
is the solution. More research is needed to resolve the issue fully. The jury is still out on this one.
Do'stlaringiz bilan baham: 
