Relational similarity

Page 1

What evidence suggests that children find relational judgements

difficult?

Introduction

Humans

are the only ones to possess the representational capacity to systematically

reinterpret first-order perceptual relations in terms of higher-order, rule-governed

relational structures. Relational similarity is a trademark of analogical

reasoning, which implies taking a relationship from one domain and applying it

to a different domain. Relational similarity refers to the fact that objects

compared in an analogy should be linked by the same relations. Children’s

knowledge of categories appears to differ from that of adults and dramatic

changes occur during development. Carey (1978) suggested that children start to

categorise by attending to surface features, such as shape, and, with

experience, move on to higher order similarities based on rules. Carey found

that young children might judge some men or some buildings to be “tall”, but do

not understand size as an abstract dimension of measurement, therefore the

concept is tied to specific objects.

Keil and

Batterman (1984) constructed stories about a variety of words, with both

clearly defined features and characteristic ones. The stories contained either correct

defining features, but incorrect characteristic ones, or vice versa. Young

children (4-5-year-olds) focused on characteristic features, whereas older

children focused on defining features. Nevertheless, some terms were judged on

characteristic features by children as old as eight years of age, and some

terms were judged on defining features by even the youngest children. This

suggests that the shift in representational structure does not happen for

everything at the same time. There are two proposed views concerning the

difficulties children experience when making relational judgements.

Conceptual difficulties

The

first view is that these judgements are difficult to understand because they

are not constant. This refers to conceptual difficulties. Keil and Carroll

(1980) conducted an experiment in which they showed children of ages 3 to 6 pictures

of houses and pictures of arrows that varied in height, and asked them to judge

which one was the tallest in each of the categories. The children could

recognize the tallest when looking at houses, but not when looking at arrows,

which usually appear on a horizontal axis. Therefore it appears that young children

understand tallness only with reference to things that are typically classified

as tall or short, and not as a concept.

Perceptual difficulties

The

second view refers to perceptual difficulties and assumes that such judgements

are difficult because objects vary in terms of several dimensions, and it may

be hard to ignore the irrelevant dimensions. Garner suggested that some

perceptual dimensions are easy to separate (e.g. shape, colour and size), while

others appear integral and cannot be judged independently (e.g. brightness and

saturation). In a variety of studies, Gentner and her colleagues have

investigated the development of analogical mapping within and between domains.

In one of the studies, they asked 4 year old children to find patterns that

varied either in size (within) or size and colour (between). In randomized

trials, the performance of children was at chance. However, when the cross-dimensional

comparisons were preceded by within-dimension trials, their performance

improved. This is because in the latter condition, children progressively

learned to attend to higher-order symmetry. They could then transfer this knowledge to

the cross-dimensional sets.

Language

A powerful

tool for learning about analogies is language. Ratterman and Gentner (1998)

conducted three studies with children of 3, 4, and 5 years of age in order to

find out whether children’s relational performance can be improved by the introduction

of relational labels. In the first experiment, they used a cross-mapping task

for rich and sparse objects, and took the results of the study as a baseline

level of performance for the second experiment. In experiment 2, the

researchers gave children labels such as “mummy”, “daddy” or “baby” and found a

predicted gain in performance. In the third experiment, Ratterman and Gentner

introduced relational adjectives such as “big”, “little” or “tiny” and also tested

for the long-term effects of labelling. They found that children trained with

the relational labels produced significantly more relational responses than

children in the no-label condition. These experiments illustrate how language –

and relational language in particular – can facilitate relational

representation. Ratterman and Gentner suggest that children initially align objects

based on common features, therefore complex objects with many feature matches

will be more compelling than simple ones (the richness effect). However, by 5

years of age, children are able to map relationally in spite of the salient

features of an object match. Younger children can do this if their attention is

drawn to the relevant dimension for comparison (through language).

An interesting finding is that children start to systematically and exhaustively sort groups

of similar looking objects by spatial location about a week before the “naming

spurt”, which is the sudden increase in vocabulary that children experience at

about 18 months of age. This link has been found in normally developing

children as well as in children with Down syndrome. Children who achieve

exhausting sorting also appear to be more likely to “fast map” new words to

novel objects.

Conclusion

There

are two different views with regard to the difficulties that children might

experience when making relational judgements. The first one is that there are

conceptual difficulties, suggesting that such judgements are difficult because

they are not constant (Keil & Carroll, 1980). The second view is that

relational judgements are difficult because objects vary on many dimensions and

it is hard to inhibit the irrelevant dimension (Gentner, 1996). There seems to

be a maturational (domain general) component to children’s ability to

generalise relational terms. Nevertheless, language helps children to attend to

the relevant features of objects for categorisation and generalisation, and

studies have shown that the spontaneous tendency to attend to similarity in

grouping objects develops at around the same time as the vocabulary spurt.

New Page