Andrew Radford - Linguistics An Introduction [Second Edition]

210

words

finding. The lexical entries for robin, sparrow and eagle, being ‘close’ to that for bird, receive a large amount of activation and are thereby primed (see above) and produced on subjects’ lists. Lexical entries for other, more remote, bird names are primed to a lesser extent or not at all.

Another very direct approach to this topic is to ask subjects to rate pairs of words for semantic similarity using, say, a five-point scale. Thus, you might be presented with sparrow and eagle and if you feel that they are very similar

semantically, you will score them at five, if you consider them not semantically similar at all, you will give them one, and if you perceive a middling amount of semantic similarity, you will use one of the intermediate numbers. It comes as a surprise to many people that a technique as simple as this produces reliable results across large populations of subjects. As far as our current interest goes, the important finding coming out of such experiments is that pairs such as robin and bird receive significantly higher scores than pairs such as ostrich and bird.

Again, this is consistent with the lexical representation for robin being ‘closer’

than that of ostrich to that of bird in psychological space, a conclusion that is not captured by supposing that lexical organisation in this area is merely taxonomic.

Finally, another twist to this story emerges from an experiment conducted by Lance Rips and his colleagues. In this study, subjects were asked to imagine a small remote island populated entirely by various species of birds and were told that all members of one species (e.g. the owls) had been infected with a particularly virulent (for birds) disease. The subjects’ task was to judge what proportion of other species succumbed to the disease. In support of what we have seen above, it was found that if the initially infected species was prototypical (e.g. robins), then greater proportions of other species were judged to contract the disease than if the initially infected species was not prototypical (e.g. ducks). Putting this crudely, if the robins started it, more sparrows, eagles, owls, etc. were judged to get the disease than if the ducks started it. Intriguingly, this result applies to specific pairs of birds. For instance, if the disease starts with robins, maybe 60 per cent of ducks are judged to get it; however, if it starts with ducks, only 40 per cent of robins fall ill. This is a rather different result to those which show that the lexical representation of robin is relatively ‘close’ to that of bird. What this seems to show is that the

‘distance’ from the lexical representation of robin to that of duck is smaller than the distance from the lexical representation of duck to that of robin, i.e. ‘distance’

in the mental lexicon might not even be symmetrical!

In general, we can conclude that lexical processing is an extremely rapid and efficient cognitive process, and psycholinguists have only just begun to develop appropriate theoretical models for understanding this process. Additionally, the organisation of the mental lexicon, while broadly in line with the ideas that linguists have developed, appears to have some rather unusual properties. Most importantly, while psycholinguists often appeal to non-linguistic notions such as memory and frequency in their studies, the proposals made by linguists on such issues as semantic similarity, categorisation and lexical representation regularly provide the basis for modelling (exercise 6).

Lexical processing and the mental lexicon

211

Exercises

1.

Obusek and Warren (1973) presented subjects with samples of speech such as those below from which individual phonemes were deleted

or replaced with a cough (the spelling we have used reflects the

missing word, but, of course, this is not available to subjects in an

aural presentation experiment). Participants often failed to detect that a

phoneme was missing and supplied appropriate phonemes in the dif-

ferent contexts. Discuss these phoneme restoration effects in the light of

the controversy between serial-autonomous and parallel-interactive

models of word recognition.

(a) It was found that the *eel was on the orange

(peel)

(b) It was found that the *eel was on the axle

(wheel)

(c) It was found that the *eel was on the fishing-rod

(reel)

(d) It was found that the *eal was on the table

(meal)

2.

In a lexical priming experiment (Carreiras, Duñabeitia and Perea

2007), where subjects had to recognise targets as words or non-words, target words (e.g. the word MATERIAL in the examples below) that

were immediately preceded by prime words containing digits (a below)

or symbols (b below) were responded to as quickly as when they were

preceded by an identity prime (c below), and responses under all three

conditions were much faster than when the target word was preceded by

an unrelated control word (d below).

Primes

Target

(a) M4T3R14l

MATERIAL

(b) MΔT€R!ΔL

MATERIAL

(c) MATERIAL

MATERIAL

(d) CORPORAL

MATERIAL

What do these findings tell us about word recognition during reading?

Do we really recognise words from left to right by identifying one letter

after another? How is it possible that NUM83R5 ΔND $YMβ0L$ C4N

B€ U$3D Δ$ L3††3R$ !N 4 $3N7€NC€ ΔND †H3 R3$UL7!NG

$3N7$NC$ C4N B$ UND3R$ †00D?

3.

With respect to the processing of lexically ambiguous words, at least

two possibilities must be considered: (A) On encountering an ambig-

uous word, only one of its possible meanings is accessed; or (B) all

meanings are accessed initially, and all but the ‘correct’ one are

discarded later. What do the findings below tell us about the way we

process lexically ambiguous words?

(a) McKay (1966) found that in a sentence-completion task, subjects would take longer to complete (i) than (ii).

212

words

(i) After taking the right turn at the intersection, I …

(ii) After taking the left turn at the intersection, I …

(b) In a cross-modal lexical decision experiment, Swinney (1979) found priming effects for all meanings of the word bug (not just

the one that fitted the context). Specifically, in (iii) below, he found that recognition of both ant and spy as words was facilitated

at the point marked *.

(iii) Rumour had it that, for years, the government building had

been plagued with problems. The man was not surprised

when he found several bugs * in the corner of his room.

4.

Below are some examples of picture-naming errors typical of a type of

aphasia called semantic anomia:

Patient’s error

Target word

(a) buffalo

lion

(b) hair

comb

(c) wash

towel

(d) sugar

coffee

(e) sink

desk

Describe the nature of the above errors. What do cases of anomia tell us

about the way words and concepts are represented in the mind/brain?

5.

Analyse the following speech errors. Which (if any) are problematic

for serial-autonomous models of speech production, and why?

(a) week at workends

(work at weekends)

(b) I’d hear one if I knew it

(I’d know one if I heard it)

(c) I hate raining on a hitchy day (I hate hitching on a rainy day)

(d) blond eyes

(blond hair)

(e) I’m making the kettle on

(making some tea / putting the

kettle on)

(f) I’ve eaten all my library books (I’ve read all my library books)

(g) It’s difficult to valify

(validate / verify)

6.

You are to conduct a small experiment to investigate the prototype