[251]

Appendices to Parts Two and Three

A Subjective and objective analysis

The analysis of linguistic units, constantly being made by the speakers of a language, may be called subjective analysis. It must not be confused with objective analysis, based on history. In a form such as Greek híppos ‘horse’, the grammarian distinguishes three elements: a root, a suffix and an ending (hípp-o-s). Greeks themselves recognised only two (hípp-os: cf. p. [213]). Objective analysis distinguishes four smaller units in Latin amābās ‘you were loving’ (am-ā-bā-s): but speakers of Latin segmented the form as amā-bā-s, or probably even treated -bās as a single flexion, as distinct from the stem. In the French words entier ‘entire’ (Latin in-teger ‘intact’), enfant ‘child’ (Latin in-fans ‘not speaking’), enceinte ‘pregnant’ (Latin in-cincta ‘un-girdled’), the historian will detect a prefix en-, identical with the privative in- of Latin. But the subjective analysis of French speakers fails to recognise it altogether.

The grammarian is often tempted to see mistakes in these spontaneous linguistic analyses. But subjective analysis is no more in error than is ‘false’ analogy (cf. p. [223]). The language does not make mistakes. Its point of view is a different one, that is all. There is no common measure between the analysis of speakers and the analysis of historians, even though both proceed in the same way, i.e. by correlating series containing a common element. Both can be justified. Each has its own value. But in the last resort the only one that matters is the speakers’, for it is based directly on the facts of linguistic structure.

[252]

Historical analysis is only a derivative form of analysis. In the end, it comes down to a projection of constructions taken from different periods on to a single plane. Like spontaneous segmentation, its aim is to identify the smaller units which go to make up a word. But it brings together the various segmentations made over a period of time, with a view to determining the oldest of these. A word is like a house of which the internal arrangement and purpose have been changed on various occasions. None the less, for those who live in it, there is only ever one. The analysis hípp-o-s examined above is not false, for it reflects facts that speakers were at some time aware of: it is simpiy ‘anachronic’, relating to a different period from that from which the word is taken. This hípp-o-s is not in contradiction with the hípp-os of Classical Greek; but it must be judged differently. Which once more comes back to the intrinsic difference between the diachronic and the synchronic.

Moreover, this allows us to deal with a methodological question still unresolved in linguistics. The comparative philologists divided words into roots, themes, suffixes, etc. and treated these distinctions as having an absolute value. To read Bopp and his disciples, one would think that the Greeks had brought with them from time immemorial a whole outfit of roots and suffixes, and that they put their words together as they went along; so that patḗr, for example, was as far as they were concerned a root pa plus a suffix ter, and dṓsō was the sum total of dō+so+ a personal ending.

It was inevitable that there should come a reaction against these aberrations, and the watchword became – quite rightly – ‘look at what happens in present-day languages, in everyday speech, and do not attribute to earlier periods any process or phenomenon which cannot be observed nowadays’. And since very frequently it is not possible to foist on modern languages analyses of the kind Bopp made, the Neo-grammarians declare in accordance with this principle that roots, themes, suffixes, etc. are pure abstractions of the mind: if we make use of these abstractions, it is solely for convenience of exposition. But if there is no justification in the establishment of these categories, why establish them? And when they are established, on what ground is it claimed that a segmentation like hípp-o-s, for example, is prefertable to hípp-os?

[253]

The new school, after pointing out the defects of the old doctrine – which was easily done – was content with rejecting it in theory, while in practice remaining encumbered with a scientific apparatus which it could not dispense with after all. As soon as one examines these ‘abstractions’ rationally, one sees to what extent they do correspond to reality, and a very simple qualification is all that is needed to give these grammarian’s devices an interpretation which is both valid and precise. An attempt has been made to do this above, by showing that objective analysis, being linked internally to subjective analysis of the living language, has a legitimate and clearly defined place in linguistic method.

