THE IMPORTANCE OF COMPOUNDS IN RESEARCH

The foundations for research on dependent clusters can be found in

Gregory Murphy’s Comprehending Complex Concepts (1988). Here,

Murphy defines the complex concept as lying between the simple – that

“can be represented as a single lexical item”, and the “lexicalized (i.e.,

idiomatic) expression”. In his paper, he quotes the example of “corporate

lawyer” which is a fixed, complex adjective-noun expression. Murphy

notes that the noun-noun expression “*corporation lawyer*” is not

available for use and expressions like “corporate stationery” mean

something very different from the term “corporate”. Murphy hints at the

fact that the listener would have to know which of the specific meanings a

non-predicating term like “corporate” has and his paper can be seen as

another stepping-stone towards acceptance of fixed collocations as a

psycholinguistic notion.

Ratcliff and McKoon (1988) go much further in their research. The

hypothesis they outline is that of compound cue priming. In terms of

90

retrieval from memory, they advance the theory that it is not concept

trees (bird – animal – flight) but words that go together that make it

possible to associate:

The theory assumes that the prime and target form a compound cue and that this compound interacts with memory to produce a value of resonance, goodness of match, or familiarity that is determined by associations in long-term memory between the prime and target. If the prime and target are directly associated in memory, then the familiarity value will be larger than if they are not associated.

(Ratcliff and McKoon 1988: 405)

This would cover a range of options. The “goodness of match” would

determine in what sense “corporate” (see above) would be used if it

compounds with “lawyer” rather than with “stationery”. Likewise, the

sense of “familiarity” would find few associations for “corporation lawyer”

– “corporate lawyer” being the familiar combination. In fact, compound

cue priming highlights that the human mind very seldom retains a single

lexical item by itself in its memory. It usually is associated with another

term. This notion of association goes beyond the confines of simple

collocation. Referring to their earlier (1981) work, Ratcliff and McKoon

(1988: 389) point out that “they have shown that priming can be obtained

between concepts that are much more than four words apart.” This raises

issues, though, about collocation since it appears to contradict Sinclair’s

(1991) claim that there are no valid collocations beyond the five-word

mark either side.

91

De Mornay Davies (1998), in his work on brain-damaged patients

65

finds that they lack the knowledge (in other words, the operating

software) to use their semantic memory.

They do tend to hyperprime

66

_{, seemingly retaining most of the semantic}

information associated with target words presented:

It has often been reported for these patients that, whereas semantic representations, as assessed by off-line tasks, are degraded or inaccessible, their performance on semantic priming tasks suggests that much of the semantic information associated with these concepts is retained” (de Mornay Davies 1998: 390)

The importance of his work in this context is that he is able to

demonstrate the long-term memory function of semantic association

67

_and

its automatic retrieval:

Automatic semantic priming assumes that, on presentation of a word, the information about that word is retrieved as a result of lexical access, rather than being retrieved explicitly as a result of subjects’ responses to task demands.

(de Mornay Davies 1998: 391)

The concept of lexical access appears to be very close to lexical priming.

De Mornay Davies is more explicit when he states:

65

described by DMD (1998: 390) as "Patients with semantic memory breakdown".

66 _{Patients with semantic memory breakdown often show increased priming on semantic priming} tasks compared to normals (``hyper-priming’ ’)

92

Even if two words are not ``semantically related’ ’ in the strictest sense (i.e. they do not come from the same superordinate category), their frequent association produces a relationship at the ``meaning’’ level. (de Mornay Davies 1998: 394)

This foreshadows Hoey, saying that each term is primed to mean

something as a result of frequent association.

De Mornay Davies finds that there is still a strong drive by researchers to

try and find a meaning-driven correlation of words. However, this would

neither explain idiomatic use, nor his findings with brain-damaged

patients. There is, however, a lexical and semantic automatism:

.. activation in the lexical network could be controlled by co-occurrence frequency, such that words that often co-occur in speech or text (`collocates’ ) would be more strongly linked in a phonological or orthographic lexical network. Lexical co-occurrence, therefore, has no connection with meaning-level representations, and many researchers argue that associative priming results from lexical-level co-occurrence.

(de Mornay Davies 1998: 402)

Regrettably, he does not specify who these “many researchers” are; the

bases of his claims are the findings of his own experiments. Being more

specific than Ratcliff and McKoon, he anticipates Hoey’s later claim that

it is the property of each word to be primed to either prefer or avoid the

company of other specific words, noting that this is the case because the

mind co-associates these words, rather than because it links each

individual word to concepts or meanings. This approach to

meaning is

also noted by the pragmaticist Siobhan Chapman:

93

Many [linguists] would argue that it does not even make sense to try to discuss ‘meaning’ as a feature independent of context. The meaning of a word is entirely defined by how speakers use it in context; (…) these linguists reject the distinction between semantics and pragmatics as an unnecessary imposition on human communication. (Chapman 2006: 116)

By 2000, researchers had gathered enough evidence to conclude that

priming is an automatic process, a single process not split into stages.

Hernandez et al. (2001) confirm that -

… No evidence was found for a stage in which lexical priming is present but sentential priming is absent – a finding that is difficult to reconcile with two-stage models of lexical versus sentential priming. We conclude that sentential context operates very early in the process of word recognition, and that it can interact with lexical priming at the earliest

time window.

(Hernandez et al. 2001: 191)

There has been, too, a body of work indicating how compounds or

collocates play an important role for the human mind in the association of

lexical items. On the basis of this, a host of new experiments and research

has been undertaken in the past since the late 1990s.

In document Evidence of lexical priming in spoken Liverpool English (Page 89-93)