So What Does
'The Sign Is Arbitrary'
Actually Mean???

Margos' Magical Letter Page

copyright 1998
by Margaret Magnus
all rights reserved

Ever since de Saussure, 'the sign is arbitrary' has been one of the primary 'catchwords' of the linguistics literature. It appears in exactly that form -- the sign is arbitrary -- in most introductory linguistics textbooks, generally within the context of the observation that one cannot predict the referent of a word just by hearing it. In other words, different languages do exist, and one must learn a foreign language. One cannot understand a foreign language simply by listening to it.

Surely this is so obvious, that it hardly requires mention at all. So why is this phrase repeated again and again? Why do we require that it be internalized and recited on tests? And why do we not instead simply require students to repeat, "There are different languages, and you have to learn them."? Why do we prefer, "The sign is arbitrary."? What lies implicit in the phrase 'the sign is arbitrary' that does not lie in the phrase, 'there are different languages'?

I think 'the sign is arbitrary' is meant primarily to emphasize that you cannot predict the meaning of a word from its form. But to what extent is it meant to imply this? Does the phrase 'the sign is arbitrary' really mean that nothing about the meaning of a word can be gleaned from its structure? Surely this is not what is meant, because it is quite universally recognized that words do have structure, and one of the 4 major branches of linguistics -- morphology -- is devoted to the study of the structure of words. And of course it is assumed that these primary components -- the morphemes -- are meaning bearing, and therefore the morphological form or shape of a word does affect its meaning in some predictable way.

So what is intended by this phrase? I think when you consider the question at some depth from this angle, it becomes clear in time that what it is intended to mean specifically is that although the morphemes of which a word is composed do affect its meaning, the phonemes of which a word is composed do not. It is intended to mean that morphemes are meaning-bearing but phonemes are not, although this conclusion can in no way be draw logically from the phrase 'the sign is arbitrary'. If the sign arbitrary, then neither morphological nor phonological form should affect it. If it's not arbitrary, there's no a priori reason why both shouldn't affect it. And of course, there is a fairly substantial literature in sound symbolism that demonstrates that phonological structure does affect meaning. A root in English which begins with 'k-' is much more likely to refer to an enclosure or cover than a root that begins with anything else. This is simply a fact that has been observed time and again by many people.

Nevertheless, this sort of data is simply dismissed or rather not even considered in the bulk of the linguistics literature. Why? Perhaps this is due to the fact that although words beginning with 'gl-' are much more likely to concern reflected light in some form, these quite universal tendencies are not absolutes. Words beginning with 'gl-', though much more likely to concern reflected light do not always do so. And words beginning with 'pr-', though much more likely to concern pointed objects do not always do so. And although there is no phonological form which does not display semantic disproportions of some kind, there is also no sequence of phonemes shorter than a morpheme that displays an absolute tendency. So one must suppose that the assumption is that morpheme meanings are more absolute than phoneme meanings in this sense. But are they?

It is not the case that every word that ends in /ing/ is a present participle. And we are quite content to resolve this, by defining a present participle /ing/ morpheme and saying that whenever the final /ing/s don't mark present participles, they are not instances of this morpheme. Since this is tautological, we have not thereby demonstrated that all present participle /ing/ are instances of a morpheme. We have simply defined the morpheme as all instances of present participle /ing/. We have therefore also not shown empirically that present participial /ing/ is meaning-bearing. We have forced the morpheme /ing/ to be meaning-bearing by simply excluding all the counter-examples. Which is fine... It is useful to define a morpheme /ing/ to consist of those instances of final /ing/ that mark the present participle. But we must bear in mind that we have not thereby observed morphemes to be meaning-bearing. We have defined them to be so.

Bolinger proposed essentially that we do the same thing for shorter strings, that we define a 'gl-' submorpheme for 'reflected light', but this has not caught on. I think this is not because the term 'sub-morpheme' is less helpful than the term 'morpheme', but because we do not wish to acknowledge that the empirical facts in both cases are analogous... that morphemes are meaning-bearing because we define them as such, that we can similarly define sub-morphemes, and that since this is possible in a completely analogous manner, morphemes are not the minimal meaning-bearing units, unless we define them as such for the sake of pure, unadulterated dogma, unrelated to any empirical fact.

For even if we consider those strings which are defined as morphemes, a case can be made that we don't know what they mean. Consider, for example, the tens of thousands of latinate suffixes which exist in English. What do they mean? It is often assumed that, for example, the latinate suffix '-ate' is verbalizing. That is, its meaning is in large part 'verbalize', as in 'amalgamate' and 'oxygenate'. One generally presumes that the 'eysh' in 'amalgamation' is an allomorph of this 'ate' in 'amalgamate'. Is then the 'eysh' in 'examination' also an allomorph of 'ate', and why, then, do we have no word 'examinate', but only 'examine'? Examples like this thoroughly pervade the latinate vocabulary of English. The fact is that we don't know what '-ate' or 'eysh' means, and that morpheme meanings are not that much more predictable than the meaning of initial 'gl-' or initial 'k-'.

So why is the literature on sound symbolism largely overlooked? Surely if word meaning is in part affected by its phonological shape in most or all words, then this is a very fundamental fact of language... the data is not, in other words, dismissed because it's unimportant. Having thought about this for quite some time, I can only conclude that this data is overlooked thanks to the slogan 'the sign is arbitrary', and for no other rational reason. This phrase is posted near the beginning of every introductory linguistics book as a sort of road block. "Do not engage in a particular form of research." And it works! People without really asking themselves rationally what 'the sign is arbitrary' means, refrain from a certain type of research. It's taboo. And as far as I can tell, 'the sign is arbitrary' really doesn't actually mean much at all beyond the observation that every 2-year-old has made, that you have to learn a language... you can't know what words refer to just by hearing them. There are different languages.