B Subjective analysis and determination of units smaller than the word

As regards linguistic analysis, one cannot establish a method or formulate definitions except by approaching the task synchronically. This we should now like to demonstrate by commenting on the parts of the word: prefixes, roots, stems, suffixes, endings.¹

Let us first consider endings: that is to say, the characteristic flexions or variable elements at the end of a word which distinguish the forms of a noun or verb paradigm. In the Greek verb ‘to harness’, zeúgnū-mi, zeúgnū-s, zeúgnū-si, zeúgnu-men, etc., the endings -mi, -s, -si, etc. are delimited simply by contrast with one another and with the preceding part of the word . It has already been noted (pp. [123] and [163]) in connexion with the Czech genitive žen, as opposed to the nominative žena, that the absence of an ending may play the same role as an ending itself does. Thus in Greek the singular zeúgnū! (‘harness!’) contrasts with the plural zeúgnu-te! (‘harness!’), and the vocative rhêtor! (‘O speaker!’) contrasts with rhḗtor-os, etc. In French marš (written marche! ‘march!’) contrasts with maršō (written marchons! ‘let us march!’). These are examples of inflected forms with a zero ending.

By discarding the ending, one obtains the theme or stem. This is, generally speaking, the common element which can be identified spontaneously by comparison of a series of related words, whether inflected or not, and which carries the idea common to the whole series. Thus in French in the series roulis (‘rolling’), rouleau (‘roller’), rouler (‘to roll’), roulage (‘rolling’), roulement (‘rolling’), it is easy to identify a stem roul-. But the analysis made by speakers often distinguishes stems of various kinds within the same family of words; or more exactly, stems of various grades. The element -, identified above in zeúgnū-mi, zeúgnū-s, etc., is a first-grade stem. It is not unanalysable: for if one compares other series (zeúgnūmi (‘I harness’), zeuktós (‘yoked’), zeûksis (‘a joining’), zeuktḗ r (‘one who harnesses’), zugón (‘yoke’), etc. vs. zeúgnūmi (‘I harness’), deíknūmi (‘I show’), órnūmi (‘I stir’), etc.) the segmentation zeug-nu becomes obvious. So zeug- (with its variants zeug-, zeuk-, zug-, cf. p. [220]) is a second-grade stem. This stem is irreducible, for by comparing related forms it is impossible to take its segmentation any further.

[255]

This irreducible element common to all the words of one family is called a root. Since any subjective synchronic segmentation can only separate out different material elements by attaching some fragment of meaning to each, the root is in this respect the element where the meaning common to the whole series of related words reaches a maximum of abstraction and generality. Naturally, this vagueness varies from one root to another, but it also depends to some extent upon the grade of the stem in question. The more segmentations involved, the more abstract the residual meaning is likely to be. Thus zeugmátion means ‘little yoke’, zeûgma ‘yoke, band’ (of no special kind), and zeug- expresses the general idea of ‘joining together’.

It follows that a root, as such, cannot constitute a word and have endings directly added to it. A word always represents a relatively specific idea, at least from a grammatical point of view, and this is incompatible with the generality and abstractness characteristic of roots. But what then is to be made of the type of case which arises very frequently, where root and stem appear to be identical? For example, the Greek word for ‘flame’ phlóks (genitive phlogós) has a root phleg- or phlog-, which is found in all words of this family (phlég-ō ‘I burn’, etc.). Is this not counterevidence to the distinction laid down above? No; for we must here distinguish the root phleg- or phlog- in its general sense from phlog- with a particular meaning. Otherwise we shall be considering just the material form and ignoring the meaning. In this case the same phonetic element has two different values, and thus constitutes two distinct linguistic elements (cf. p. [147]). Just as previously zeúgnū! (‘harness!’) was seen to be an inflected form with a zero ending, we can say that phlóg- (‘flame’) is a stem with a zero suffix. Thus any confusion is avoided: the stem is kept distinct from the root, even if the two are phonetically identical.

[256]

The root, then, is a reality which speakers of a language recognise. They do not always, it is true, identify it equally clearly in all cases. There are differences in this respect from one language to another, as well as within individual languages.

In certain languages, specific features make speakers aware of roots. This is the case in German, where the root tends to take a regular form. It is nearly always monosyllabic (e.g. streit-, bind-, haft-, etc.), and conforms to certain rules of structure. There are restrictions on the order of sounds. Certain combinations, such as stop + liquid, are not allowed in final position. Thus werk- is a possible root, but not wekr-. Similarly helf- and werd- are found, but not hefl- or wedr-.

