Imageability

There is little evidence that imagery effects lexical access. However, since there could be an interaction between the visual channels and imagery (argued for below) this was thought to be a possibility worth exploring even against doubts (creating the required factoring of the stimuli with

imageability was not that difficult and so was judged worth including even though there was a low probability it might yield interesting findings).

Some words make reference to visually identifiable objects or properties like BLUE or HOTEL. They are

imageable. Other words, in contrast, do not. They are not readily visualised and are abstract. examples are TRUTH and HOT. The property of imageability is closely related with that of concreteness: the degree to which a word's reference has a concrete referent. The difference between them can be illustrated by the words BLUE and HOTEL both of which are high imageability words though only HOTEL is a highly

concrete one. Concreteness and imageability values for words are closely correlated — r = .78 (Gilhooly & Logie, 1980). In some of the following discussion I make the assumption that the high correlation between concreteness and

imageability enables research upon one to a limited extent to be generalised to the other.

As noted, imageability is of interest since like word frequency it might tap lexical processes. In addition, it is conceivable that the lexical processes of highly imageable words may have a special relationship with the parvo

channel. Though conceivable it is however most likely that imageable and abstract words are processed the same way with any differences between them occurring postlexically. It is however possible that they may be processed differently.

The reason for raising this possibility is that the lexical processes responsible for recognising concrete and abstract words appear from two case studies to be

neurologically doubly dissociable with the reported

existence of lesions producing deficits in reading low but not high concrete words (Shallice & Warrington, 1975) and conversely lesions producing defects affecting the reading of high concrete words more than low ones (Warrington,

1981). One interpretation of this double dissociation is that the lexicons responsible for representing imageable and abstract words are differently or separately stored or

connected with the semantic system. This suggests that there may exist different processes or connections underlying the

identification of concrete and abstract words.

There are a number of possibilities for the nature of these processes or connections. One candidate (other

interpretations are possible, indeed they are more likely) for the difference between them could derived from the

parallel existence to the reading lexicon of an image lexicon (Paivio, 1971, 1986) used to recognise visual

objects and properties. The existence of this extra lexicon could change (one way is described below) the processing of imageable words making their processing different from

abstract ones.

Kroll and Merves (1986) suggest one possibility. In suggesting an explanation for the reaction time advantage in lexical decision for high concrete words over frequency and length matched abstract words (Kroll & Merves, 1986; James, 1975), Kroll and Merves (1986) hypothesised that

'concreteness effects might arise as a function of

prelexical factors that influence earlier stages of lexical retrieval' page 100. One possible influence upon this

prelexical stage might lie in the visual channels. As mentioned in chapter 3, the main function of the parvo channel is the visual recognition of objects, something which involves it with a special relationship to the image lexicon. In contrast, the functions of the magno channel primarily involve aspects of vision other than image recognition and so it would not be expected to have a

special relationship with the image lexicon. The existence of a special relationship between the recognition of

imageable words and images and special relationship between the parvo channel and image processing could result in a third between between the parvo channel and imageable words.

If this was the case then an interaction might appear between the 1 cha n n e l ' technique of presentation and the

imageability/ abstractness effect upon lexical decision. For instance, if there was a special relationship between the parvo channel and imageable words then a visual presentation which blocked parvo channel perception might affect

imageable words more than abstract ones.

Failure to find this interaction however would not imply 'parvo1 stimuli do not affect the reading process. Concreteness/imageability correlates with other variables such as age of acquisition (Gilhooly & Logie, 1980) and ease of predication (Jones, 1985). Moreover, effects attributed to concreteness/imageability have been attributed to age of acquisition (Coltheart, Laxon, and Keating, 1988) and ease of predication (Jones, 1985). Therefore the

concrete/imageable word advantage in lexical decision could be the product of processes unrelated to the conjectured special relationship between imageable words and the image lexicon with this advantage instead deriving from the

effects of age of acquisition or ease of predication.

5.2.5. Letter length

Letter length affects lexical decision reaction times with no or slight length effects for words but strong ones

Figure 5.2. Illustration of Ellis, Young and Anderson's (1988) model of processes which give rise to length effects. With

alternations from their illustration on page 269.

LEXICAL SUBLEXICAL (SEMANTIC) PROCESSING PROCESSING A VISUAL INPUT LEXICON

I

I

are several ideas why this might be the case. The letter length effect for wordlike nonwords has been attributed by Seymour (1987a; 1987b) to their being parsed into-sublexical units, a process not required for words. Ellis, Young and Anderson (1988) have suggested, not incompatibly with

Seymour, that nonwords are buffered (by a process they refer to as mode B) in an abstract letter storage. Words in this account however do not need storage and thus are directly inputted by a quick and parallel process they call mode A

(see illustration). Though there might be length effects for this process they are slight. Why should nonwords be parsed or buffered in abstract letter storage? One possibility is that words and nonwords are processed differently in lexical decision with word responses originating through a process