Swiss linguist Ferdinand de Saussure (1857–1913) devised a model of linguistic meaning involving what he called the signifier (a symbolic or phonological form) and what it signifies. Their association is a basic unit of communication he referred to as a linguistic sign, and it is fundamentally arbitrary.
For example, rose signifies a flower with a pleasant smell, but by any other name it would, per Romeo, smell as sweet. Generally speaking, the meaning of a word cannot be predicted from its form, nor its form from its meaning.
Saussure also drew a useful distinction between two approaches to linguistic study, which he called diachronic and synchronic – essentially historical and ahistorical. How he knitted these concepts together may be seen in this passage by Jonathan Culler in his book Saussure (Fontana Modern Masters, 1976):
What is the connection between the arbitrary nature of the sign and the profoundly historical nature of language? We can put it this way: if there were some essential or natural connection between signifier and signified, then the sign would have an essential core which would be unaffected by time or which at least would resist change. This unchanging essence could be opposed to those ‘accidental’ features which did alter from one period to another. But in fact, as we have seen, there is no aspect of the sign which is a necessary property and which therefore lies outside time. Any aspect of sound or meaning can alter; the history of languages is full of radical evolutionary alterations of both sound and meaning. . . . In short, neither signifier nor signified contains any essential core which time cannot touch. Because it is arbitrary, the sign is totally subject to history, and the combination at a particular moment of a given signifier and signified is a contingent result of the historical process.
The fact that the sign is arbitrary or wholly contingent makes it subject to history but also means that signs require an ahistorical analysis. This is not as paradoxical as it might seem. Since the sign has no necessary core which must persist, it must be defined as a relational entity, in its relations to other signs. And the relevant relations are those which obtain at a particular time.
There are exceptions to the arbitrary nature of the sign, such as onomatopoeia or sound symbolism, but even these may have aspects that are arbitrary or informed by the cultures in which they exist. And they are greatly outnumbered by the arbitrary signs.
John Lyons notes in Language and Linguistics that this arbitrary quality makes languages more difficult to learn, but it also gives them great flexibility and adaptability.