It will be recalled that regular alternations, especially between vowels, reinforce much more than they weaken a feeling for roots and smaller units in general. In this respect too, German with its Ablaut variations (cf. p. [217]) differs profoundly from French. Semitic roots have similar characteristics to an even higher degree. There the alternations are very regular, and form the basis of a large number of complex oppositions (e.g. Hebrew qāt.al, qt.altem, qt.ōl, qit.lū, etc., – all forms of the same verb ‘to kill’). Moreover, they show a feature reminiscent of the monosyllabic character of roots in German, but more striking; namely, they invariably have three consonants (cf. p. [315] ff.).

From this point of view, French is quite different. It has few alternations, and many two- or three-syllable roots, such as commenc-, hésit-, épouvant-, in addition to monosyllabic ones like roul-, march-, and mang-. Furthermore, French roots are too diverse in the combinations of sounds they allow, especially in final position, to be reduced to rules (cf. tu-er, régn-er, guid-er, grond-er, souffl-er, tard-er, entr-er, hurl-er). Thus it is not surprising that in French a feeling for roots is relatively weak.

[257]

If roots can be identified, so too can prefixes and suffixes. The prefix precedes that part of the word recognised as its stem: e.g. hupo- ‘under’ in Greek hupo-zeúgnūmi (‘bring under (the yoke)’). The suffix is the element added to the root in order to form a stem (e.g. Greek zeug-mat-), or to one stem in order to form a second-grade stem (e.g. zeugmat-io-).¹ This element, as already noted above, may be represented by zero, just as an ending may. Identifying the suffix is thus simply another facet of the analysis of stems.

Sometimes the suffix has a concrete meaning, a semantic² value, as in zeuk-tēr-, where -tēr- designates the agent or instigator of the action (‘one who, or that which (yokes)’). Sometimes it has a purely grammatical function, as in zeúg-nū-(mi), where nū marks the notion of present tense. The prefix may also play either role, although it is rare for it to have a grammatical function in Indo-European languages, as it does in the ge- of the German past participle (ge-setzt from setzen ‘to put, place’), or the perfective prefixes of Slavic (Russian nāpisát’, etc.).

The prefix also differs from the suffix in another respect, which is not universal but very common nevertheless: namely, it is more easily identifiable because it stands out more clearly from the word as a whole. This has to do with the nature of the prefix as a linguistic element: for in most cases when a prefix is removed from a form, what is left over itself appears as a complete word (e.g. in French recommencer (‘begin again’) vs. commencer (‘begin’), indigne (‘worthy’) vs. digne (‘worthy’), maladroit (‘unskilful’) vs. adroit (‘skilful’), contrepoids (‘counterweight’) vs. poids (‘weight’). This feature is even more striking in Latin, Greek and German. It should be added that a prefix may often function as an independent word: e.g. French contre, mal, avant, sur, German unter, vor, Greek katá, pró.¹ But this is not the case with suffixes. If a suffix is removed, what is left is an incomplete word: e.g. French organisation – organis-, German Trennung – trenn-, Greek zeûgma – zeug-.² Furthermore, the suffix has no independent existence as a form.

[258]

Thus it comes about that the stem usually has its starting point clearly marked out. Before there is any comparison with other forms, the speaker knows where to place the boundary between the prefix and what follows it. But it is not the same at the end of a word. No boundary emerges except as the result of comparison between forms having the same stem or the same suffix, and such comparisons will give different results depending on the kinds of forms compared.

