One of the basic assumptions of latter-day linguists is that literal, linguistic symbols are essentially arbitrary, that is, that there is no connection whatsoever between the sound of a word (the signal) and the thing it denotes or connotes (the message). Only a small number of onomatopoeic words are recognized by them as exceptions. On their 'conventionalist' view the names of things are due purely to convention and have no deeper appropriateness. Linguistic conventionalists once had to fight the so-called 'naturalist' doctrine that there would be a correct name for everything by nature. According to this doctrine words denoting movement, for instance, would actually and necessarily contain an r or l (two sounds which are acoustically similar and may be allophones of the same phoneme, albeit not in the present language). Nowadays it is said that naturalism was 'inluenced by a primitive belief in the magic properties of names'; or that in 'primitive societies' a thing was believed to be its name.

Conventionalists may conveniently have assumed that they settled the issue forever, and yet the rejection of supernaturalist 'naturalism' by no means forces us to adopt an entirely arbitrary conventionalism with respect to new words or the introduction of new morphemes, and with respect to a selective use of old ones. There may be no necessary link between the sound of a word and the thing it refers to, such does not mean that a certain linguistic symbol could not be connected with a certain thing in a fashion which is not arbitrary. It is only then that the sound of a word does not refer to a particular thing, but symbolizes it in the most appropriate way. Thus, given that certain phonemes are more stable than other ones, it is a straightforward case of analogy that --if possible, and if wanted-- words denoting or connoting stability or things which are stable, or which belong to the same associative field, should contain one or more stable phonemes (like n or m) rather than unstable ones (like s or h). Making use of such analogies is quite something else than arguing that existing words do actually contain these phonemes in a particular language as linguistic naturalists were once so audacious to profess. We shall call this new position "the symbolist position". Altho not incompatible with conventionalism, it treats the language user not merely as a consumer of conventional products but also as a creator of new products and as a selective user of old ones.

Names may not have magic properties, they do have properties, that is, whole-properties and part-properties. In the spoken language a whole-property is, for example, the place of the primary stress. The part-properties are, then, those of the individual sounds making up the name. In the present language at least one of these sounds is always a vowel, and it may also be possible to link one or more of the properties of such a vowel to certain things in the reality the total word is about. Usually the thing will not have the same property, yet it may have an attribute with the same conceptual position, or a similar attribute in a different respect.

The word symbol itself derives from symballein meaning (to) throw together. This throwing together refers to an ancient custom of breaking a coin or ring in two when friends would part for a long time or forever. Would one of them, or one of their other friends or relatives, return after many years, the two parts of the coin or ring could be compared with each other. This would give the possessor a token of identity and a right to the other person's hospitality. Now, the two halves which our own linguistic symbolism reunites are, on the one hand, the view that the names of things are due to convention, and on the other, the view that the names of things need not be completely arbitrary combinations of sounds or characters. It is in their separated conditions, when broken away from each other, that these views existed, or still exist, as conventionalism and naturalism respectively.