Semantic Features

The identification of semantic features is an attempt to relate lexical items by decomposing meanings into features that recur in the meanings of several lexical items. These features tend to be universal categories of meaning, reflecting the nature of the universe in which we live. One such feature is concrete (features are conventionally written with capital letters), which serves to distinguish words like dog, plant and stone from words like truth, joy and ability. The latter are often called “abstract” nouns; but in feature analysis it is usual for each feature to represent a binary choice. So the first set of words would be called +concrete, and the second set – concrete. Words that have the feature +concrete may be further differentiated by the feature organic, which distinguishes dog and plant from stone and table, the former being +organic and the latter – organic. Words that have the feature +organic may be further differentiated by means of the feature animate: plant is –animate, while dog is +animate. And +animate words may be further differentiated by the feature human: dog is–human, while boy is+human. The features associated with boy are, then, +human, +animate, +organic, +concrete.

Considered like this, the features appear to be hierarchically ordered (Figure 1):

Semantic features

Figure 1. Hierarchy of semantic features

This means that a feature like human includes all those above it in the hierarchy. To specify a word as having the feature +human implies that it also has the features +animate, +organic and +concrete. But not all features of meaning that can be identified may be hierarchically ordered in this way. To further differentiate the meanings of words with the feature +human, e. g. boy, girl, man, woman, we need two further features: male and adult. Not only may these not be hierarchically ordered themselves, since boy and girl are – adult, while girl and woman are – male; but these features are also needed to differentiate words with the feature +animate which are – human, e. g. mare, stallion, foal. Incidentally, this set of words also shows that features may be marked ±, since foal would be ±male.

The implication of distinctive feature analysis, as it is called, is that the whole vocabulary may be differentiated and related by means of features. In this view each lexical item would be decomposed into its distinctive semantic features in the same way that in phonology a phoneme may be decomposed into its distinctive phonetic features, e. g. [p] into bilabial voiceless and plosive. However, it is not clear that lexical items and their meanings may be totally analyzed in this way. Take the much analyzed example bachelor. In terms of the features so far considered bachelor is presumably +human, +male and +adult. We shall now require a further feature married, which will also be required for the word spinster, and both of these will be – married. This analysis accounts for only one sense of the word bachelor, i. e. that meaning ‘an unmarried man’; it does not, forexample, account for the sense ‘holder of a first degree of a university’.

Our analysis for this one sense is, however, reasonably successful in terms of features: the meaning of bachelor is well represented by the features +human, +male, +adult, – married. But take the word table. In terms of the features already considered, this is presumably analyzable as +concrete and – organic. Beyond this, we probably need a feature furniture, and table will be marked as +furniture. After this we begin to get stuck. Is the fact that a table has a flat surface and stands on legs of criterial significance? Or should we concentrate on its use? But many items of furniture have flat surfaces and stand on legs —cabinets, desks, cupboards, bookcases. And could we define its use precisely enough? —for eating off? for working at? And would these be universal semantic features?

Some areas of vocabulary do, however, lend themselves well to a feature type of analysis. This is particularly the case with kinship terms. To distinguish the words mother, father, son, daughter, brother, sister, uncle, aunt, cousin, niece, nephew the following features are needed: male, same generation, ascending generation, collateral. [...]