Andrew Radford - Linguistics An Introduction [Second Edition]

Our thought experiment is complete, and it is time to confess that the genuine experiment has been done by Eleanor Rosch and her colleagues in the 1970s with the results hinted at above. What do these results mean? It appears that the basic level is the most abstract level at which (relatively) large numbers of diagnostic (i.e. fairly reliable) properties are associated with categories. In this sense, basic categories are ‘informationally rich’ – there are many properties which will give you fairly reliable cues that you are confronted by a chair, and not a bed or a table, and this will in turn enable you to predict that the object you are concerned with has the range of properties generally associated with chairs, even if, for whatever reason, you don’t get a good look at it, say. By contrast, there are few, if any, properties that enable you to decide that something is a piece of furniture (excluding, of course, knowing already that what you have is a chair, a table or a bed). Finally, the problem of our relationship with kitchen chairs and other subordinate categories is that the vast majority of properties we associate with them will not serve to distinguish them from other varieties of chair. In short, it appears that categorisation at the basic level can be achieved with reasonable reliability on the basis of partial information, whereas this is not true at either the superordinate or the subordinate levels. It is not true at the superordinate level because there are no properties which predict category membership at this level; it is not true at the subordinate level because the predictors of category membership here are not reliable.

The suggestion that a certain level of taxonomic categorisation is informationally rich in this way leads to the provocative idea that children are somehow geared to informational richness (clusters of co-ordinated properties) in their environments. And adults ‘know’ unconsciously that this is the case. As a consequence, they provide words which small children are ‘ready’ for. Much remains to be understood in connection with this suggestion, but if it is along the right lines, it provides an illustration of how the maximisation of the informativeness of categories provides children with ready-made meanings to be matched by the

words supplied by their linguistic environment. From this perspective, absence of error is precisely what we would expect in the acquisition of early vocabulary

(exercises 7 and 8).

Exercises

1.

In section 11 we distinguished between word-based and stem-based morphology. Consider the acquisition of a language with stem-based

morphology, such as Italian or Spanish. Would you expect this process

Children and words

197

to support the hypothesis that the development of functional cate-

gories (including tense and agreement inflections) is delayed, this

delay being the consequence of the gradual unfolding of a genetically

determined programme? Are these difficulties fatal to the proposal?

2.

Devise your own small experiment to test children’s control of past

tense allomorphy. To do this, you will need to invent a number of verbs

referring to actions which can be easily depicted in drawings. For

instance, you might draw a cat balancing on its tail and have accom-

panying text along the lines of the following: ‘This is a cat who knows

how to zid. He does it most days. Yesterday he did it; yesterday he …’

The child’s task is to complete the sentence, and evidence for control

of the relevant morphological processes would be provided by a child

saying zidded.

If you have access to a small group of children, try your experiment

on them and summarise the results. As an alternative, the class can

co-operate in devising the experiment and each member of the class

for whom it is possible can run the experiment on one or more

children, with a subsequent pooling of the results.

3.

Thinking along similar lines to those you have pursued in exercise 2,

devise experiments for testing children’s control of the comparative (-er)

and superlative (-est) suffixes. Make sure that you include adjectives for

which an adult would use the more and most constructions. Again, run

your experiment if you have the opportunity.

4.

It has been noted that when children overregularise past tense mor-

phology, they are more likely to do this with certain types of irregular

verbs than with others. For example, English verbs which undergo

ablaut (see section 11) and no other process in forming their past tense (sing/sang, ring/rang, etc.) are more likely to be overregularised than

are verbs which undergo no change (hit/hit, shut/shut, etc.). When

adults are asked to supply past tense forms under time pressure, a

similar difference in the amount of overregularisation occurs. Why do

you think this might be? (Hint: think of as many no-change verbs as

you can and pay close attention to their phonological characteristics in

the light of what you know about regular past tense formation.)

5.

The two classes of irregular verbs in exercise 4 do not exhaust the full set of English verbs which have irregular past tense forms. Think of as

many irregular past tense forms as you can, and classify them in terms

of the morphological processes they involve. Test adults informally on

their ability to supply irregular forms from your various classes under

time pressure (you do this by saying that you are going to present them

198

words

with a verb and they have to produce the past tense form immediately –

give them some examples, so that they are clear on what is required).

Do you think that the results of such informal testing will generalise to

predict frequencies of overregularisation errors in child speech? Does

your answer to the question in exercise 4 help in understanding the data you have collected?

6.

In the text, we introduced the compounds geese-quietener and goose-

quietener and suggested that adults are likely to find the former

reasonably acceptable (cf. *ducks-quietener). By making up a small

set of compounds involving regular and irregular plurals, test whether

this is so. You might, for example, provide a context for each of your

compounds, and then ask adults to rate them for acceptability on a

scale of 1 to 5. Note that the important observations to make are

comparative: even if people don’t like geese-quietener much, do

they nonetheless clearly prefer it to ducks-quietener?

7.

You are to conduct a simple naming experiment with children to

ascertain whether they use superordinate-, basic- or subordinate-

level expressions to name a variety of common objects. The simplest

way to do this is to cut pictures of objects out of magazines and show

them to children with the question ‘What’s this?’ Present your results

in a systematic way.

How might you deal with the objection that the children you tested

simply did not know the superordinate and subordinate terms you

were interested in eliciting?

8.

Construct partial taxonomies for which the most abstract term is

(a) fish; (b) reptile, indicating clearly where in the taxonomy morpho-

logically complex forms appear. Now consider the claims in (i) and (ii): (i) children acquiring English acquire fish before they acquire cod,

trout, etc.

(ii) children acquiring English acquire snake, lizard, etc. before they

acquire reptile.

If (i) and (ii) are true, what are the implications of this for the view presented in the text that the child’s entry to a taxonomy is always at