[11.39]
a. The psychology of success—sex
b. (from a politician) I like Heath. He’s tough—like Hitler. (Shocked silence from reporters)—Did I say Hitler? I
meant Churchill.
(from Ellis 1980)
In successful word retrieval, embarrassment of this kind is avoided. Only the word with the intended
meaning, which is selected by the conscious mind, is slotted into the sentence. Any other words that are
looked up are suppressed.
11.6
THE SPREADING ACTIVATION MODEL
We have established that the mental lexicon has a number of distinct modules in which different bits of
lexical information are kept. There are separate components for (i) the meaning and (ii) the phonological
representation. Further, with respect to meaning, content words are kept separate from grammatical function
words and inflectional affixes. We have seen that errors in speech processing give us an insight into the
workings of the mental lexicon.
The question we will now consider is how is all this information retrieved successfully? One possibility
could be that the speaker retrieves words going in a straight line, doing one thing at a time: (i) selecting the
meaning; (ii) selecting the inflectional morphemes and grammatical function words; and (iii) finally
mapping the morphemes making up the word on to a phonological representation. But this is unlikely to be
correct. Evidence from speech errors, like blends where semantic confusion seems to be triggered by
phonological resemblance, suggests that retrieval does not go step by step—pronunciation is not considered
in isolation from semantic choices.
A superior model that has been proposed is the SPREADING ACTIVATION MODEL (cf. Dell and
Reich (1980), Hörmann (1986), Aitchison (1987)). In this model, the starting point is the broad area of
meaning that the speaker intends to convey. Initially, the SEMANTIC FIELD targeted may be very broad
and a lot more words may be activated than the speaker actually needs. Any promising candidate is
considered before the choice is eventually narrowed to a word whose meaning
and phonological shape have
the best fit (cf. Aitchison (1987), Matthei and Roeper (1983), Marslen-Wilson (1989b)).
For instance, one may target people, then children, then one’s children, then one’s daughters. I know a
mother who, when at all agitated, goes through the names of her daughters before she eventually finds the
name of the right girl to give a good telling off. Often, if she wants to scold
Louise she will say,
‘Barbara—
eh, Jane—eh, Louise’
. She knows the child, she sees her right in front of her, but in her fury she cannot
locate the child’s name in the mental lexicon. Similarly, a student might say,
‘I haven’t finished my history—
English essay’
. Obviously, here the chosen topic of the sentence is school subjects. What is not zeroed in on
initially is the exact subject. And the first attempt to zero in on a subject is unsuccessful.
Psychologists have developed the spreading activation model to account for the way in which words are
represented in the mental lexicon. The representation they have put forward is much richer than the typical
dictionary definitions of words. They have proposed that encyclopaedic information is stored in the form of
an interlocking hierarchical network (cf. Aitchison (1987), Hörmann (1986) and Lipka (1990) and
references cited there). The dominance relation in the hierarchy is expressed in terms of ‘X is a Y’ e.g. ‘a
mammal is an animal’; ‘the cow is a mammal’, ‘the Friesian is a cow’, and ‘the Jersey is a cow’. This ‘X is
THE MENTAL LEXICON 179
a Y’ relation is called a relation of INCLUSION. So, we can say that the class of animals includes
mammals; the class of mammals includes cows and the class of cows
includes Friesians and Jerseys, etc.
In addition to inclusion, there are many other semantic relations in the network, e.g. ‘X has Y’, ‘X
contains Y’, ‘X is offspring of Y’, etc. An example of a simplified version of the representation of the
meanings of
cow and
tortoise in the mental lexicon is given in the figure on p. 257.
We can say that the nodes of the tree that have been activated are, metaphorically speaking, ‘hot’. The
greater the amount of active consideration a node receives from the speaker, the greater is its degree of
activation—hence the ‘hotter
’
it is. A hot node transfers its heat to nearby nodes that are linked to it. But a
node ‘cools’ as attention shifts elsewhere to activated nodes that are remotely, or not at all, linked to it. In
short, the zeroing in on the meaning of a word on a particular active node is a crucial step. A large number
of words are activated but few are considered actively. Then most of these are suppressed and, if all goes
well, only one word is eventually selected.
WORD FREQUENCY is a very important factor that affects the speed of retrieval. Savin (1963) showed
that it is easier to recognise frequently used words when they are presented in less than optimum conditions
where they are difficult to see or hear. Even in optimum conditions when subjects are presented with both
frequent and rare words, they find it easier to identify and classify the more frequent words (cf. Frederiksen
and Kroll 1976). You would expect that. Assuming that words are listed, those that are used very rarely
(e.g.
gazebo, zymotic) are put at the bottom of the stack and words that are used several hundred times a day
(e.g.
the, I, am) are put at the top of the stack for easy access. Frequently used words are more active; they
are hot in terms of
the spreading activation model; they are always available for use.
The spreading activation model is not one-directional, going in a straight line one step at a time. It allows
constant movement to and fro between the different sub-components. After having activated a range of
words with the desired meaning, the speaker can look up pronunciations then zip back to the meanings and
narrow the selection of words, then look up the pronunciation again. In between, the speaker may also be
sorting out the inflectional morphemes to attach to the stems selected. And remember, at any time the
speaker may peep back at aspects of the representation of a word and revise them in the light of new
information. Alternatively, the speaker may look forward to see what choices are on offer later on, if a
Do'stlaringiz bilan baham: