Let us now consider how the words of the language of experience apply to the referents as we have ordered them into spaces. Much work has been done on color terminology and its relationship to the color space. Our task is twofold: we might start with a collection of color words in use by English speakers (for instance, by asking a sample of speakers to write down all the words for colors they can think of) and then try to assign each word a region in the color space; or we might sample the color space itself by selecting two hundred colors, evenly distributed throughout the space, and then show each color to a representative sample of speakers of English and ask them to write down the English word or words that best describe that color. Let us call the first procedure Approach A and the second Approach B.

When Approach A is used, we obtain a name-map that reflects the meaning of certain words in the lexicon of a natural language. This type of information must be distinguished from the actual use that speakers make of these words. It does not completely reveal the mechanism of naming. This becomes most obvious from the fact that Approach A leaves certain parts of the color space unmapped. There will be certain colors to which none of the words collected earlier (in the absence of color samples) properly refers. The operations that lead to such results are quite simple. Subjects are presented with comprehensive color charts (Brown and Lenneberg, 1954, used the Munsell book of colors) and are asked to point to all the colors that might be called x (brown, lavender, orange, etc.). There will be some words which some subjects cannot locate at all (say, heliotrope); he may not know what the word means, or there may be wide disagreement between subjects what the proper location of a word is (say, magenta). After having given all the words in our compilation to all our subjects we will discover that there is a residue of physical colors to which no one assigned any of the names - the ‘innominate regions’.

Now, when Approach B is used the ‘ innominate regions ’ of the color space all but disappear. Although one or another subject may say ‘ I don’t know what one calls that color ’, there will be others who quite readily assign some descriptive phrase to any color shown. Approach B reflects much better the versatility or basic productivity inherent in the act of naming, while it provides at the same time some statistic on the actual usage in a given language.

Since subjects will differ among themselves in their precise naming-habit and since there will be colors that are ambiguous with respect to common names in English, colors will differ in their probability of being called x. In fact, many colors will be called by more than one name, each name having a different probability of being assigned to a given color.

I have shown experimentally (Lenneberg, 1957) that speakers of English do not only distribute their words - in this case for colors - in a characteristic and language- specific way over the stimulus continuum, but that one can also elicit from subjects correct estimates of probability that a given stimulus will be given a specific name.

If we take the color space and ask how does the word red fit into this continuum, it will appear that a certain circumscribed region of points (each point represents a different shade of color) is covered by this word. However, only some points are felt to be ‘ good reds ’; other points are more orangy reds or brownish reds. That means that the map or region of the word red has a focus or center (points that are the most typical examples of the word red) and that the borders of the map are never sharp; they fade out and are overlapped by at least one, usually several other word maps.

It is difficult to visualize the fourth dimension of probability in the three- dimensional maps of words inside the color space. But the overlapping, encompassing, and subdividing nature of the probabilistic gradients of name maps is clearly seen if one takes out of the color space a linear array of colors (for example, stimuli that vary only in hue, holding brightness and saturation constant) so as to produce a unidimensional subsample of the color space, and then maps the vocabulary on to this continuum as shown in Fig. 3. The procedure for obtaining this map was that of Approach B. Notice that there are certain colors which all subjects call by essentially the same name whereas other colors are sometimes called by this and sometimes by that name. Colors nos. 4 and 13 in Fig. 3 were given so many individual names that it was not possible to represent them all by a distinct curve.

Thus, when words are mapped into a referent space, a probabilistic variable emerges that we might call name-determinacy. The loci in the referent space vary with regard to the name-determinacy. The determinacy is greatest in the center of the foci and decreases centrifugally from these points; it is lowest at the intercept of name maps. The region which appeared to be ‘ innominate ’ under Approach A is one of great and widespread name-indeterminacy. Name-determinacy is, of course, a social phenomenon.

The procedures outlined here are a convenient way of demonstrating some of the semantic differences between languages. For instance, color name maps have been worked out for Zuni (Lenneberg and Roberts, 1956) and partially for Navaho (Landar et ah, i960), and Conklin (1955) adopted the Lenneberg-Roberts technique for the study of Hanunoo color categories.

This type of empirical research on semantic relationships in various languages can only be carried out within a very small part of the total vocabulary of a language, namely only on the language of experience. Therefore, it is not possible to verify empirically the far-reaching claims that have been made about the ‘total incomparability of semantics ’ among unrelated languages. The languages investigated all have names for colors, and apart from finding mergers of two of our English color maps into one, or the substructuring of one of our maps, or shifts in border lines, nothing particularly astonishing has been discovered. We gather that man, everywhere, can make reference to colors and communicate about color fairly efficiently.