From the point of view of subjective analysis, suffixes and stems exist only in virtue of syntagmatic and associative oppositions. It is possible in some cases to find a formative element and a stem in two different parts of a word, whatever they may be, provided they give rise to an opposition. In the Latin accusative dictātōrem (‘dictator’), for instance, comparison with accusatives like consul-em (‘consul’), ped-em (‘foot’), etc. suggests a stem dictātōr-: whereas comparison with lic-tō-rem (‘lictor’), scrip-tōrem (‘writer’) etc. suggests a stem dictā-, and comparison with pō-tātōrem (‘drinker’) and can-tātōrem (‘singer’) suggests a stem dic-. Given the right circumstances, the speaker may be led to segment forms in any or all of a number of conceivable ways, e.g. dicāt-ōrem on the model of am-ōrem (‘love’), ard-ōrem (‘flame’), or dict-ātōrem on the model of ōr-ātōrem (‘orator’), ar-ātōrem (‘ploughman’). As we know (cf. p. [233]), the results of these spontaneous analyses can be seen in analogical formations of every period. These formations allow us to identify the smaller units (roots, prefixes, suffixes, endings) which the language recognises, and the values it attaches to them.

C Etymology

[259]

Etymology is neither a separate discipline nor a part of evolutionary linguistics. It is only a special application of principles relating to synchronic and diachronic facts. It traces back the past history of words until it finds something that explains them.

When one speaks of the origin of a word and says that it ‘comes from’ another word, there are several things that may be meant. French sel (‘salt’) comes from Latin sal (‘salt’) by simply a change of sound. French labourer (‘to plough’) comes from Old French labourer (‘to work’) simply by a change of meaning. French couver (‘to hatch’) comes from Latin cubāre (‘to lie down’) by changes of both meaning and sound. Finally, to say that French pommier (‘apple tree’) comes from pomme (‘apple’) is to state a relationship of grammatical derivation. In the first three cases we are dealing with diachronic identities, whereas the fourth is based on a synchronic relation between various different terms. This latter is the most important part of etymological research, as is shown by everything which has been said above in connexion with analogy.

The etymology of Latin bonus (‘good’) is not established simply by the discovery of an earlier form dvenos. But if one finds that Latin bis (‘twice’) goes back to dvis, thus uncovering a connexion with duo (‘two’), that is what may be called an etymological operation. The same applies to tracing French oiseau (‘bird’) back to Latin avicellus (‘little bird’), which brings to light the connexion between the French word and the Latin word for ‘bird’ (avis).

Etymology is thus first and foremost the explanation of words by investigating their connexions with other words. Explaining means relating to terms already known. In linguistics, to explain a word is to relate it to other words: for there are no necessary relations between sound and meaning. (Cf. p. [100] on the arbitrary nature of the linguistic sign.)

[260]

Etymology is not satisfied with explaining isolated words. It traces the history of word families, and also of formative elements (prefixes, suffixes, etc.).

Like static and evolutionary linguistics, etymology describes facts: but the description is not methodical, since it follows no particular direction. In investigating a word, etymology may take its information from phonetics, morphology, semantics, etc. as the need arises. To achieve its aims, etymology makes use of all the means which linguistics makes available, but does not scrutinise the nature of the processes it is obliged to engage in.

Notes

¹ Saussure did not deal with the question of compound words, at least from a synchronic viewpoint. Consequently, judgment must be suspended on this topic. What is clear enough is that the diachronic distinction earlier drawn between compounds and agglutinated forms cannot simply be carried over here, where one is dealing with the analysis of a linguistic state. It need hardly be added that this discussion of smaller units is not intended to resolve the more ticklish question raised on pp. [147] and [154] concerning the definition of the word as a unit. (Editorial note)

[254]

¹ In the example previously given (p. [254]), zeug- itself was described as a ‘second-grade’ stem. The discussion throughout this section bears witness to a failure to distinguish consistently between the type of relationship which links a root to the word forms based upon that root, on the one hand, and the quite different type of relationship which links the stem of any given word form with its ending, on the other. (Translator’s note)

² The distinction here between what is ‘semantic’ and what is ‘purely grammatical’ is unfortunate in view of the sense given elsewhere in the text to the term ‘grammatical’. Saussurean linguistics, as expounded in the Cours, lays no foundation for a distinction between meanings like ‘instigator of action’ and meanings like ‘present tense’. On the contrary, it is difficult to see how such a distinction could be validated other than by reference to strictly non-linguistic factors. (Translator’s note)

¹ English examples would be in, over, under. (Translator’s note)

² English examples would be eld-est, wid-th, burgl-ar. But English also provides counter instances. (Translator’s note)